Preprocessor.h revision 1c2e9332fa69727425a3a2b912e36e2ab62083f8
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  DiagnosticsEngine        *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  IdentifierInfo *Ident__has_warning;              // __has_warning
97
98  SourceLocation DATELoc, TIMELoc;
99  unsigned CounterValue;  // Next __COUNTER__ value.
100
101  enum {
102    /// MaxIncludeStackDepth - Maximum depth of #includes.
103    MaxAllowedIncludeStackDepth = 200
104  };
105
106  // State that is set before the preprocessor begins.
107  bool KeepComments : 1;
108  bool KeepMacroComments : 1;
109  bool SuppressIncludeNotFoundError : 1;
110  bool AutoModuleImport : 1;
111
112  // State that changes while the preprocessor runs:
113  bool InMacroArgs : 1;            // True if parsing fn macro invocation args.
114
115  /// Whether the preprocessor owns the header search object.
116  bool OwnsHeaderSearch : 1;
117
118  /// DisableMacroExpansion - True if macro expansion is disabled.
119  bool DisableMacroExpansion : 1;
120
121  /// \brief Whether we have already loaded macros from the external source.
122  mutable bool ReadMacrosFromExternalSource : 1;
123
124  /// Identifiers - This is mapping/lookup information for all identifiers in
125  /// the program, including program keywords.
126  mutable IdentifierTable Identifiers;
127
128  /// Selectors - This table contains all the selectors in the program. Unlike
129  /// IdentifierTable above, this table *isn't* populated by the preprocessor.
130  /// It is declared/expanded here because it's role/lifetime is
131  /// conceptually similar the IdentifierTable. In addition, the current control
132  /// flow (in clang::ParseAST()), make it convenient to put here.
133  /// FIXME: Make sure the lifetime of Identifiers/Selectors *isn't* tied to
134  /// the lifetime of the preprocessor.
135  SelectorTable Selectors;
136
137  /// BuiltinInfo - Information about builtins.
138  Builtin::Context BuiltinInfo;
139
140  /// PragmaHandlers - This tracks all of the pragmas that the client registered
141  /// with this preprocessor.
142  PragmaNamespace *PragmaHandlers;
143
144  /// \brief Tracks all of the comment handlers that the client registered
145  /// with this preprocessor.
146  std::vector<CommentHandler *> CommentHandlers;
147
148  /// \brief The code-completion handler.
149  CodeCompletionHandler *CodeComplete;
150
151  /// \brief The file that we're performing code-completion for, if any.
152  const FileEntry *CodeCompletionFile;
153
154  /// \brief The offset in file for the code-completion point.
155  unsigned CodeCompletionOffset;
156
157  /// \brief The location for the code-completion point. This gets instantiated
158  /// when the CodeCompletionFile gets #include'ed for preprocessing.
159  SourceLocation CodeCompletionLoc;
160
161  /// \brief The start location for the file of the code-completion point.
162  /// This gets instantiated when the CodeCompletionFile gets #include'ed
163  /// for preprocessing.
164  SourceLocation CodeCompletionFileLoc;
165
166  /// \brief The source location of the __import_module__ keyword we just
167  /// lexed, if any.
168  SourceLocation ModuleImportLoc;
169
170  /// \brief The source location of the currently-active
171  /// #pragma clang arc_cf_code_audited begin.
172  SourceLocation PragmaARCCFCodeAuditedLoc;
173
174  /// \brief True if we hit the code-completion point.
175  bool CodeCompletionReached;
176
177  /// \brief The number of bytes that we will initially skip when entering the
178  /// main file, which is used when loading a precompiled preamble, along
179  /// with a flag that indicates whether skipping this number of bytes will
180  /// place the lexer at the start of a line.
181  std::pair<unsigned, bool> SkipMainFilePreamble;
182
183  /// CurLexer - This is the current top of the stack that we're lexing from if
184  /// not expanding a macro and we are lexing directly from source code.
185  ///  Only one of CurLexer, CurPTHLexer, or CurTokenLexer will be non-null.
186  llvm::OwningPtr<Lexer> CurLexer;
187
188  /// CurPTHLexer - This is the current top of stack that we're lexing from if
189  ///  not expanding from a macro and we are lexing from a PTH cache.
190  ///  Only one of CurLexer, CurPTHLexer, or CurTokenLexer will be non-null.
191  llvm::OwningPtr<PTHLexer> CurPTHLexer;
192
193  /// CurPPLexer - This is the current top of the stack what we're lexing from
194  ///  if not expanding a macro.  This is an alias for either CurLexer or
195  ///  CurPTHLexer.
196  PreprocessorLexer *CurPPLexer;
197
198  /// CurLookup - The DirectoryLookup structure used to find the current
199  /// FileEntry, if CurLexer is non-null and if applicable.  This allows us to
200  /// implement #include_next and find directory-specific properties.
201  const DirectoryLookup *CurDirLookup;
202
203  /// CurTokenLexer - This is the current macro we are expanding, if we are
204  /// expanding a macro.  One of CurLexer and CurTokenLexer must be null.
205  llvm::OwningPtr<TokenLexer> CurTokenLexer;
206
207  /// \brief The kind of lexer we're currently working with.
208  enum CurLexerKind {
209    CLK_Lexer,
210    CLK_PTHLexer,
211    CLK_TokenLexer,
212    CLK_CachingLexer,
213    CLK_LexAfterModuleImport
214  } CurLexerKind;
215
216  /// IncludeMacroStack - This keeps track of the stack of files currently
217  /// #included, and macros currently being expanded from, not counting
218  /// CurLexer/CurTokenLexer.
219  struct IncludeStackInfo {
220    enum CurLexerKind     CurLexerKind;
221    Lexer                 *TheLexer;
222    PTHLexer              *ThePTHLexer;
223    PreprocessorLexer     *ThePPLexer;
224    TokenLexer            *TheTokenLexer;
225    const DirectoryLookup *TheDirLookup;
226
227    IncludeStackInfo(enum CurLexerKind K, Lexer *L, PTHLexer* P,
228                     PreprocessorLexer* PPL,
229                     TokenLexer* TL, const DirectoryLookup *D)
230      : CurLexerKind(K), TheLexer(L), ThePTHLexer(P), ThePPLexer(PPL),
231        TheTokenLexer(TL), TheDirLookup(D) {}
232  };
233  std::vector<IncludeStackInfo> IncludeMacroStack;
234
235  /// Callbacks - These are actions invoked when some preprocessor activity is
236  /// encountered (e.g. a file is #included, etc).
237  PPCallbacks *Callbacks;
238
239  /// Macros - For each IdentifierInfo with 'HasMacro' set, we keep a mapping
240  /// to the actual definition of the macro.
