Preprocessor.h revision b8db7cd9ac05c522855631670ec2e97255384f5a
1//===--- Preprocessor.h - C Language Family Preprocessor --------*- C++ -*-===//
2//
3//                     The LLVM Compiler Infrastructure
4//
5// This file is distributed under the University of Illinois Open Source
6// License. See LICENSE.TXT for details.
7//
8//===----------------------------------------------------------------------===//
9//
10//  This file defines the Preprocessor interface.
11//
12//===----------------------------------------------------------------------===//
13
14#ifndef LLVM_CLANG_LEX_PREPROCESSOR_H
15#define LLVM_CLANG_LEX_PREPROCESSOR_H
16
17#include "clang/Lex/MacroInfo.h"
18#include "clang/Lex/Lexer.h"
19#include "clang/Lex/PTHLexer.h"
20#include "clang/Lex/PPCallbacks.h"
21#include "clang/Lex/TokenLexer.h"
22#include "clang/Lex/PTHManager.h"
23#include "clang/Basic/Builtins.h"
24#include "clang/Basic/Diagnostic.h"
25#include "clang/Basic/IdentifierTable.h"
26#include "clang/Basic/SourceLocation.h"
27#include "llvm/ADT/DenseMap.h"
28#include "llvm/ADT/IntrusiveRefCntPtr.h"
29#include "llvm/ADT/SmallPtrSet.h"
30#include "llvm/ADT/OwningPtr.h"
31#include "llvm/ADT/SmallVector.h"
32#include "llvm/ADT/ArrayRef.h"
33#include "llvm/Support/Allocator.h"
34#include <vector>
35
36namespace clang {
37
38class SourceManager;
39class ExternalPreprocessorSource;
40class FileManager;
41class FileEntry;
42class HeaderSearch;
43class PragmaNamespace;
44class PragmaHandler;
45class CommentHandler;
46class ScratchBuffer;
47class TargetInfo;
48class PPCallbacks;
49class CodeCompletionHandler;
50class DirectoryLookup;
51class PreprocessingRecord;
52class ModuleLoader;
53
54/// Preprocessor - This object engages in a tight little dance with the lexer to
55/// efficiently preprocess tokens.  Lexers know only about tokens within a
56/// single source file, and don't know anything about preprocessor-level issues
57/// like the #include stack, token expansion, etc.
58///
59class Preprocessor : public llvm::RefCountedBase<Preprocessor> {
60  Diagnostic        *Diags;
61  LangOptions       &Features;
62  const TargetInfo  *Target;
63  FileManager       &FileMgr;
64  SourceManager     &SourceMgr;
65  ScratchBuffer     *ScratchBuf;
66  HeaderSearch      &HeaderInfo;
67  ModuleLoader      &TheModuleLoader;
68
69  /// \brief External source of macros.
70  ExternalPreprocessorSource *ExternalSource;
71
72
73  /// PTH - An optional PTHManager object used for getting tokens from
74  ///  a token cache rather than lexing the original source file.
75  llvm::OwningPtr<PTHManager> PTH;
76
77  /// BP - A BumpPtrAllocator object used to quickly allocate and release
78  ///  objects internal to the Preprocessor.
79  llvm::BumpPtrAllocator BP;
80
81  /// Identifiers for builtin macros and other builtins.
82  IdentifierInfo *Ident__LINE__, *Ident__FILE__;   // __LINE__, __FILE__
83  IdentifierInfo *Ident__DATE__, *Ident__TIME__;   // __DATE__, __TIME__
84  IdentifierInfo *Ident__INCLUDE_LEVEL__;          // __INCLUDE_LEVEL__
85  IdentifierInfo *Ident__BASE_FILE__;              // __BASE_FILE__
86  IdentifierInfo *Ident__TIMESTAMP__;              // __TIMESTAMP__
87  IdentifierInfo *Ident__COUNTER__;                // __COUNTER__
88  IdentifierInfo *Ident_Pragma, *Ident__pragma;    // _Pragma, __pragma
89  IdentifierInfo *Ident__VA_ARGS__;                // __VA_ARGS__
90  IdentifierInfo *Ident__has_feature;              // __has_feature
91  IdentifierInfo *Ident__has_extension;            // __has_extension
92  IdentifierInfo *Ident__has_builtin;              // __has_builtin
93  IdentifierInfo *Ident__has_attribute;            // __has_attribute
94  IdentifierInfo *Ident__has_include;              // __has_include
95  IdentifierInfo *Ident__has_include_next;         // __has_include_next
96
97  SourceLocation DATELoc, TIMELoc;
98  unsigned CounterValue;  // Next __COUNTER__ value.
99
100  enum {
101    /// MaxIncludeStackDepth - Maximum depth of #includes.
102    MaxAllowedIncludeStackDepth = 200
103  };
104
105  // State that is set before the preprocessor begins.
106  bool KeepComments : 1;
107  bool KeepMacroComments : 1;
108  bool SuppressIncludeNotFoundError : 1;
109
110  // State that changes while the preprocessor runs:
111  bool InMacroArgs : 1;            // True if parsing fn macro invocation args.
112
113  /// Whether the preprocessor owns the header search object.
114  bool OwnsHeaderSearch : 1;
115
116  /// DisableMacroExpansion - True if macro expansion is disabled.
117  bool DisableMacroExpansion : 1;
118
119  /// \brief Whether we have already loaded macros from the external source.
120  mutable bool ReadMacrosFromExternalSource : 1;
121
122  /// Identifiers - This is mapping/lookup information for all identifiers in
123  /// the program, including program keywords.
124  mutable IdentifierTable Identifiers;
125
126  /// Selectors - This table contains all the selectors in the program. Unlike
127  /// IdentifierTable above, this table *isn't* populated by the preprocessor.
128  /// It is declared/expanded here because it's role/lifetime is
129  /// conceptually similar the IdentifierTable. In addition, the current control
130  /// flow (in clang::ParseAST()), make it convenient to put here.
131  /// FIXME: Make sure the lifetime of Identifiers/Selectors *isn't* tied to
132  /// the lifetime of the preprocessor.
133  SelectorTable Selectors;
134
135  /// BuiltinInfo - Information about builtins.
136  Builtin::Context BuiltinInfo;
137
138  /// PragmaHandlers - This tracks all of the pragmas that the client registered
139  /// with this preprocessor.
140  PragmaNamespace *PragmaHandlers;
141
142  /// \brief Tracks all of the comment handlers that the client registered
143  /// with this preprocessor.
144  std::vector<CommentHandler *> CommentHandlers;
145
146  /// \brief The code-completion handler.
147  CodeCompletionHandler *CodeComplete;
148
149  /// \brief The file that we're performing code-completion for, if any.
150  const FileEntry *CodeCompletionFile;
151
152  /// \brief The offset in file for the code-completion point.
153  unsigned CodeCompletionOffset;
154
155  /// \brief The location for the code-completion point. This gets instantiated
156  /// when the CodeCompletionFile gets #include'ed for preprocessing.
157  SourceLocation CodeCompletionLoc;
158
159  /// \brief The start location for the file of the code-completion point.
160  /// This gets instantiated when the CodeCompletionFile gets #include'ed
161  /// for preprocessing.
162  SourceLocation CodeCompletionFileLoc;
163
164  /// \brief The source location of the __import_module__ keyword we just
165  /// lexed, if any.
166  SourceLocation ModuleImportLoc;
167
168  /// \brief True if we hit the code-completion point.
169  bool CodeCompletionReached;
170
171  /// \brief The number of bytes that we will initially skip when entering the
172  /// main file, which is used when loading a precompiled preamble, along
173  /// with a flag that indicates whether skipping this number of bytes will
174  /// place the lexer at the start of a line.
175  std::pair<unsigned, bool> SkipMainFilePreamble;
176
177  /// CurLexer - This is the current top of the stack that we're lexing from if
178  /// not expanding a macro and we are lexing directly from source code.
179  ///  Only one of CurLexer, CurPTHLexer, or CurTokenLexer will be non-null.
180  llvm::OwningPtr<Lexer> CurLexer;
181
182  /// CurPTHLexer - This is the current top of stack that we're lexing from if
183  ///  not expanding from a macro and we are lexing from a PTH cache.
184  ///  Only one of CurLexer, CurPTHLexer, or CurTokenLexer will be non-null.
185  llvm::OwningPtr<PTHLexer> CurPTHLexer;
186
187  /// CurPPLexer - This is the current top of the stack what we're lexing from
188  ///  if not expanding a macro.  This is an alias for either CurLexer or
189  ///  CurPTHLexer.
190  PreprocessorLexer *CurPPLexer;
191
192  /// CurLookup - The DirectoryLookup structure used to find the current
193  /// FileEntry, if CurLexer is non-null and if applicable.  This allows us to
194  /// implement #include_next and find directory-specific properties.
195  const DirectoryLookup *CurDirLookup;
196
197  /// CurTokenLexer - This is the current macro we are expanding, if we are
198  /// expanding a macro.  One of CurLexer and CurTokenLexer must be null.
199  llvm::OwningPtr<TokenLexer> CurTokenLexer;
200
201  /// \brief The kind of lexer we're currently working with.
202  enum CurLexerKind {
203    CLK_Lexer,
204    CLK_PTHLexer,
205    CLK_TokenLexer,
206    CLK_CachingLexer,
207    CLK_LexAfterModuleImport
208  } CurLexerKind;
209
210  /// IncludeMacroStack - This keeps track of the stack of files currently
211  /// #included, and macros currently being expanded from, not counting
212  /// CurLexer/CurTokenLexer.
213  struct IncludeStackInfo {
214    enum CurLexerKind     CurLexerKind;
215    Lexer                 *TheLexer;
216    PTHLexer              *ThePTHLexer;
217    PreprocessorLexer     *ThePPLexer;
218    TokenLexer            *TheTokenLexer;
219    const DirectoryLookup *TheDirLookup;
220
221    IncludeStackInfo(enum CurLexerKind K, Lexer *L, PTHLexer* P,
222                     PreprocessorLexer* PPL,
223                     TokenLexer* TL, const DirectoryLookup *D)
224      : CurLexerKind(K), TheLexer(L), ThePTHLexer(P), ThePPLexer(PPL),
225        TheTokenLexer(TL), TheDirLookup(D) {}
226  };
227  std::vector<IncludeStackInfo> IncludeMacroStack;
228
229  /// Callbacks - These are actions invoked when some preprocessor activity is
230  /// encountered (e.g. a file is #included, etc).
231  PPCallbacks *Callbacks;
232
233  /// Macros - For each IdentifierInfo with 'HasMacro' set, we keep a mapping
234  /// to the actual definition of the macro.
235  llvm::DenseMap<IdentifierInfo*, MacroInfo*> Macros;
236
237  /// \brief Macros that we want to warn because they are not used at the end
238  /// of the translation unit; we store just their SourceLocations instead
239  /// something like MacroInfo*. The benefit of this is that when we are
240  /// deserializing from PCH, we don't need to deserialize identifier & macros
241  /// just so that we can report that they are unused, we just warn using
242  /// the SourceLocations of this set (that will be filled by the ASTReader).
243  /// We are using SmallPtrSet instead of a vector for faster removal.
244  typedef llvm::SmallPtrSet<SourceLocation, 32> WarnUnusedMacroLocsTy;
245  WarnUnusedMacroLocsTy WarnUnusedMacroLocs;
246
247  /// MacroArgCache - This is a "freelist" of MacroArg objects that can be
248  /// reused for quick allocation.
249  MacroArgs *MacroArgCache;
250  friend class MacroArgs;
251
252  /// PragmaPushMacroInfo - For each IdentifierInfo used in a #pragma
253  /// push_macro directive, we keep a MacroInfo stack used to restore
254  /// previous macro value.
255  llvm::DenseMap<IdentifierInfo*, std::vector<MacroInfo*> > PragmaPushMacroInfo;
256
257  // Various statistics we track for performance analysis.
258  unsigned NumDirectives, NumIncluded, NumDefined, NumUndefined, NumPragma;
259  unsigned NumIf, NumElse, NumEndif;
260  unsigned NumEnteredSourceFiles, MaxIncludeStackDepth;
261  unsigned NumMacroExpanded, NumFnMacroExpanded, NumBuiltinMacroExpanded;
262  unsigned NumFastMacroExpanded, NumTokenPaste, NumFastTokenPaste;
263  unsigned NumSkipped;
264
265  /// Predefines - This string is the predefined macros that preprocessor
266  /// should use from the command line etc.
267  std::string Predefines;
268
269  /// TokenLexerCache - Cache macro expanders to reduce malloc traffic.
270  enum { TokenLexerCacheSize = 8 };
271  unsigned NumCachedTokenLexers;
272  TokenLexer *TokenLexerCache[TokenLexerCacheSize];
273
274  /// \brief Keeps macro expanded tokens for TokenLexers.
275  //
276  /// Works like a stack; a TokenLexer adds the macro expanded tokens that is
277  /// going to lex in the cache and when it finishes the tokens are removed
278  /// from the end of the cache.
279  SmallVector<Token, 16> MacroExpandedTokens;
280  std::vector<std::pair<TokenLexer *, size_t> > MacroExpandingLexersStack;
281
282  /// \brief A record of the macro definitions and expansions that
283  /// occurred during preprocessing.
284  ///
285  /// This is an optional side structure that can be enabled with
286  /// \c createPreprocessingRecord() prior to preprocessing.
287  PreprocessingRecord *Record;
288
289private:  // Cached tokens state.
290  typedef SmallVector<Token, 1> CachedTokensTy;
291
292  /// CachedTokens - Cached tokens are stored here when we do backtracking or
293  /// lookahead. They are "lexed" by the CachingLex() method.
294  CachedTokensTy CachedTokens;
295
296  /// CachedLexPos - The position of the cached token that CachingLex() should
297  /// "lex" next. If it points beyond the CachedTokens vector, it means that
298  /// a normal Lex() should be invoked.
299  CachedTokensTy::size_type CachedLexPos;
300
301  /// BacktrackPositions - Stack of backtrack positions, allowing nested
302  /// backtracks. The EnableBacktrackAtThisPos() method pushes a position to
303  /// indicate where CachedLexPos should be set when the BackTrack() method is
304  /// invoked (at which point the last position is popped).