241  llvm::DenseMap<IdentifierInfo*, MacroInfo*> Macros;
242
243  /// \brief Macros that we want to warn because they are not used at the end
244  /// of the translation unit; we store just their SourceLocations instead
245  /// something like MacroInfo*. The benefit of this is that when we are
246  /// deserializing from PCH, we don't need to deserialize identifier & macros
247  /// just so that we can report that they are unused, we just warn using
248  /// the SourceLocations of this set (that will be filled by the ASTReader).
249  /// We are using SmallPtrSet instead of a vector for faster removal.
250  typedef llvm::SmallPtrSet<SourceLocation, 32> WarnUnusedMacroLocsTy;
251  WarnUnusedMacroLocsTy WarnUnusedMacroLocs;
252
253  /// MacroArgCache - This is a "freelist" of MacroArg objects that can be
254  /// reused for quick allocation.
255  MacroArgs *MacroArgCache;
256  friend class MacroArgs;
257
258  /// PragmaPushMacroInfo - For each IdentifierInfo used in a #pragma
259  /// push_macro directive, we keep a MacroInfo stack used to restore
260  /// previous macro value.
261  llvm::DenseMap<IdentifierInfo*, std::vector<MacroInfo*> > PragmaPushMacroInfo;
262
263  // Various statistics we track for performance analysis.
264  unsigned NumDirectives, NumIncluded, NumDefined, NumUndefined, NumPragma;
265  unsigned NumIf, NumElse, NumEndif;
266  unsigned NumEnteredSourceFiles, MaxIncludeStackDepth;
267  unsigned NumMacroExpanded, NumFnMacroExpanded, NumBuiltinMacroExpanded;
268  unsigned NumFastMacroExpanded, NumTokenPaste, NumFastTokenPaste;
269  unsigned NumSkipped;
270
271  /// Predefines - This string is the predefined macros that preprocessor
272  /// should use from the command line etc.
273  std::string Predefines;
274
275  /// TokenLexerCache - Cache macro expanders to reduce malloc traffic.
276  enum { TokenLexerCacheSize = 8 };
277  unsigned NumCachedTokenLexers;
278  TokenLexer *TokenLexerCache[TokenLexerCacheSize];
279
280  /// \brief Keeps macro expanded tokens for TokenLexers.
281  //
282  /// Works like a stack; a TokenLexer adds the macro expanded tokens that is
283  /// going to lex in the cache and when it finishes the tokens are removed
284  /// from the end of the cache.
285  SmallVector<Token, 16> MacroExpandedTokens;
286  std::vector<std::pair<TokenLexer *, size_t> > MacroExpandingLexersStack;
287
288  /// \brief A record of the macro definitions and expansions that
289  /// occurred during preprocessing.
290  ///
291  /// This is an optional side structure that can be enabled with
292  /// \c createPreprocessingRecord() prior to preprocessing.
293  PreprocessingRecord *Record;
294
295private:  // Cached tokens state.
296  typedef SmallVector<Token, 1> CachedTokensTy;
297
298  /// CachedTokens - Cached tokens are stored here when we do backtracking or
299  /// lookahead. They are "lexed" by the CachingLex() method.
300  CachedTokensTy CachedTokens;
301
302  /// CachedLexPos - The position of the cached token that CachingLex() should
303  /// "lex" next. If it points beyond the CachedTokens vector, it means that
304  /// a normal Lex() should be invoked.
305  CachedTokensTy::size_type CachedLexPos;
306
307  /// BacktrackPositions - Stack of backtrack positions, allowing nested
308  /// backtracks. The EnableBacktrackAtThisPos() method pushes a position to
309  /// indicate where CachedLexPos should be set when the BackTrack() method is
310  /// invoked (at which point the last position is popped).
311  std::vector<CachedTokensTy::size_type> BacktrackPositions;
312
313  struct MacroInfoChain {
314    MacroInfo MI;
315    MacroInfoChain *Next;
316    MacroInfoChain *Prev;
317  };
318
319  /// MacroInfos are managed as a chain for easy disposal.  This is the head
320  /// of that list.
321  MacroInfoChain *MIChainHead;
322
323  /// MICache - A "freelist" of MacroInfo objects that can be reused for quick
324  /// allocation.
325  MacroInfoChain *MICache;
326
327  MacroInfo *getInfoForMacro(IdentifierInfo *II) const;
328
329public:
330  Preprocessor(DiagnosticsEngine &diags, LangOptions &opts,
331               const TargetInfo *target,
332               SourceManager &SM, HeaderSearch &Headers,
333               ModuleLoader &TheModuleLoader,
334               IdentifierInfoLookup *IILookup = 0,
335               bool OwnsHeaderSearch = false,
336               bool DelayInitialization = false);
337
338  ~Preprocessor();
339
340  /// \brief Initialize the preprocessor, if the constructor did not already
341  /// perform the initialization.
342  ///
343  /// \param Target Information about the target.
344  void Initialize(const TargetInfo &Target);
345
346  DiagnosticsEngine &getDiagnostics() const { return *Diags; }
347  void setDiagnostics(DiagnosticsEngine &D) { Diags = &D; }
348
349  const LangOptions &getLangOptions() const { return Features; }
350  const TargetInfo &getTargetInfo() const { return *Target; }
351  FileManager &getFileManager() const { return FileMgr; }
352  SourceManager &getSourceManager() const { return SourceMgr; }
353  HeaderSearch &getHeaderSearchInfo() const { return HeaderInfo; }
354
355  IdentifierTable &getIdentifierTable() { return Identifiers; }
356  SelectorTable &getSelectorTable() { return Selectors; }
357  Builtin::Context &getBuiltinInfo() { return BuiltinInfo; }
358  llvm::BumpPtrAllocator &getPreprocessorAllocator() { return BP; }
359
360  void setPTHManager(PTHManager* pm);
361
362  PTHManager *getPTHManager() { return PTH.get(); }
363
364  void setExternalSource(ExternalPreprocessorSource *Source) {
365    ExternalSource = Source;
366  }
367
368  ExternalPreprocessorSource *getExternalSource() const {
369    return ExternalSource;
370  }
371
372  /// \brief Retrieve the module loader associated with this preprocessor.
373  ModuleLoader &getModuleLoader() const { return TheModuleLoader; }
374
375  /// SetCommentRetentionState - Control whether or not the preprocessor retains
376  /// comments in output.
377  void SetCommentRetentionState(bool KeepComments, bool KeepMacroComments) {
378    this->KeepComments = KeepComments | KeepMacroComments;
379    this->KeepMacroComments = KeepMacroComments;
380  }
381
382  bool getCommentRetentionState() const { return KeepComments; }
383
384  void SetSuppressIncludeNotFoundError(bool Suppress) {
385    SuppressIncludeNotFoundError = Suppress;
386  }
387
388  bool GetSuppressIncludeNotFoundError() {
389    return SuppressIncludeNotFoundError;
390  }
391
392  /// \brief Specify whether automatic module imports are enabled.
393  void setAutoModuleImport(bool AutoModuleImport = true) {
394    this->AutoModuleImport = AutoModuleImport;
395  }
396
397  /// isCurrentLexer - Return true if we are lexing directly from the specified
398  /// lexer.
399  bool isCurrentLexer(const PreprocessorLexer *L) const {
400    return CurPPLexer == L;
401  }
402
403  /// getCurrentLexer - Return the current lexer being lexed from.  Note
404  /// that this ignores any potentially active macro expansions and _Pragma
405  /// expansions going on at the time.
406  PreprocessorLexer *getCurrentLexer() const { return CurPPLexer; }
407
408  /// getCurrentFileLexer - Return the current file lexer being lexed from.
409  /// Note that this ignores any potentially active macro expansions and _Pragma
410  /// expansions going on at the time.
411  PreprocessorLexer *getCurrentFileLexer() const;
412
413  /// getPPCallbacks/addPPCallbacks - Accessors for preprocessor callbacks.
414  /// Note that this class takes ownership of any PPCallbacks object given to
415  /// it.
416  PPCallbacks *getPPCallbacks() const { return Callbacks; }
417  void addPPCallbacks(PPCallbacks *C) {
418    if (Callbacks)
419      C = new PPChainedCallbacks(C, Callbacks);
420    Callbacks = C;
421  }
422
423  /// getMacroInfo - Given an identifier, return the MacroInfo it is #defined to
424  /// or null if it isn't #define'd.
425  MacroInfo *getMacroInfo(IdentifierInfo *II) const {
426    if (!II->hasMacroDefinition())
427      return 0;
428
429    return getInfoForMacro(II);
430  }
431
432  /// setMacroInfo - Specify a macro for this identifier.
433  ///
434  void setMacroInfo(IdentifierInfo *II, MacroInfo *MI);
435
436  /// macro_iterator/macro_begin/macro_end - This allows you to walk the current
437  /// state of the macro table.  This visits every currently-defined macro.
438  typedef llvm::DenseMap<IdentifierInfo*,
439                         MacroInfo*>::const_iterator macro_iterator;
440  macro_iterator macro_begin(bool IncludeExternalMacros = true) const;
441  macro_iterator macro_end(bool IncludeExternalMacros = true) const;
442
443  const std::string &getPredefines() const { return Predefines; }
444  /// setPredefines - Set the predefines for this Preprocessor.  These
445  /// predefines are automatically injected when parsing the main file.
446  void setPredefines(const char *P) { Predefines = P; }
447  void setPredefines(const std::string &P) { Predefines = P; }
448
449  /// getIdentifierInfo - Return information about the specified preprocessor
450  /// identifier token.  The version of this method that takes two character
451  /// pointers is preferred unless the identifier is already available as a
452  /// string (this avoids allocation and copying of memory to construct an
453  /// std::string).
454  IdentifierInfo *getIdentifierInfo(StringRef Name) const {
455    return &Identifiers.get(Name);
456  }
457
458  /// AddPragmaHandler - Add the specified pragma handler to the preprocessor.
459  /// If 'Namespace' is non-null, then it is a token required to exist on the
460  /// pragma line before the pragma string starts, e.g. "STDC" or "GCC".
461  void AddPragmaHandler(StringRef Namespace, PragmaHandler *Handler);
462  void AddPragmaHandler(PragmaHandler *Handler) {
463    AddPragmaHandler(StringRef(), Handler);
464  }
465
466  /// RemovePragmaHandler - Remove the specific pragma handler from
467  /// the preprocessor. If \arg Namespace is non-null, then it should
468  /// be the namespace that \arg Handler was added to. It is an error
469  /// to remove a handler that has not been registered.
470  void RemovePragmaHandler(StringRef Namespace, PragmaHandler *Handler);
471  void RemovePragmaHandler(PragmaHandler *Handler) {
472    RemovePragmaHandler(StringRef(), Handler);
473  }
474
475  /// \brief Add the specified comment handler to the preprocessor.
476  void AddCommentHandler(CommentHandler *Handler);
477
478  /// \brief Remove the specified comment handler.
479  ///
480  /// It is an error to remove a handler that has not been registered.
481  void RemoveCommentHandler(CommentHandler *Handler);
482
483  /// \brief Set the code completion handler to the given object.
484  void setCodeCompletionHandler(CodeCompletionHandler &Handler) {
485    CodeComplete = &Handler;
486  }
487
488  /// \brief Retrieve the current code-completion handler.
489  CodeCompletionHandler *getCodeCompletionHandler() const {
490    return CodeComplete;
491  }
492
493  /// \brief Clear out the code completion handler.
494  void clearCodeCompletionHandler() {
495    CodeComplete = 0;
496  }
497
498  /// \brief Hook used by the lexer to invoke the "natural language" code
499  /// completion point.
500  void CodeCompleteNaturalLanguage();
501
502  /// \brief Retrieve the preprocessing record, or NULL if there is no
503  /// preprocessing record.
504  PreprocessingRecord *getPreprocessingRecord() const { return Record; }
505
506  /// \brief Create a new preprocessing record, which will keep track of
507  /// all macro expansions, macro definitions, etc.
508  void createPreprocessingRecord(bool IncludeNestedMacroExpansions);
509
510  /// EnterMainSourceFile - Enter the specified FileID as the main source file,
511  /// which implicitly adds the builtin defines etc.
512  void EnterMainSourceFile();
513
514  /// EndSourceFile - Inform the preprocessor callbacks that processing is
515  /// complete.
516  void EndSourceFile();
517
518  /// EnterSourceFile - Add a source file to the top of the include stack and
519  /// start lexing tokens from it instead of the current buffer.  Emit an error
520  /// and don't enter the file on error.
521  void EnterSourceFile(FileID CurFileID, const DirectoryLookup *Dir,
522                       SourceLocation Loc);
523
524  /// EnterMacro - Add a Macro to the top of the include stack and start lexing
525  /// tokens from it instead of the current buffer.  Args specifies the
526  /// tokens input to a function-like macro.
527  ///
528  /// ILEnd specifies the location of the ')' for a function-like macro or the
529  /// identifier for an object-like macro.
530  void EnterMacro(Token &Identifier, SourceLocation ILEnd, MacroArgs *Args);
531
532  /// EnterTokenStream - Add a "macro" context to the top of the include stack,
533  /// which will cause the lexer to start returning the specified tokens.
534  ///
535  /// If DisableMacroExpansion is true, tokens lexed from the token stream will
536  /// not be subject to further macro expansion.  Otherwise, these tokens will
537  /// be re-macro-expanded when/if expansion is enabled.
538  ///
539  /// If OwnsTokens is false, this method assumes that the specified stream of
540  /// tokens has a permanent owner somewhere, so they do not need to be copied.
541  /// If it is true, it assumes the array of tokens is allocated with new[] and
542  /// must be freed.