305  std::vector<CachedTokensTy::size_type> BacktrackPositions;
306
307  struct MacroInfoChain {
308    MacroInfo MI;
309    MacroInfoChain *Next;
310    MacroInfoChain *Prev;
311  };
312
313  /// MacroInfos are managed as a chain for easy disposal.  This is the head
314  /// of that list.
315  MacroInfoChain *MIChainHead;
316
317  /// MICache - A "freelist" of MacroInfo objects that can be reused for quick
318  /// allocation.
319  MacroInfoChain *MICache;
320
321  MacroInfo *getInfoForMacro(IdentifierInfo *II) const;
322
323public:
324  Preprocessor(Diagnostic &diags, LangOptions &opts,
325               const TargetInfo *target,
326               SourceManager &SM, HeaderSearch &Headers,
327               ModuleLoader &TheModuleLoader,
328               IdentifierInfoLookup *IILookup = 0,
329               bool OwnsHeaderSearch = false,
330               bool DelayInitialization = false);
331
332  ~Preprocessor();
333
334  /// \brief Initialize the preprocessor, if the constructor did not already
335  /// perform the initialization.
336  ///
337  /// \param Target Information about the target.
338  void Initialize(const TargetInfo &Target);
339
340  Diagnostic &getDiagnostics() const { return *Diags; }
341  void setDiagnostics(Diagnostic &D) { Diags = &D; }
342
343  const LangOptions &getLangOptions() const { return Features; }
344  const TargetInfo &getTargetInfo() const { return *Target; }
345  FileManager &getFileManager() const { return FileMgr; }
346  SourceManager &getSourceManager() const { return SourceMgr; }
347  HeaderSearch &getHeaderSearchInfo() const { return HeaderInfo; }
348
349  IdentifierTable &getIdentifierTable() { return Identifiers; }
350  SelectorTable &getSelectorTable() { return Selectors; }
351  Builtin::Context &getBuiltinInfo() { return BuiltinInfo; }
352  llvm::BumpPtrAllocator &getPreprocessorAllocator() { return BP; }
353
354  void setPTHManager(PTHManager* pm);
355
356  PTHManager *getPTHManager() { return PTH.get(); }
357
358  void setExternalSource(ExternalPreprocessorSource *Source) {
359    ExternalSource = Source;
360  }
361
362  ExternalPreprocessorSource *getExternalSource() const {
363    return ExternalSource;
364  }
365
366  /// \brief Retrieve the module loader associated with this preprocessor.
367  ModuleLoader &getModuleLoader() const { return TheModuleLoader; }
368
369  /// SetCommentRetentionState - Control whether or not the preprocessor retains
370  /// comments in output.
371  void SetCommentRetentionState(bool KeepComments, bool KeepMacroComments) {
372    this->KeepComments = KeepComments | KeepMacroComments;
373    this->KeepMacroComments = KeepMacroComments;
374  }
375
376  bool getCommentRetentionState() const { return KeepComments; }
377
378  void SetSuppressIncludeNotFoundError(bool Suppress) {
379    SuppressIncludeNotFoundError = Suppress;
380  }
381
382  bool GetSuppressIncludeNotFoundError() {
383    return SuppressIncludeNotFoundError;
384  }
385
386  /// isCurrentLexer - Return true if we are lexing directly from the specified
387  /// lexer.
388  bool isCurrentLexer(const PreprocessorLexer *L) const {
389    return CurPPLexer == L;
390  }
391
392  /// getCurrentLexer - Return the current lexer being lexed from.  Note
393  /// that this ignores any potentially active macro expansions and _Pragma
394  /// expansions going on at the time.
395  PreprocessorLexer *getCurrentLexer() const { return CurPPLexer; }
396
397  /// getCurrentFileLexer - Return the current file lexer being lexed from.
398  /// Note that this ignores any potentially active macro expansions and _Pragma
399  /// expansions going on at the time.
400  PreprocessorLexer *getCurrentFileLexer() const;
401
402  /// getPPCallbacks/addPPCallbacks - Accessors for preprocessor callbacks.
403  /// Note that this class takes ownership of any PPCallbacks object given to
404  /// it.
405  PPCallbacks *getPPCallbacks() const { return Callbacks; }
406  void addPPCallbacks(PPCallbacks *C) {
407    if (Callbacks)
408      C = new PPChainedCallbacks(C, Callbacks);
409    Callbacks = C;
410  }
411
412  /// getMacroInfo - Given an identifier, return the MacroInfo it is #defined to
413  /// or null if it isn't #define'd.
414  MacroInfo *getMacroInfo(IdentifierInfo *II) const {
415    if (!II->hasMacroDefinition())
416      return 0;
417
418    return getInfoForMacro(II);
419  }
420
421  /// setMacroInfo - Specify a macro for this identifier.
422  ///
423  void setMacroInfo(IdentifierInfo *II, MacroInfo *MI);
424
425  /// macro_iterator/macro_begin/macro_end - This allows you to walk the current
426  /// state of the macro table.  This visits every currently-defined macro.
427  typedef llvm::DenseMap<IdentifierInfo*,
428                         MacroInfo*>::const_iterator macro_iterator;
429  macro_iterator macro_begin(bool IncludeExternalMacros = true) const;
430  macro_iterator macro_end(bool IncludeExternalMacros = true) const;
431
432  const std::string &getPredefines() const { return Predefines; }
433  /// setPredefines - Set the predefines for this Preprocessor.  These
434  /// predefines are automatically injected when parsing the main file.
435  void setPredefines(const char *P) { Predefines = P; }
436  void setPredefines(const std::string &P) { Predefines = P; }
437
438  /// getIdentifierInfo - Return information about the specified preprocessor
439  /// identifier token.  The version of this method that takes two character
440  /// pointers is preferred unless the identifier is already available as a
441  /// string (this avoids allocation and copying of memory to construct an
442  /// std::string).
443  IdentifierInfo *getIdentifierInfo(StringRef Name) const {
444    return &Identifiers.get(Name);
445  }
446
447  /// AddPragmaHandler - Add the specified pragma handler to the preprocessor.
448  /// If 'Namespace' is non-null, then it is a token required to exist on the
449  /// pragma line before the pragma string starts, e.g. "STDC" or "GCC".
450  void AddPragmaHandler(StringRef Namespace, PragmaHandler *Handler);
451  void AddPragmaHandler(PragmaHandler *Handler) {
452    AddPragmaHandler(StringRef(), Handler);
453  }
454
455  /// RemovePragmaHandler - Remove the specific pragma handler from
456  /// the preprocessor. If \arg Namespace is non-null, then it should
457  /// be the namespace that \arg Handler was added to. It is an error
458  /// to remove a handler that has not been registered.
459  void RemovePragmaHandler(StringRef Namespace, PragmaHandler *Handler);
460  void RemovePragmaHandler(PragmaHandler *Handler) {
461    RemovePragmaHandler(StringRef(), Handler);
462  }
463
464  /// \brief Add the specified comment handler to the preprocessor.
465  void AddCommentHandler(CommentHandler *Handler);
466
467  /// \brief Remove the specified comment handler.
468  ///
469  /// It is an error to remove a handler that has not been registered.