543  ///
544  void EnterTokenStream(const Token *Toks, unsigned NumToks,
545                        bool DisableMacroExpansion, bool OwnsTokens);
546
547  /// RemoveTopOfLexerStack - Pop the current lexer/macro exp off the top of the
548  /// lexer stack.  This should only be used in situations where the current
549  /// state of the top-of-stack lexer is known.
550  void RemoveTopOfLexerStack();
551
552  /// EnableBacktrackAtThisPos - From the point that this method is called, and
553  /// until CommitBacktrackedTokens() or Backtrack() is called, the Preprocessor
554  /// keeps track of the lexed tokens so that a subsequent Backtrack() call will
555  /// make the Preprocessor re-lex the same tokens.
556  ///
557  /// Nested backtracks are allowed, meaning that EnableBacktrackAtThisPos can
558  /// be called multiple times and CommitBacktrackedTokens/Backtrack calls will
559  /// be combined with the EnableBacktrackAtThisPos calls in reverse order.
560  ///
561  /// NOTE: *DO NOT* forget to call either CommitBacktrackedTokens or Backtrack
562  /// at some point after EnableBacktrackAtThisPos. If you don't, caching of
563  /// tokens will continue indefinitely.
564  ///
565  void EnableBacktrackAtThisPos();
566
567  /// CommitBacktrackedTokens - Disable the last EnableBacktrackAtThisPos call.
568  void CommitBacktrackedTokens();
569
570  /// Backtrack - Make Preprocessor re-lex the tokens that were lexed since
571  /// EnableBacktrackAtThisPos() was previously called.
572  void Backtrack();
573
574  /// isBacktrackEnabled - True if EnableBacktrackAtThisPos() was called and
575  /// caching of tokens is on.
576  bool isBacktrackEnabled() const { return !BacktrackPositions.empty(); }
577
578  /// Lex - To lex a token from the preprocessor, just pull a token from the
579  /// current lexer or macro object.
580  void Lex(Token &Result) {
581    switch (CurLexerKind) {
582    case CLK_Lexer: CurLexer->Lex(Result); break;
583    case CLK_PTHLexer: CurPTHLexer->Lex(Result); break;
584    case CLK_TokenLexer: CurTokenLexer->Lex(Result); break;
585    case CLK_CachingLexer: CachingLex(Result); break;
586    case CLK_LexAfterModuleImport: LexAfterModuleImport(Result); break;
587    }
588  }
589
590  void LexAfterModuleImport(Token &Result);
591
592  /// LexNonComment - Lex a token.  If it's a comment, keep lexing until we get
593  /// something not a comment.  This is useful in -E -C mode where comments
594  /// would foul up preprocessor directive handling.
595  void LexNonComment(Token &Result) {
596    do
597      Lex(Result);
598    while (Result.getKind() == tok::comment);
599  }
600
601  /// LexUnexpandedToken - This is just like Lex, but this disables macro
602  /// expansion of identifier tokens.
603  void LexUnexpandedToken(Token &Result) {
604    // Disable macro expansion.
605    bool OldVal = DisableMacroExpansion;
606    DisableMacroExpansion = true;
607    // Lex the token.
608    Lex(Result);
609
610    // Reenable it.
611    DisableMacroExpansion = OldVal;
612  }
613
614  /// LexUnexpandedNonComment - Like LexNonComment, but this disables macro
615  /// expansion of identifier tokens.
616  void LexUnexpandedNonComment(Token &Result) {
617    do
618      LexUnexpandedToken(Result);
619    while (Result.getKind() == tok::comment);
620  }
621
622  /// LookAhead - This peeks ahead N tokens and returns that token without
623  /// consuming any tokens.  LookAhead(0) returns the next token that would be
624  /// returned by Lex(), LookAhead(1) returns the token after it, etc.  This
625  /// returns normal tokens after phase 5.  As such, it is equivalent to using
626  /// 'Lex', not 'LexUnexpandedToken'.
627  const Token &LookAhead(unsigned N) {
628    if (CachedLexPos + N < CachedTokens.size())
629      return CachedTokens[CachedLexPos+N];
630    else
631      return PeekAhead(N+1);
632  }
633
634  /// RevertCachedTokens - When backtracking is enabled and tokens are cached,
635  /// this allows to revert a specific number of tokens.
636  /// Note that the number of tokens being reverted should be up to the last
637  /// backtrack position, not more.
638  void RevertCachedTokens(unsigned N) {
639    assert(isBacktrackEnabled() &&
640           "Should only be called when tokens are cached for backtracking");
641    assert(signed(CachedLexPos) - signed(N) >= signed(BacktrackPositions.back())
642         && "Should revert tokens up to the last backtrack position, not more");
643    assert(signed(CachedLexPos) - signed(N) >= 0 &&
644           "Corrupted backtrack positions ?");
645    CachedLexPos -= N;
646  }
647
648  /// EnterToken - Enters a token in the token stream to be lexed next. If
649  /// BackTrack() is called afterwards, the token will remain at the insertion
650  /// point.
651  void EnterToken(const Token &Tok) {
652    EnterCachingLexMode();
653    CachedTokens.insert(CachedTokens.begin()+CachedLexPos, Tok);
654  }
655
656  /// AnnotateCachedTokens - We notify the Preprocessor that if it is caching
657  /// tokens (because backtrack is enabled) it should replace the most recent
658  /// cached tokens with the given annotation token. This function has no effect
659  /// if backtracking is not enabled.
660  ///
661  /// Note that the use of this function is just for optimization; so that the
662  /// cached tokens doesn't get re-parsed and re-resolved after a backtrack is
663  /// invoked.
664  void AnnotateCachedTokens(const Token &Tok) {
665    assert(Tok.isAnnotation() && "Expected annotation token");
666    if (CachedLexPos != 0 && isBacktrackEnabled())
667      AnnotatePreviousCachedTokens(Tok);
668  }
669
670  /// \brief Replace the last token with an annotation token.
671  ///
672  /// Like AnnotateCachedTokens(), this routine replaces an
673  /// already-parsed (and resolved) token with an annotation
674  /// token. However, this routine only replaces the last token with
675  /// the annotation token; it does not affect any other cached
676  /// tokens. This function has no effect if backtracking is not
677  /// enabled.
678  void ReplaceLastTokenWithAnnotation(const Token &Tok) {
679    assert(Tok.isAnnotation() && "Expected annotation token");
680    if (CachedLexPos != 0 && isBacktrackEnabled())
681      CachedTokens[CachedLexPos-1] = Tok;
682  }
683
684  /// \brief Specify the point at which code-completion will be performed.
685  ///
686  /// \param File the file in which code completion should occur. If
687  /// this file is included multiple times, code-completion will
688  /// perform completion the first time it is included. If NULL, this
689  /// function clears out the code-completion point.
690  ///
691  /// \param Line the line at which code completion should occur
692  /// (1-based).
693  ///
694  /// \param Column the column at which code completion should occur
695  /// (1-based).