470  void RemoveCommentHandler(CommentHandler *Handler);
471
472  /// \brief Set the code completion handler to the given object.
473  void setCodeCompletionHandler(CodeCompletionHandler &Handler) {
474    CodeComplete = &Handler;
475  }
476
477  /// \brief Retrieve the current code-completion handler.
478  CodeCompletionHandler *getCodeCompletionHandler() const {
479    return CodeComplete;
480  }
481
482  /// \brief Clear out the code completion handler.
483  void clearCodeCompletionHandler() {
484    CodeComplete = 0;
485  }
486
487  /// \brief Hook used by the lexer to invoke the "natural language" code
488  /// completion point.
489  void CodeCompleteNaturalLanguage();
490
491  /// \brief Retrieve the preprocessing record, or NULL if there is no
492  /// preprocessing record.
493  PreprocessingRecord *getPreprocessingRecord() const { return Record; }
494
495  /// \brief Create a new preprocessing record, which will keep track of
496  /// all macro expansions, macro definitions, etc.
497  void createPreprocessingRecord(bool IncludeNestedMacroExpansions);
498
499  /// EnterMainSourceFile - Enter the specified FileID as the main source file,
500  /// which implicitly adds the builtin defines etc.
501  void EnterMainSourceFile();
502
503  /// EndSourceFile - Inform the preprocessor callbacks that processing is
504  /// complete.
505  void EndSourceFile();
506
507  /// EnterSourceFile - Add a source file to the top of the include stack and
508  /// start lexing tokens from it instead of the current buffer.  Emit an error
509  /// and don't enter the file on error.
510  void EnterSourceFile(FileID CurFileID, const DirectoryLookup *Dir,
511                       SourceLocation Loc);
512
513  /// EnterMacro - Add a Macro to the top of the include stack and start lexing
514  /// tokens from it instead of the current buffer.  Args specifies the
515  /// tokens input to a function-like macro.
516  ///
517  /// ILEnd specifies the location of the ')' for a function-like macro or the
518  /// identifier for an object-like macro.
519  void EnterMacro(Token &Identifier, SourceLocation ILEnd, MacroArgs *Args);
520
521  /// EnterTokenStream - Add a "macro" context to the top of the include stack,
522  /// which will cause the lexer to start returning the specified tokens.
523  ///
524  /// If DisableMacroExpansion is true, tokens lexed from the token stream will
525  /// not be subject to further macro expansion.  Otherwise, these tokens will
526  /// be re-macro-expanded when/if expansion is enabled.
527  ///
528  /// If OwnsTokens is false, this method assumes that the specified stream of
529  /// tokens has a permanent owner somewhere, so they do not need to be copied.
530  /// If it is true, it assumes the array of tokens is allocated with new[] and
531  /// must be freed.
532  ///
533  void EnterTokenStream(const Token *Toks, unsigned NumToks,
534                        bool DisableMacroExpansion, bool OwnsTokens);
535
536  /// RemoveTopOfLexerStack - Pop the current lexer/macro exp off the top of the
537  /// lexer stack.  This should only be used in situations where the current
538  /// state of the top-of-stack lexer is known.
539  void RemoveTopOfLexerStack();
540
541  /// EnableBacktrackAtThisPos - From the point that this method is called, and
542  /// until CommitBacktrackedTokens() or Backtrack() is called, the Preprocessor
543  /// keeps track of the lexed tokens so that a subsequent Backtrack() call will
544  /// make the Preprocessor re-lex the same tokens.
545  ///
546  /// Nested backtracks are allowed, meaning that EnableBacktrackAtThisPos can
547  /// be called multiple times and CommitBacktrackedTokens/Backtrack calls will
548  /// be combined with the EnableBacktrackAtThisPos calls in reverse order.
549  ///
550  /// NOTE: *DO NOT* forget to call either CommitBacktrackedTokens or Backtrack
551  /// at some point after EnableBacktrackAtThisPos. If you don't, caching of
552  /// tokens will continue indefinitely.
553  ///
554  void EnableBacktrackAtThisPos();
555
556  /// CommitBacktrackedTokens - Disable the last EnableBacktrackAtThisPos call.
557  void CommitBacktrackedTokens();
558
559  /// Backtrack - Make Preprocessor re-lex the tokens that were lexed since
560  /// EnableBacktrackAtThisPos() was previously called.
561  void Backtrack();
562
563  /// isBacktrackEnabled - True if EnableBacktrackAtThisPos() was called and
564  /// caching of tokens is on.
565  bool isBacktrackEnabled() const { return !BacktrackPositions.empty(); }
566
567  /// Lex - To lex a token from the preprocessor, just pull a token from the
568  /// current lexer or macro object.
569  void Lex(Token &Result) {
570    switch (CurLexerKind) {
571    case CLK_Lexer: CurLexer->Lex(Result); break;
572    case CLK_PTHLexer: CurPTHLexer->Lex(Result); break;
573    case CLK_TokenLexer: CurTokenLexer->Lex(Result); break;
574    case CLK_CachingLexer: CachingLex(Result); break;
575    case CLK_LexAfterModuleImport: LexAfterModuleImport(Result); break;
576    }
577  }
578
579  void LexAfterModuleImport(Token &Result);
580
581  /// LexNonComment - Lex a token.  If it's a comment, keep lexing until we get
582  /// something not a comment.  This is useful in -E -C mode where comments
583  /// would foul up preprocessor directive handling.
584  void LexNonComment(Token &Result) {
585    do
586      Lex(Result);
587    while (Result.getKind() == tok::comment);
588  }
589
590  /// LexUnexpandedToken - This is just like Lex, but this disables macro
591  /// expansion of identifier tokens.
592  void LexUnexpandedToken(Token &Result) {
593    // Disable macro expansion.
594    bool OldVal = DisableMacroExpansion;
595    DisableMacroExpansion = true;
596    // Lex the token.
597    Lex(Result);
598
599    // Reenable it.
600    DisableMacroExpansion = OldVal;
601  }
602
603  /// LexUnexpandedNonComment - Like LexNonComment, but this disables macro
604  /// expansion of identifier tokens.
605  void LexUnexpandedNonComment(Token &Result) {
606    do
607      LexUnexpandedToken(Result);
608    while (Result.getKind() == tok::comment);
609  }
610
611  /// LookAhead - This peeks ahead N tokens and returns that token without
612  /// consuming any tokens.  LookAhead(0) returns the next token that would be
613  /// returned by Lex(), LookAhead(1) returns the token after it, etc.  This
614  /// returns normal tokens after phase 5.  As such, it is equivalent to using
615  /// 'Lex', not 'LexUnexpandedToken'.
616  const Token &LookAhead(unsigned N) {
617    if (CachedLexPos + N < CachedTokens.size())
618      return CachedTokens[CachedLexPos+N];
619    else
620      return PeekAhead(N+1);
621  }
622
623  /// RevertCachedTokens - When backtracking is enabled and tokens are cached,
624  /// this allows to revert a specific number of tokens.
625  /// Note that the number of tokens being reverted should be up to the last
626  /// backtrack position, not more.
627  void RevertCachedTokens(unsigned N) {
628    assert(isBacktrackEnabled() &&
629           "Should only be called when tokens are cached for backtracking");
630    assert(signed(CachedLexPos) - signed(N) >= signed(BacktrackPositions.back())
631         && "Should revert tokens up to the last backtrack position, not more");
632    assert(signed(CachedLexPos) - signed(N) >= 0 &&
633           "Corrupted backtrack positions ?");
634    CachedLexPos -= N;
635  }
636
637  /// EnterToken - Enters a token in the token stream to be lexed next. If
638  /// BackTrack() is called afterwards, the token will remain at the insertion
639  /// point.
640  void EnterToken(const Token &Tok) {
641    EnterCachingLexMode();
642    CachedTokens.insert(CachedTokens.begin()+CachedLexPos, Tok);
643  }
644
645  /// AnnotateCachedTokens - We notify the Preprocessor that if it is caching
646  /// tokens (because backtrack is enabled) it should replace the most recent
647  /// cached tokens with the given annotation token. This function has no effect
648  /// if backtracking is not enabled.
649  ///
650  /// Note that the use of this function is just for optimization; so that the
651  /// cached tokens doesn't get re-parsed and re-resolved after a backtrack is
652  /// invoked.
653  void AnnotateCachedTokens(const Token &Tok) {
654    assert(Tok.isAnnotation() && "Expected annotation token");
655    if (CachedLexPos != 0 && isBacktrackEnabled())
656      AnnotatePreviousCachedTokens(Tok);
657  }
658
659  /// \brief Replace the last token with an annotation token.
660  ///
661  /// Like AnnotateCachedTokens(), this routine replaces an
662  /// already-parsed (and resolved) token with an annotation
663  /// token. However, this routine only replaces the last token with
664  /// the annotation token; it does not affect any other cached
665  /// tokens. This function has no effect if backtracking is not
666  /// enabled.
667  void ReplaceLastTokenWithAnnotation(const Token &Tok) {
668    assert(Tok.isAnnotation() && "Expected annotation token");
669    if (CachedLexPos != 0 && isBacktrackEnabled())
670      CachedTokens[CachedLexPos-1] = Tok;
671  }
672
673  /// \brief Specify the point at which code-completion will be performed.
674  ///
675  /// \param File the file in which code completion should occur. If
676  /// this file is included multiple times, code-completion will
677  /// perform completion the first time it is included. If NULL, this
678  /// function clears out the code-completion point.
679  ///
680  /// \param Line the line at which code completion should occur
681  /// (1-based).
682  ///
683  /// \param Column the column at which code completion should occur
684  /// (1-based).
685  ///
686  /// \returns true if an error occurred, false otherwise.
687  bool SetCodeCompletionPoint(const FileEntry *File,
688                              unsigned Line, unsigned Column);
689
690  /// \brief Determine if we are performing code completion.
691  bool isCodeCompletionEnabled() const { return CodeCompletionFile != 0; }
692
693  /// \brief Returns the location of the code-completion point.
694  /// Returns an invalid location if code-completion is not enabled or the file
695  /// containing the code-completion point has not been lexed yet.
696  SourceLocation getCodeCompletionLoc() const { return CodeCompletionLoc; }
697
698  /// \brief Returns the start location of the file of code-completion point.
699  /// Returns an invalid location if code-completion is not enabled or the file
700  /// containing the code-completion point has not been lexed yet.
701  SourceLocation getCodeCompletionFileLoc() const {
702    return CodeCompletionFileLoc;
703  }
704
705  /// \brief Returns true if code-completion is enabled and we have hit the
706  /// code-completion point.
707  bool isCodeCompletionReached() const { return CodeCompletionReached; }
708
709  /// \brief Note that we hit the code-completion point.
710  void setCodeCompletionReached() {
711    assert(isCodeCompletionEnabled() && "Code-completion not enabled!");
712    CodeCompletionReached = true;
713    // Silence any diagnostics that occur after we hit the code-completion.
714    getDiagnostics().setSuppressAllDiagnostics(true);
715  }
716
717  /// \brief Instruct the preprocessor to skip part of the main
718  /// the main source file.
719  ///
720  /// \brief Bytes The number of bytes in the preamble to skip.
721  ///
722  /// \brief StartOfLine Whether skipping these bytes puts the lexer at the
723  /// start of a line.
724  void setSkipMainFilePreamble(unsigned Bytes, bool StartOfLine) {
725    SkipMainFilePreamble.first = Bytes;
726    SkipMainFilePreamble.second = StartOfLine;
727  }
728
729  /// Diag - Forwarding function for diagnostics.  This emits a diagnostic at
730  /// the specified Token's location, translating the token's start
731  /// position in the current buffer into a SourcePosition object for rendering.
732  DiagnosticBuilder Diag(SourceLocation Loc, unsigned DiagID) {
733    return Diags->Report(Loc, DiagID);
734  }
735
736  DiagnosticBuilder Diag(const Token &Tok, unsigned DiagID) {
737    return Diags->Report(Tok.getLocation(), DiagID);
738  }
739
740  /// getSpelling() - Return the 'spelling' of the token at the given
741  /// location; does not go up to the spelling location or down to the
742  /// expansion location.
743  ///
744  /// \param buffer A buffer which will be used only if the token requires
745  ///   "cleaning", e.g. if it contains trigraphs or escaped newlines
746  /// \param invalid If non-null, will be set \c true if an error occurs.
747  StringRef getSpelling(SourceLocation loc,
748                              SmallVectorImpl<char> &buffer,
749                              bool *invalid = 0) const {
750    return Lexer::getSpelling(loc, buffer, SourceMgr, Features, invalid);
751  }
752
753  /// getSpelling() - Return the 'spelling' of the Tok token.  The spelling of a
754  /// token is the characters used to represent the token in the source file
755  /// after trigraph expansion and escaped-newline folding.  In particular, this
756  /// wants to get the true, uncanonicalized, spelling of things like digraphs
757  /// UCNs, etc.
758  ///
759  /// \param Invalid If non-null, will be set \c true if an error occurs.
760  std::string getSpelling(const Token &Tok, bool *Invalid = 0) const {
761    return Lexer::getSpelling(Tok, SourceMgr, Features, Invalid);
762  }
763
764  /// getSpelling - This method is used to get the spelling of a token into a
765  /// preallocated buffer, instead of as an std::string.  The caller is required
766  /// to allocate enough space for the token, which is guaranteed to be at least
767  /// Tok.getLength() bytes long.  The length of the actual result is returned.
768  ///
769  /// Note that this method may do two possible things: it may either fill in
770  /// the buffer specified with characters, or it may *change the input pointer*
771  /// to point to a constant buffer with the data already in it (avoiding a
772  /// copy).  The caller is not allowed to modify the returned buffer pointer
773  /// if an internal buffer is returned.
774  unsigned getSpelling(const Token &Tok, const char *&Buffer,
775                       bool *Invalid = 0) const {
776    return Lexer::getSpelling(Tok, Buffer, SourceMgr, Features, Invalid);
777  }
778
779  /// getSpelling - This method is used to get the spelling of a token into a
780  /// SmallVector. Note that the returned StringRef may not point to the
781  /// supplied buffer if a copy can be avoided.