696  ///
697  /// \returns true if an error occurred, false otherwise.
698  bool SetCodeCompletionPoint(const FileEntry *File,
699                              unsigned Line, unsigned Column);
700
701  /// \brief Determine if we are performing code completion.
702  bool isCodeCompletionEnabled() const { return CodeCompletionFile != 0; }
703
704  /// \brief Returns the location of the code-completion point.
705  /// Returns an invalid location if code-completion is not enabled or the file
706  /// containing the code-completion point has not been lexed yet.
707  SourceLocation getCodeCompletionLoc() const { return CodeCompletionLoc; }
708
709  /// \brief Returns the start location of the file of code-completion point.
710  /// Returns an invalid location if code-completion is not enabled or the file
711  /// containing the code-completion point has not been lexed yet.
712  SourceLocation getCodeCompletionFileLoc() const {
713    return CodeCompletionFileLoc;
714  }
715
716  /// \brief Returns true if code-completion is enabled and we have hit the
717  /// code-completion point.
718  bool isCodeCompletionReached() const { return CodeCompletionReached; }
719
720  /// \brief Note that we hit the code-completion point.
721  void setCodeCompletionReached() {
722    assert(isCodeCompletionEnabled() && "Code-completion not enabled!");
723    CodeCompletionReached = true;
724    // Silence any diagnostics that occur after we hit the code-completion.
725    getDiagnostics().setSuppressAllDiagnostics(true);
726  }
727
728  /// \brief The location of the currently-active #pragma clang
729  /// arc_cf_code_audited begin.  Returns an invalid location if there
730  /// is no such pragma active.
731  SourceLocation getPragmaARCCFCodeAuditedLoc() const {
732    return PragmaARCCFCodeAuditedLoc;
733  }
734
735  /// \brief Set the location of the currently-active #pragma clang
736  /// arc_cf_code_audited begin.  An invalid location ends the pragma.
737  void setPragmaARCCFCodeAuditedLoc(SourceLocation Loc) {
738    PragmaARCCFCodeAuditedLoc = Loc;
739  }
740
741  /// \brief Instruct the preprocessor to skip part of the main
742  /// the main source file.
743  ///
744  /// \brief Bytes The number of bytes in the preamble to skip.
745  ///
746  /// \brief StartOfLine Whether skipping these bytes puts the lexer at the
747  /// start of a line.
748  void setSkipMainFilePreamble(unsigned Bytes, bool StartOfLine) {
749    SkipMainFilePreamble.first = Bytes;
750    SkipMainFilePreamble.second = StartOfLine;
751  }
752
753  /// Diag - Forwarding function for diagnostics.  This emits a diagnostic at
754  /// the specified Token's location, translating the token's start
755  /// position in the current buffer into a SourcePosition object for rendering.
756  DiagnosticBuilder Diag(SourceLocation Loc, unsigned DiagID) const {
757    return Diags->Report(Loc, DiagID);
758  }
759
760  DiagnosticBuilder Diag(const Token &Tok, unsigned DiagID) const {
761    return Diags->Report(Tok.getLocation(), DiagID);
762  }
763
764  /// getSpelling() - Return the 'spelling' of the token at the given
765  /// location; does not go up to the spelling location or down to the
766  /// expansion location.
767  ///
768  /// \param buffer A buffer which will be used only if the token requires
769  ///   "cleaning", e.g. if it contains trigraphs or escaped newlines
770  /// \param invalid If non-null, will be set \c true if an error occurs.
771  StringRef getSpelling(SourceLocation loc,
772                              SmallVectorImpl<char> &buffer,
773                              bool *invalid = 0) const {
774    return Lexer::getSpelling(loc, buffer, SourceMgr, Features, invalid);
775  }
776
777  /// getSpelling() - Return the 'spelling' of the Tok token.  The spelling of a
778  /// token is the characters used to represent the token in the source file
779  /// after trigraph expansion and escaped-newline folding.  In particular, this
780  /// wants to get the true, uncanonicalized, spelling of things like digraphs
781  /// UCNs, etc.
782  ///
783  /// \param Invalid If non-null, will be set \c true if an error occurs.
784  std::string getSpelling(const Token &Tok, bool *Invalid = 0) const {
785    return Lexer::getSpelling(Tok, SourceMgr, Features, Invalid);
786  }
787
788  /// getSpelling - This method is used to get the spelling of a token into a
789  /// preallocated buffer, instead of as an std::string.  The caller is required
790  /// to allocate enough space for the token, which is guaranteed to be at least
791  /// Tok.getLength() bytes long.  The length of the actual result is returned.
792  ///
793  /// Note that this method may do two possible things: it may either fill in
794  /// the buffer specified with characters, or it may *change the input pointer*
795  /// to point to a constant buffer with the data already in it (avoiding a
796  /// copy).  The caller is not allowed to modify the returned buffer pointer
797  /// if an internal buffer is returned.
798  unsigned getSpelling(const Token &Tok, const char *&Buffer,
799                       bool *Invalid = 0) const {
800    return Lexer::getSpelling(Tok, Buffer, SourceMgr, Features, Invalid);
801  }
802
803  /// getSpelling - This method is used to get the spelling of a token into a
804  /// SmallVector. Note that the returned StringRef may not point to the
805  /// supplied buffer if a copy can be avoided.
806  StringRef getSpelling(const Token &Tok,
807                              SmallVectorImpl<char> &Buffer,
808                              bool *Invalid = 0) const;
809
810  /// getSpellingOfSingleCharacterNumericConstant - Tok is a numeric constant
811  /// with length 1, return the character.
812  char getSpellingOfSingleCharacterNumericConstant(const Token &Tok,
813                                                   bool *Invalid = 0) const {
814    assert(Tok.is(tok::numeric_constant) &&
815           Tok.getLength() == 1 && "Called on unsupported token");
816    assert(!Tok.needsCleaning() && "Token can't need cleaning with length 1");
817
818    // If the token is carrying a literal data pointer, just use it.
819    if (const char *D = Tok.getLiteralData())
820      return *D;
821
822    // Otherwise, fall back on getCharacterData, which is slower, but always
823    // works.
824    return *SourceMgr.getCharacterData(Tok.getLocation(), Invalid);
825  }
826
827  /// CreateString - Plop the specified string into a scratch buffer and set the
828  /// specified token's location and length to it.  If specified, the source
829  /// location provides a location of the expansion point of the token.
830  void CreateString(const char *Buf, unsigned Len, Token &Tok,
831                    SourceLocation ExpansionLocStart = SourceLocation(),
832                    SourceLocation ExpansionLocEnd = SourceLocation());
833
834  /// \brief Computes the source location just past the end of the
835  /// token at this source location.