782  StringRef getSpelling(const Token &Tok,
783                              SmallVectorImpl<char> &Buffer,
784                              bool *Invalid = 0) const;
785
786  /// getSpellingOfSingleCharacterNumericConstant - Tok is a numeric constant
787  /// with length 1, return the character.
788  char getSpellingOfSingleCharacterNumericConstant(const Token &Tok,
789                                                   bool *Invalid = 0) const {
790    assert(Tok.is(tok::numeric_constant) &&
791           Tok.getLength() == 1 && "Called on unsupported token");
792    assert(!Tok.needsCleaning() && "Token can't need cleaning with length 1");
793
794    // If the token is carrying a literal data pointer, just use it.
795    if (const char *D = Tok.getLiteralData())
796      return *D;
797
798    // Otherwise, fall back on getCharacterData, which is slower, but always
799    // works.
800    return *SourceMgr.getCharacterData(Tok.getLocation(), Invalid);
801  }
802
803  /// CreateString - Plop the specified string into a scratch buffer and set the
804  /// specified token's location and length to it.  If specified, the source
805  /// location provides a location of the expansion point of the token.
806  void CreateString(const char *Buf, unsigned Len,
807                    Token &Tok, SourceLocation SourceLoc = SourceLocation());
808
809  /// \brief Computes the source location just past the end of the
810  /// token at this source location.
811  ///
812  /// This routine can be used to produce a source location that
813  /// points just past the end of the token referenced by \p Loc, and
814  /// is generally used when a diagnostic needs to point just after a
815  /// token where it expected something different that it received. If
816  /// the returned source location would not be meaningful (e.g., if
817  /// it points into a macro), this routine returns an invalid
818  /// source location.
819  ///
820  /// \param Offset an offset from the end of the token, where the source
821  /// location should refer to. The default offset (0) produces a source
822  /// location pointing just past the end of the token; an offset of 1 produces
823  /// a source location pointing to the last character in the token, etc.
824  SourceLocation getLocForEndOfToken(SourceLocation Loc, unsigned Offset = 0) {
825    return Lexer::getLocForEndOfToken(Loc, Offset, SourceMgr, Features);
826  }
827
828  /// \brief Returns true if the given MacroID location points at the first
829  /// token of the macro expansion.
830  bool isAtStartOfMacroExpansion(SourceLocation loc) const {
831    return Lexer::isAtStartOfMacroExpansion(loc, SourceMgr, Features);
832  }
833
834  /// \brief Returns true if the given MacroID location points at the last
835  /// token of the macro expansion.
836  bool isAtEndOfMacroExpansion(SourceLocation loc) const {
837    return Lexer::isAtEndOfMacroExpansion(loc, SourceMgr, Features);
838  }
839
840  /// DumpToken - Print the token to stderr, used for debugging.
841  ///
842  void DumpToken(const Token &Tok, bool DumpFlags = false) const;
843  void DumpLocation(SourceLocation Loc) const;
844  void DumpMacro(const MacroInfo &MI) const;
845
846  /// AdvanceToTokenCharacter - Given a location that specifies the start of a
847  /// token, return a new location that specifies a character within the token.
848  SourceLocation AdvanceToTokenCharacter(SourceLocation TokStart,
849                                         unsigned Char) const {
850    return Lexer::AdvanceToTokenCharacter(TokStart, Char, SourceMgr, Features);
851  }
852
853  /// IncrementPasteCounter - Increment the counters for the number of token
854  /// paste operations performed.  If fast was specified, this is a 'fast paste'
855  /// case we handled.
856  ///
857  void IncrementPasteCounter(bool isFast) {
858    if (isFast)
859      ++NumFastTokenPaste;
860    else
861      ++NumTokenPaste;
862  }
863
864  void PrintStats();
865
866  size_t getTotalMemory() const;
867
868  /// HandleMicrosoftCommentPaste - When the macro expander pastes together a
869  /// comment (/##/) in microsoft mode, this method handles updating the current
870  /// state, returning the token on the next source line.
871  void HandleMicrosoftCommentPaste(Token &Tok);
872
873  //===--------------------------------------------------------------------===//
874  // Preprocessor callback methods.  These are invoked by a lexer as various
875  // directives and events are found.
876
877  /// LookUpIdentifierInfo - Given a tok::raw_identifier token, look up the
878  /// identifier information for the token and install it into the token,
879  /// updating the token kind accordingly.
880  IdentifierInfo *LookUpIdentifierInfo(Token &Identifier) const;
881
882private:
883  llvm::DenseMap<IdentifierInfo*,unsigned> PoisonReasons;
884
885public:
886
887  // SetPoisonReason - Call this function to indicate the reason for
888  // poisoning an identifier. If that identifier is accessed while
889  // poisoned, then this reason will be used instead of the default
890  // "poisoned" diagnostic.
891  void SetPoisonReason(IdentifierInfo *II, unsigned DiagID);
892
893  // HandlePoisonedIdentifier - Display reason for poisoned
894  // identifier.
895  void HandlePoisonedIdentifier(Token & Tok);
896
897  void MaybeHandlePoisonedIdentifier(Token & Identifier) {
898    if(IdentifierInfo * II = Identifier.getIdentifierInfo()) {
899      if(II->isPoisoned()) {
900        HandlePoisonedIdentifier(Identifier);
901      }
902    }
903  }
904
905private:
906  /// Identifiers used for SEH handling in Borland. These are only
907  /// allowed in particular circumstances
908  IdentifierInfo *Ident__exception_code, *Ident___exception_code, *Ident_GetExceptionCode; // __except block
909  IdentifierInfo *Ident__exception_info, *Ident___exception_info, *Ident_GetExceptionInfo; // __except filter expression
910  IdentifierInfo *Ident__abnormal_termination, *Ident___abnormal_termination, *Ident_AbnormalTermination; // __finally
911public:
912  void PoisonSEHIdentifiers(bool Poison = true); // Borland
913
914  /// HandleIdentifier - This callback is invoked when the lexer reads an
915  /// identifier and has filled in the tokens IdentifierInfo member.  This
916  /// callback potentially macro expands it or turns it into a named token (like
917  /// 'for').
918  void HandleIdentifier(Token &Identifier);
919
920
921  /// HandleEndOfFile - This callback is invoked when the lexer hits the end of
922  /// the current file.  This either returns the EOF token and returns true, or
923  /// pops a level off the include stack and returns false, at which point the
924  /// client should call lex again.
925  bool HandleEndOfFile(Token &Result, bool isEndOfMacro = false);
926
927  /// HandleEndOfTokenLexer - This callback is invoked when the current
928  /// TokenLexer hits the end of its token stream.
929  bool HandleEndOfTokenLexer(Token &Result);
930
931  /// HandleDirective - This callback is invoked when the lexer sees a # token
932  /// at the start of a line.  This consumes the directive, modifies the
933  /// lexer/preprocessor state, and advances the lexer(s) so that the next token
934  /// read is the correct one.
935  void HandleDirective(Token &Result);
936
937  /// CheckEndOfDirective - Ensure that the next token is a tok::eod token.  If
938  /// not, emit a diagnostic and consume up until the eod.  If EnableMacros is
939  /// true, then we consider macros that expand to zero tokens as being ok.
940  void CheckEndOfDirective(const char *Directive, bool EnableMacros = false);
941
942  /// DiscardUntilEndOfDirective - Read and discard all tokens remaining on the
943  /// current line until the tok::eod token is found.
944  void DiscardUntilEndOfDirective();
945
946  /// SawDateOrTime - This returns true if the preprocessor has seen a use of
947  /// __DATE__ or __TIME__ in the file so far.
948  bool SawDateOrTime() const {
949    return DATELoc != SourceLocation() || TIMELoc != SourceLocation();
950  }
951  unsigned getCounterValue() const { return CounterValue; }
952  void setCounterValue(unsigned V) { CounterValue = V; }
953
954  /// AllocateMacroInfo - Allocate a new MacroInfo object with the provide
955  ///  SourceLocation.
956  MacroInfo *AllocateMacroInfo(SourceLocation L);
957
958  /// CloneMacroInfo - Allocate a new MacroInfo object which is clone of MI.
959  MacroInfo *CloneMacroInfo(const MacroInfo &MI);
960
961  /// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully
962  /// checked and spelled filename, e.g. as an operand of #include. This returns
963  /// true if the input filename was in <>'s or false if it were in ""'s.  The
964  /// caller is expected to provide a buffer that is large enough to hold the
965  /// spelling of the filename, but is also expected to handle the case when
966  /// this method decides to use a different buffer.
967  bool GetIncludeFilenameSpelling(SourceLocation Loc,StringRef &Filename);
968
969  /// LookupFile - Given a "foo" or <foo> reference, look up the indicated file,
970  /// return null on failure.  isAngled indicates whether the file reference is
971  /// for system #include's or not (i.e. using <> instead of "").
972  const FileEntry *LookupFile(StringRef Filename,
973                              bool isAngled, const DirectoryLookup *FromDir,
974                              const DirectoryLookup *&CurDir,
975                              SmallVectorImpl<char> *SearchPath,
976                              SmallVectorImpl<char> *RelativePath);
977
978  /// GetCurLookup - The DirectoryLookup structure used to find the current
979  /// FileEntry, if CurLexer is non-null and if applicable.  This allows us to
980  /// implement #include_next and find directory-specific properties.
981  const DirectoryLookup *GetCurDirLookup() { return CurDirLookup; }
982
983  /// isInPrimaryFile - Return true if we're in the top-level file, not in a
984  /// #include.
985  bool isInPrimaryFile() const;
986
987  /// ConcatenateIncludeName - Handle cases where the #include name is expanded
988  /// from a macro as multiple tokens, which need to be glued together.  This
989  /// occurs for code like:
990  ///    #define FOO <a/b.h>
991  ///    #include FOO
992  /// because in this case, "<a/b.h>" is returned as 7 tokens, not one.
993  ///
994  /// This code concatenates and consumes tokens up to the '>' token.  It
995  /// returns false if the > was found, otherwise it returns true if it finds
996  /// and consumes the EOD marker.
997  bool ConcatenateIncludeName(llvm::SmallString<128> &FilenameBuffer,
998                              SourceLocation &End);
999
1000  /// LexOnOffSwitch - Lex an on-off-switch (C99 6.10.6p2) and verify that it is
1001  /// followed by EOD.  Return true if the token is not a valid on-off-switch.
1002  bool LexOnOffSwitch(tok::OnOffSwitch &OOS);
1003
1004private:
1005
1006  void PushIncludeMacroStack() {
1007    IncludeMacroStack.push_back(IncludeStackInfo(CurLexerKind,
1008                                                 CurLexer.take(),
1009                                                 CurPTHLexer.take(),
1010                                                 CurPPLexer,
1011                                                 CurTokenLexer.take(),
1012                                                 CurDirLookup));
1013    CurPPLexer = 0;
1014  }
1015
1016  void PopIncludeMacroStack() {
1017    CurLexer.reset(IncludeMacroStack.back().TheLexer);
1018    CurPTHLexer.reset(IncludeMacroStack.back().ThePTHLexer);
1019    CurPPLexer = IncludeMacroStack.back().ThePPLexer;
1020    CurTokenLexer.reset(IncludeMacroStack.back().TheTokenLexer);
1021    CurDirLookup  = IncludeMacroStack.back().TheDirLookup;
1022    CurLexerKind = IncludeMacroStack.back().CurLexerKind;
1023    IncludeMacroStack.pop_back();
1024  }
1025
1026  /// AllocateMacroInfo - Allocate a new MacroInfo object.
1027  MacroInfo *AllocateMacroInfo();
1028
1029  /// ReleaseMacroInfo - Release the specified MacroInfo.  This memory will
1030  ///  be reused for allocating new MacroInfo objects.
1031  void ReleaseMacroInfo(MacroInfo* MI);
1032
1033  /// ReadMacroName - Lex and validate a macro name, which occurs after a
1034  /// #define or #undef.  This emits a diagnostic, sets the token kind to eod,
1035  /// and discards the rest of the macro line if the macro name is invalid.
1036  void ReadMacroName(Token &MacroNameTok, char isDefineUndef = 0);
1037
1038  /// ReadMacroDefinitionArgList - The ( starting an argument list of a macro
1039  /// definition has just been read.  Lex the rest of the arguments and the
1040  /// closing ), updating MI with what we learn.  Return true if an error occurs
1041  /// parsing the arg list.
1042  bool ReadMacroDefinitionArgList(MacroInfo *MI);
1043
1044  /// SkipExcludedConditionalBlock - We just read a #if or related directive and
1045  /// decided that the subsequent tokens are in the #if'd out portion of the
1046  /// file.  Lex the rest of the file, until we see an #endif.  If
1047  /// FoundNonSkipPortion is true, then we have already emitted code for part of
1048  /// this #if directive, so #else/#elif blocks should never be entered. If
1049  /// FoundElse is false, then #else directives are ok, if not, then we have
1050  /// already seen one so a #else directive is a duplicate.  When this returns,
1051  /// the caller can lex the first valid token.
1052  void SkipExcludedConditionalBlock(SourceLocation IfTokenLoc,
1053                                    bool FoundNonSkipPortion, bool FoundElse);
1054
1055  /// PTHSkipExcludedConditionalBlock - A fast PTH version of
1056  ///  SkipExcludedConditionalBlock.
1057  void PTHSkipExcludedConditionalBlock();
1058
1059  /// EvaluateDirectiveExpression - Evaluate an integer constant expression that
1060  /// may occur after a #if or #elif directive and return it as a bool.  If the
1061  /// expression is equivalent to "!defined(X)" return X in IfNDefMacro.
1062  bool EvaluateDirectiveExpression(IdentifierInfo *&IfNDefMacro);
1063
1064  /// RegisterBuiltinPragmas - Install the standard preprocessor pragmas:
1065  /// #pragma GCC poison/system_header/dependency and #pragma once.
1066  void RegisterBuiltinPragmas();
1067
1068  /// RegisterBuiltinMacros - Register builtin macros, such as __LINE__ with the
1069  /// identifier table.
1070  void RegisterBuiltinMacros();
1071
1072  /// HandleMacroExpandedIdentifier - If an identifier token is read that is to
1073  /// be expanded as a macro, handle it and return the next token as 'Tok'.  If
1074  /// the macro should not be expanded return true, otherwise return false.