836  ///
837  /// This routine can be used to produce a source location that
838  /// points just past the end of the token referenced by \p Loc, and
839  /// is generally used when a diagnostic needs to point just after a
840  /// token where it expected something different that it received. If
841  /// the returned source location would not be meaningful (e.g., if
842  /// it points into a macro), this routine returns an invalid
843  /// source location.
844  ///
845  /// \param Offset an offset from the end of the token, where the source
846  /// location should refer to. The default offset (0) produces a source
847  /// location pointing just past the end of the token; an offset of 1 produces
848  /// a source location pointing to the last character in the token, etc.
849  SourceLocation getLocForEndOfToken(SourceLocation Loc, unsigned Offset = 0) {
850    return Lexer::getLocForEndOfToken(Loc, Offset, SourceMgr, Features);
851  }
852
853  /// \brief Returns true if the given MacroID location points at the first
854  /// token of the macro expansion.
855  bool isAtStartOfMacroExpansion(SourceLocation loc) const {
856    return Lexer::isAtStartOfMacroExpansion(loc, SourceMgr, Features);
857  }
858
859  /// \brief Returns true if the given MacroID location points at the last
860  /// token of the macro expansion.
861  bool isAtEndOfMacroExpansion(SourceLocation loc) const {
862    return Lexer::isAtEndOfMacroExpansion(loc, SourceMgr, Features);
863  }
864
865  /// DumpToken - Print the token to stderr, used for debugging.
866  ///
867  void DumpToken(const Token &Tok, bool DumpFlags = false) const;
868  void DumpLocation(SourceLocation Loc) const;
869  void DumpMacro(const MacroInfo &MI) const;
870
871  /// AdvanceToTokenCharacter - Given a location that specifies the start of a
872  /// token, return a new location that specifies a character within the token.
873  SourceLocation AdvanceToTokenCharacter(SourceLocation TokStart,
874                                         unsigned Char) const {
875    return Lexer::AdvanceToTokenCharacter(TokStart, Char, SourceMgr, Features);
876  }
877
878  /// IncrementPasteCounter - Increment the counters for the number of token
879  /// paste operations performed.  If fast was specified, this is a 'fast paste'
880  /// case we handled.
881  ///
882  void IncrementPasteCounter(bool isFast) {
883    if (isFast)
884      ++NumFastTokenPaste;
885    else
886      ++NumTokenPaste;
887  }
888
889  void PrintStats();
890
891  size_t getTotalMemory() const;
892
893  /// HandleMicrosoftCommentPaste - When the macro expander pastes together a
894  /// comment (/##/) in microsoft mode, this method handles updating the current
895  /// state, returning the token on the next source line.
896  void HandleMicrosoftCommentPaste(Token &Tok);
897
898  //===--------------------------------------------------------------------===//
899  // Preprocessor callback methods.  These are invoked by a lexer as various
900  // directives and events are found.
901
902  /// LookUpIdentifierInfo - Given a tok::raw_identifier token, look up the
903  /// identifier information for the token and install it into the token,
904  /// updating the token kind accordingly.
905  IdentifierInfo *LookUpIdentifierInfo(Token &Identifier) const;
906
907private:
908  llvm::DenseMap<IdentifierInfo*,unsigned> PoisonReasons;
909
910public:
911
912  // SetPoisonReason - Call this function to indicate the reason for
913  // poisoning an identifier. If that identifier is accessed while
914  // poisoned, then this reason will be used instead of the default
915  // "poisoned" diagnostic.
916  void SetPoisonReason(IdentifierInfo *II, unsigned DiagID);
917
918  // HandlePoisonedIdentifier - Display reason for poisoned
919  // identifier.
920  void HandlePoisonedIdentifier(Token & Tok);
921
922  void MaybeHandlePoisonedIdentifier(Token & Identifier) {
923    if(IdentifierInfo * II = Identifier.getIdentifierInfo()) {
924      if(II->isPoisoned()) {
925        HandlePoisonedIdentifier(Identifier);
926      }
927    }
928  }
929
930private:
931  /// Identifiers used for SEH handling in Borland. These are only
932  /// allowed in particular circumstances
933  // __except block
934  IdentifierInfo *Ident__exception_code,
935                 *Ident___exception_code,
936                 *Ident_GetExceptionCode;
937  // __except filter expression
938  IdentifierInfo *Ident__exception_info,
939                 *Ident___exception_info,
940                 *Ident_GetExceptionInfo;
941  // __finally
942  IdentifierInfo *Ident__abnormal_termination,
943                 *Ident___abnormal_termination,
944                 *Ident_AbnormalTermination;
945public:
946  void PoisonSEHIdentifiers(bool Poison = true); // Borland
947
948  /// HandleIdentifier - This callback is invoked when the lexer reads an
949  /// identifier and has filled in the tokens IdentifierInfo member.  This
950  /// callback potentially macro expands it or turns it into a named token (like
951  /// 'for').
952  void HandleIdentifier(Token &Identifier);
953
954
955  /// HandleEndOfFile - This callback is invoked when the lexer hits the end of
956  /// the current file.  This either returns the EOF token and returns true, or
957  /// pops a level off the include stack and returns false, at which point the
958  /// client should call lex again.
959  bool HandleEndOfFile(Token &Result, bool isEndOfMacro = false);
960
961  /// HandleEndOfTokenLexer - This callback is invoked when the current
962  /// TokenLexer hits the end of its token stream.
963  bool HandleEndOfTokenLexer(Token &Result);
964
965  /// HandleDirective - This callback is invoked when the lexer sees a # token
966  /// at the start of a line.  This consumes the directive, modifies the
967  /// lexer/preprocessor state, and advances the lexer(s) so that the next token
968  /// read is the correct one.
969  void HandleDirective(Token &Result);
970
971  /// CheckEndOfDirective - Ensure that the next token is a tok::eod token.  If
972  /// not, emit a diagnostic and consume up until the eod.  If EnableMacros is
973  /// true, then we consider macros that expand to zero tokens as being ok.
974  void CheckEndOfDirective(const char *Directive, bool EnableMacros = false);
975
976  /// DiscardUntilEndOfDirective - Read and discard all tokens remaining on the
977  /// current line until the tok::eod token is found.
978  void DiscardUntilEndOfDirective();
979
980  /// SawDateOrTime - This returns true if the preprocessor has seen a use of
981  /// __DATE__ or __TIME__ in the file so far.
982  bool SawDateOrTime() const {
983    return DATELoc != SourceLocation() || TIMELoc != SourceLocation();
984  }
985  unsigned getCounterValue() const { return CounterValue; }
986  void setCounterValue(unsigned V) { CounterValue = V; }
987
988  /// AllocateMacroInfo - Allocate a new MacroInfo object with the provide
989  ///  SourceLocation.
990  MacroInfo *AllocateMacroInfo(SourceLocation L);
991
992  /// CloneMacroInfo - Allocate a new MacroInfo object which is clone of MI.