1075  bool HandleMacroExpandedIdentifier(Token &Tok, MacroInfo *MI);
1076
1077  /// \brief Cache macro expanded tokens for TokenLexers.
1078  //
1079  /// Works like a stack; a TokenLexer adds the macro expanded tokens that is
1080  /// going to lex in the cache and when it finishes the tokens are removed
1081  /// from the end of the cache.
1082  Token *cacheMacroExpandedTokens(TokenLexer *tokLexer,
1083                                  ArrayRef<Token> tokens);
1084  void removeCachedMacroExpandedTokensOfLastLexer();
1085  friend void TokenLexer::ExpandFunctionArguments();
1086
1087  /// isNextPPTokenLParen - Determine whether the next preprocessor token to be
1088  /// lexed is a '('.  If so, consume the token and return true, if not, this
1089  /// method should have no observable side-effect on the lexed tokens.
1090  bool isNextPPTokenLParen();
1091
1092  /// ReadFunctionLikeMacroArgs - After reading "MACRO(", this method is
1093  /// invoked to read all of the formal arguments specified for the macro
1094  /// invocation.  This returns null on error.
1095  MacroArgs *ReadFunctionLikeMacroArgs(Token &MacroName, MacroInfo *MI,
1096                                       SourceLocation &ExpansionEnd);
1097
1098  /// ExpandBuiltinMacro - If an identifier token is read that is to be expanded
1099  /// as a builtin macro, handle it and return the next token as 'Tok'.
1100  void ExpandBuiltinMacro(Token &Tok);
1101
1102  /// Handle_Pragma - Read a _Pragma directive, slice it up, process it, then
1103  /// return the first token after the directive.  The _Pragma token has just
1104  /// been read into 'Tok'.
1105  void Handle_Pragma(Token &Tok);
1106
1107  /// HandleMicrosoft__pragma - Like Handle_Pragma except the pragma text
1108  /// is not enclosed within a string literal.
1109  void HandleMicrosoft__pragma(Token &Tok);
1110
1111  /// EnterSourceFileWithLexer - Add a lexer to the top of the include stack and
1112  /// start lexing tokens from it instead of the current buffer.
1113  void EnterSourceFileWithLexer(Lexer *TheLexer, const DirectoryLookup *Dir);
1114
1115  /// EnterSourceFileWithPTH - Add a lexer to the top of the include stack and
1116  /// start getting tokens from it using the PTH cache.
1117  void EnterSourceFileWithPTH(PTHLexer *PL, const DirectoryLookup *Dir);
1118
1119  /// IsFileLexer - Returns true if we are lexing from a file and not a
1120  ///  pragma or a macro.
1121  static bool IsFileLexer(const Lexer* L, const PreprocessorLexer* P) {
1122    return L ? !L->isPragmaLexer() : P != 0;
1123  }
1124
1125  static bool IsFileLexer(const IncludeStackInfo& I) {
1126    return IsFileLexer(I.TheLexer, I.ThePPLexer);
1127  }
1128
1129  bool IsFileLexer() const {
1130    return IsFileLexer(CurLexer.get(), CurPPLexer);
1131  }
1132
1133  //===--------------------------------------------------------------------===//
1134  // Caching stuff.
1135  void CachingLex(Token &Result);
1136  bool InCachingLexMode() const {
1137    // If the Lexer pointers are 0 and IncludeMacroStack is empty, it means
1138    // that we are past EOF, not that we are in CachingLex mode.
1139    return CurPPLexer == 0 && CurTokenLexer == 0 && CurPTHLexer == 0 &&
1140           !IncludeMacroStack.empty();
1141  }
1142  void EnterCachingLexMode();
1143  void ExitCachingLexMode() {
1144    if (InCachingLexMode())
1145      RemoveTopOfLexerStack();
1146  }
1147  const Token &PeekAhead(unsigned N);
1148  void AnnotatePreviousCachedTokens(const Token &Tok);
1149
1150  //===--------------------------------------------------------------------===//
1151  /// Handle*Directive - implement the various preprocessor directives.  These
1152  /// should side-effect the current preprocessor object so that the next call
1153  /// to Lex() will return the appropriate token next.
1154  void HandleLineDirective(Token &Tok);
1155  void HandleDigitDirective(Token &Tok);
1156  void HandleUserDiagnosticDirective(Token &Tok, bool isWarning);
1157  void HandleIdentSCCSDirective(Token &Tok);
1158  void HandleMacroExportDirective(Token &Tok);
1159
1160  // File inclusion.
1161  void HandleIncludeDirective(SourceLocation HashLoc,
1162                              Token &Tok,
1163                              const DirectoryLookup *LookupFrom = 0,
1164                              bool isImport = false);
1165  void HandleIncludeNextDirective(SourceLocation HashLoc, Token &Tok);
1166  void HandleIncludeMacrosDirective(SourceLocation HashLoc, Token &Tok);
1167  void HandleImportDirective(SourceLocation HashLoc, Token &Tok);
1168
1169  // Macro handling.
1170  void HandleDefineDirective(Token &Tok);
1171  void HandleUndefDirective(Token &Tok);
1172
1173  // Conditional Inclusion.
1174  void HandleIfdefDirective(Token &Tok, bool isIfndef,
1175                            bool ReadAnyTokensBeforeDirective);
1176  void HandleIfDirective(Token &Tok, bool ReadAnyTokensBeforeDirective);
1177  void HandleEndifDirective(Token &Tok);
1178  void HandleElseDirective(Token &Tok);
1179  void HandleElifDirective(Token &Tok);
1180
1181  // Pragmas.
1182  void HandlePragmaDirective(unsigned Introducer);
1183public:
1184  void HandlePragmaOnce(Token &OnceTok);
1185  void HandlePragmaMark();
1186  void HandlePragmaPoison(Token &PoisonTok);
1187  void HandlePragmaSystemHeader(Token &SysHeaderTok);
1188  void HandlePragmaDependency(Token &DependencyTok);
1189  void HandlePragmaComment(Token &CommentTok);
1190  void HandlePragmaMessage(Token &MessageTok);
1191  void HandlePragmaPushMacro(Token &Tok);
1192  void HandlePragmaPopMacro(Token &Tok);
1193  IdentifierInfo *ParsePragmaPushOrPopMacro(Token &Tok);
1194
1195  // Return true and store the first token only if any CommentHandler
1196  // has inserted some tokens and getCommentRetentionState() is false.
1197  bool HandleComment(Token &Token, SourceRange Comment);
1198
1199  /// \brief A macro is used, update information about macros that need unused
1200  /// warnings.
1201  void markMacroAsUsed(MacroInfo *MI);
1202};
1203
1204/// \brief Abstract base class that describes a handler that will receive
1205/// source ranges for each of the comments encountered in the source file.
1206class CommentHandler {
1207public:
1208  virtual ~CommentHandler();
1209
1210  // The handler shall return true if it has pushed any tokens
1211  // to be read using e.g. EnterToken or EnterTokenStream.
1212  virtual bool HandleComment(Preprocessor &PP, SourceRange Comment) = 0;
1213};
1214
1215}  // end namespace clang
1216
1217#endif
1218