993  MacroInfo *CloneMacroInfo(const MacroInfo &MI);
994
995  /// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully
996  /// checked and spelled filename, e.g. as an operand of #include. This returns
997  /// true if the input filename was in <>'s or false if it were in ""'s.  The
998  /// caller is expected to provide a buffer that is large enough to hold the
999  /// spelling of the filename, but is also expected to handle the case when
1000  /// this method decides to use a different buffer.
1001  bool GetIncludeFilenameSpelling(SourceLocation Loc,StringRef &Filename);
1002
1003  /// LookupFile - Given a "foo" or <foo> reference, look up the indicated file,
1004  /// return null on failure.  isAngled indicates whether the file reference is
1005  /// for system #include's or not (i.e. using <> instead of "").
1006  const FileEntry *LookupFile(StringRef Filename,
1007                              bool isAngled, const DirectoryLookup *FromDir,
1008                              const DirectoryLookup *&CurDir,
1009                              SmallVectorImpl<char> *SearchPath,
1010                              SmallVectorImpl<char> *RelativePath,
1011                              ModuleMap::Module **SuggestedModule,
1012                              bool SkipCache = false);
1013
1014  /// GetCurLookup - The DirectoryLookup structure used to find the current
1015  /// FileEntry, if CurLexer is non-null and if applicable.  This allows us to
1016  /// implement #include_next and find directory-specific properties.
1017  const DirectoryLookup *GetCurDirLookup() { return CurDirLookup; }
1018
1019  /// isInPrimaryFile - Return true if we're in the top-level file, not in a
1020  /// #include.
1021  bool isInPrimaryFile() const;
1022
1023  /// ConcatenateIncludeName - Handle cases where the #include name is expanded
1024  /// from a macro as multiple tokens, which need to be glued together.  This
1025  /// occurs for code like:
1026  ///    #define FOO <a/b.h>
1027  ///    #include FOO
1028  /// because in this case, "<a/b.h>" is returned as 7 tokens, not one.
1029  ///
1030  /// This code concatenates and consumes tokens up to the '>' token.  It
1031  /// returns false if the > was found, otherwise it returns true if it finds
1032  /// and consumes the EOD marker.
1033  bool ConcatenateIncludeName(llvm::SmallString<128> &FilenameBuffer,
1034                              SourceLocation &End);
1035
1036  /// LexOnOffSwitch - Lex an on-off-switch (C99 6.10.6p2) and verify that it is
1037  /// followed by EOD.  Return true if the token is not a valid on-off-switch.
1038  bool LexOnOffSwitch(tok::OnOffSwitch &OOS);
1039
1040private:
1041
1042  void PushIncludeMacroStack() {
1043    IncludeMacroStack.push_back(IncludeStackInfo(CurLexerKind,
1044                                                 CurLexer.take(),
1045                                                 CurPTHLexer.take(),
1046                                                 CurPPLexer,
1047                                                 CurTokenLexer.take(),
1048                                                 CurDirLookup));
1049    CurPPLexer = 0;
1050  }
1051
1052  void PopIncludeMacroStack() {
1053    CurLexer.reset(IncludeMacroStack.back().TheLexer);
1054    CurPTHLexer.reset(IncludeMacroStack.back().ThePTHLexer);
1055    CurPPLexer = IncludeMacroStack.back().ThePPLexer;
1056    CurTokenLexer.reset(IncludeMacroStack.back().TheTokenLexer);
1057    CurDirLookup  = IncludeMacroStack.back().TheDirLookup;
1058    CurLexerKind = IncludeMacroStack.back().CurLexerKind;
1059    IncludeMacroStack.pop_back();
1060  }
1061
1062  /// AllocateMacroInfo - Allocate a new MacroInfo object.
1063  MacroInfo *AllocateMacroInfo();
1064
1065  /// ReleaseMacroInfo - Release the specified MacroInfo.  This memory will
1066  ///  be reused for allocating new MacroInfo objects.
1067  void ReleaseMacroInfo(MacroInfo* MI);
1068
1069  /// ReadMacroName - Lex and validate a macro name, which occurs after a
1070  /// #define or #undef.  This emits a diagnostic, sets the token kind to eod,
1071  /// and discards the rest of the macro line if the macro name is invalid.
1072  void ReadMacroName(Token &MacroNameTok, char isDefineUndef = 0);
1073
1074  /// ReadMacroDefinitionArgList - The ( starting an argument list of a macro
1075  /// definition has just been read.  Lex the rest of the arguments and the
1076  /// closing ), updating MI with what we learn.  Return true if an error occurs
1077  /// parsing the arg list.
1078  bool ReadMacroDefinitionArgList(MacroInfo *MI);
1079
1080  /// SkipExcludedConditionalBlock - We just read a #if or related directive and
1081  /// decided that the subsequent tokens are in the #if'd out portion of the
1082  /// file.  Lex the rest of the file, until we see an #endif.  If
1083  /// FoundNonSkipPortion is true, then we have already emitted code for part of
1084  /// this #if directive, so #else/#elif blocks should never be entered. If
1085  /// FoundElse is false, then #else directives are ok, if not, then we have
1086  /// already seen one so a #else directive is a duplicate.  When this returns,
1087  /// the caller can lex the first valid token.
1088  void SkipExcludedConditionalBlock(SourceLocation IfTokenLoc,
1089                                    bool FoundNonSkipPortion, bool FoundElse,
1090                                    SourceLocation ElseLoc = SourceLocation());
1091
1092  /// PTHSkipExcludedConditionalBlock - A fast PTH version of
1093  ///  SkipExcludedConditionalBlock.
1094  void PTHSkipExcludedConditionalBlock();
1095
1096  /// EvaluateDirectiveExpression - Evaluate an integer constant expression that
1097  /// may occur after a #if or #elif directive and return it as a bool.  If the
1098  /// expression is equivalent to "!defined(X)" return X in IfNDefMacro.
1099  bool EvaluateDirectiveExpression(IdentifierInfo *&IfNDefMacro);
1100
1101  /// RegisterBuiltinPragmas - Install the standard preprocessor pragmas:
1102  /// #pragma GCC poison/system_header/dependency and #pragma once.
1103  void RegisterBuiltinPragmas();
1104
1105  /// RegisterBuiltinMacros - Register builtin macros, such as __LINE__ with the
1106  /// identifier table.
1107  void RegisterBuiltinMacros();
1108
1109  /// HandleMacroExpandedIdentifier - If an identifier token is read that is to
1110  /// be expanded as a macro, handle it and return the next token as 'Tok'.  If
1111  /// the macro should not be expanded return true, otherwise return false.
1112  bool HandleMacroExpandedIdentifier(Token &Tok, MacroInfo *MI);
1113
1114  /// \brief Cache macro expanded tokens for TokenLexers.
1115  //
1116  /// Works like a stack; a TokenLexer adds the macro expanded tokens that is
1117  /// going to lex in the cache and when it finishes the tokens are removed
1118  /// from the end of the cache.
1119  Token *cacheMacroExpandedTokens(TokenLexer *tokLexer,
1120                                  ArrayRef<Token> tokens);
1121  void removeCachedMacroExpandedTokensOfLastLexer();
1122  friend void TokenLexer::ExpandFunctionArguments();
1123
1124  /// isNextPPTokenLParen - Determine whether the next preprocessor token to be
1125  /// lexed is a '('.  If so, consume the token and return true, if not, this
1126  /// method should have no observable side-effect on the lexed tokens.
1127  bool isNextPPTokenLParen();
1128
1129  /// ReadFunctionLikeMacroArgs - After reading "MACRO(", this method is
1130  /// invoked to read all of the formal arguments specified for the macro
1131  /// invocation.  This returns null on error.
1132  MacroArgs *ReadFunctionLikeMacroArgs(Token &MacroName, MacroInfo *MI,
1133                                       SourceLocation &ExpansionEnd);
1134
1135  /// ExpandBuiltinMacro - If an identifier token is read that is to be expanded
1136  /// as a builtin macro, handle it and return the next token as 'Tok'.
1137  void ExpandBuiltinMacro(Token &Tok);
1138
1139  /// Handle_Pragma - Read a _Pragma directive, slice it up, process it, then
1140  /// return the first token after the directive.  The _Pragma token has just
1141  /// been read into 'Tok'.
1142  void Handle_Pragma(Token &Tok);
1143
1144  /// HandleMicrosoft__pragma - Like Handle_Pragma except the pragma text
1145  /// is not enclosed within a string literal.
1146  void HandleMicrosoft__pragma(Token &Tok);
1147
1148  /// EnterSourceFileWithLexer - Add a lexer to the top of the include stack and
1149  /// start lexing tokens from it instead of the current buffer.
1150  void EnterSourceFileWithLexer(Lexer *TheLexer, const DirectoryLookup *Dir);
1151
1152  /// EnterSourceFileWithPTH - Add a lexer to the top of the include stack and
1153  /// start getting tokens from it using the PTH cache.
1154  void EnterSourceFileWithPTH(PTHLexer *PL, const DirectoryLookup *Dir);
1155
1156  /// IsFileLexer - Returns true if we are lexing from a file and not a
1157  ///  pragma or a macro.
1158  static bool IsFileLexer(const Lexer* L, const PreprocessorLexer* P) {
1159    return L ? !L->isPragmaLexer() : P != 0;
1160  }
1161
1162  static bool IsFileLexer(const IncludeStackInfo& I) {
1163    return IsFileLexer(I.TheLexer, I.ThePPLexer);
1164  }
1165
1166  bool IsFileLexer() const {
1167    return IsFileLexer(CurLexer.get(), CurPPLexer);
1168  }
1169
1170  //===--------------------------------------------------------------------===//
1171  // Caching stuff.
1172  void CachingLex(Token &Result);
1173  bool InCachingLexMode() const {
1174    // If the Lexer pointers are 0 and IncludeMacroStack is empty, it means
1175    // that we are past EOF, not that we are in CachingLex mode.
1176    return CurPPLexer == 0 && CurTokenLexer == 0 && CurPTHLexer == 0 &&
1177           !IncludeMacroStack.empty();
1178  }
1179  void EnterCachingLexMode();
1180  void ExitCachingLexMode() {
1181    if (InCachingLexMode())
1182      RemoveTopOfLexerStack();
1183  }
1184  const Token &PeekAhead(unsigned N);
1185  void AnnotatePreviousCachedTokens(const Token &Tok);
1186
1187  //===--------------------------------------------------------------------===//
1188  /// Handle*Directive - implement the various preprocessor directives.  These
1189  /// should side-effect the current preprocessor object so that the next call
1190  /// to Lex() will return the appropriate token next.
1191  void HandleLineDirective(Token &Tok);
1192  void HandleDigitDirective(Token &Tok);
1193  void HandleUserDiagnosticDirective(Token &Tok, bool isWarning);
1194  void HandleIdentSCCSDirective(Token &Tok);
1195  void HandleMacroExportDirective(Token &Tok);
1196  void HandleMacroPrivateDirective(Token &Tok);
1197
1198  // File inclusion.
1199  void HandleIncludeDirective(SourceLocation HashLoc,
1200                              Token &Tok,
1201                              const DirectoryLookup *LookupFrom = 0,
1202                              bool isImport = false);
1203  void HandleIncludeNextDirective(SourceLocation HashLoc, Token &Tok);
1204  void HandleIncludeMacrosDirective(SourceLocation HashLoc, Token &Tok);
1205  void HandleImportDirective(SourceLocation HashLoc, Token &Tok);
1206
1207  // Macro handling.
1208  void HandleDefineDirective(Token &Tok);
1209  void HandleUndefDirective(Token &Tok);
1210
1211  // Conditional Inclusion.
1212  void HandleIfdefDirective(Token &Tok, bool isIfndef,
1213                            bool ReadAnyTokensBeforeDirective);
1214  void HandleIfDirective(Token &Tok, bool ReadAnyTokensBeforeDirective);
1215  void HandleEndifDirective(Token &Tok);
1216  void HandleElseDirective(Token &Tok);
1217  void HandleElifDirective(Token &Tok);
1218
1219  // Pragmas.
1220  void HandlePragmaDirective(unsigned Introducer);
1221public:
1222  void HandlePragmaOnce(Token &OnceTok);
1223  void HandlePragmaMark();
1224  void HandlePragmaPoison(Token &PoisonTok);
1225  void HandlePragmaSystemHeader(Token &SysHeaderTok);
1226  void HandlePragmaDependency(Token &DependencyTok);
1227  void HandlePragmaComment(Token &CommentTok);
1228  void HandlePragmaMessage(Token &MessageTok);
1229  void HandlePragmaPushMacro(Token &Tok);
1230  void HandlePragmaPopMacro(Token &Tok);
1231  IdentifierInfo *ParsePragmaPushOrPopMacro(Token &Tok);
1232
1233  // Return true and store the first token only if any CommentHandler
1234  // has inserted some tokens and getCommentRetentionState() is false.
1235  bool HandleComment(Token &Token, SourceRange Comment);
1236
1237  /// \brief A macro is used, update information about macros that need unused
1238  /// warnings.
1239  void markMacroAsUsed(MacroInfo *MI);
1240};
1241
1242/// \brief Abstract base class that describes a handler that will receive
1243/// source ranges for each of the comments encountered in the source file.
1244class CommentHandler {
1245public:
1246  virtual ~CommentHandler();
1247
1248  // The handler shall return true if it has pushed any tokens
1249  // to be read using e.g. EnterToken or EnterTokenStream.
1250  virtual bool HandleComment(Preprocessor &PP, SourceRange Comment) = 0;
1251};
1252
1253}  // end namespace clang
1254
1255#endif
1256