Preprocessor.h revision fba18aa8f2cd1994dc65e8cb9f4be201c560dc0b
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  bool AutoModuleImport : 1;
110
111  // State that changes while the preprocessor runs:
112  bool InMacroArgs : 1;            // True if parsing fn macro invocation args.
113
114  /// Whether the preprocessor owns the header search object.
115  bool OwnsHeaderSearch : 1;
116
117  /// DisableMacroExpansion - True if macro expansion is disabled.
118  bool DisableMacroExpansion : 1;
119
120  /// \brief Whether we have already loaded macros from the external source.
121  mutable bool ReadMacrosFromExternalSource : 1;
122
123  /// Identifiers - This is mapping/lookup information for all identifiers in
124  /// the program, including program keywords.
125  mutable IdentifierTable Identifiers;
126
127  /// Selectors - This table contains all the selectors in the program. Unlike
128  /// IdentifierTable above, this table *isn't* populated by the preprocessor.
129  /// It is declared/expanded here because it's role/lifetime is
130  /// conceptually similar the IdentifierTable. In addition, the current control
131  /// flow (in clang::ParseAST()), make it convenient to put here.
132  /// FIXME: Make sure the lifetime of Identifiers/Selectors *isn't* tied to
133  /// the lifetime of the preprocessor.
134  SelectorTable Selectors;
135
136  /// BuiltinInfo - Information about builtins.
137  Builtin::Context BuiltinInfo;
138
139  /// PragmaHandlers - This tracks all of the pragmas that the client registered
140  /// with this preprocessor.
141  PragmaNamespace *PragmaHandlers;
142
143  /// \brief Tracks all of the comment handlers that the client registered
144  /// with this preprocessor.
145  std::vector<CommentHandler *> CommentHandlers;
146
147  /// \brief The code-completion handler.
148  CodeCompletionHandler *CodeComplete;
149
150  /// \brief The file that we're performing code-completion for, if any.
151  const FileEntry *CodeCompletionFile;
152
153  /// \brief The offset in file for the code-completion point.
154  unsigned CodeCompletionOffset;
155
156  /// \brief The location for the code-completion point. This gets instantiated
157  /// when the CodeCompletionFile gets #include'ed for preprocessing.
158  SourceLocation CodeCompletionLoc;
159
160  /// \brief The start location for the file of the code-completion point.
161  /// This gets instantiated when the CodeCompletionFile gets #include'ed
162  /// for preprocessing.
163  SourceLocation CodeCompletionFileLoc;
164
165  /// \brief The source location of the __import_module__ keyword we just
166  /// lexed, if any.
167  SourceLocation ModuleImportLoc;
168
169  /// \brief True if we hit the code-completion point.
170  bool CodeCompletionReached;
171
172  /// \brief The number of bytes that we will initially skip when entering the
173  /// main file, which is used when loading a precompiled preamble, along
174  /// with a flag that indicates whether skipping this number of bytes will
175  /// place the lexer at the start of a line.
176  std::pair<unsigned, bool> SkipMainFilePreamble;
177
178  /// CurLexer - This is the current top of the stack that we're lexing from if
179  /// not expanding a macro and we are lexing directly from source code.
180  ///  Only one of CurLexer, CurPTHLexer, or CurTokenLexer will be non-null.
181  llvm::OwningPtr<Lexer> CurLexer;
182
183  /// CurPTHLexer - This is the current top of stack that we're lexing from if
184  ///  not expanding from a macro and we are lexing from a PTH cache.
185  ///  Only one of CurLexer, CurPTHLexer, or CurTokenLexer will be non-null.
186  llvm::OwningPtr<PTHLexer> CurPTHLexer;
187
188  /// CurPPLexer - This is the current top of the stack what we're lexing from
189  ///  if not expanding a macro.  This is an alias for either CurLexer or
190  ///  CurPTHLexer.
191  PreprocessorLexer *CurPPLexer;
192
193  /// CurLookup - The DirectoryLookup structure used to find the current
194  /// FileEntry, if CurLexer is non-null and if applicable.  This allows us to
195  /// implement #include_next and find directory-specific properties.
196  const DirectoryLookup *CurDirLookup;
197
198  /// CurTokenLexer - This is the current macro we are expanding, if we are
199  /// expanding a macro.  One of CurLexer and CurTokenLexer must be null.
200  llvm::OwningPtr<TokenLexer> CurTokenLexer;
201
202  /// \brief The kind of lexer we're currently working with.
203  enum CurLexerKind {
204    CLK_Lexer,
205    CLK_PTHLexer,
206    CLK_TokenLexer,
207    CLK_CachingLexer,
208    CLK_LexAfterModuleImport
209  } CurLexerKind;
210
211  /// IncludeMacroStack - This keeps track of the stack of files currently
212  /// #included, and macros currently being expanded from, not counting
213  /// CurLexer/CurTokenLexer.
214  struct IncludeStackInfo {
215    enum CurLexerKind     CurLexerKind;
216    Lexer                 *TheLexer;
217    PTHLexer              *ThePTHLexer;
218    PreprocessorLexer     *ThePPLexer;
219    TokenLexer            *TheTokenLexer;
220    const DirectoryLookup *TheDirLookup;
221
222    IncludeStackInfo(enum CurLexerKind K, Lexer *L, PTHLexer* P,
223                     PreprocessorLexer* PPL,
224                     TokenLexer* TL, const DirectoryLookup *D)
225      : CurLexerKind(K), TheLexer(L), ThePTHLexer(P), ThePPLexer(PPL),
226        TheTokenLexer(TL), TheDirLookup(D) {}
227  };
228  std::vector<IncludeStackInfo> IncludeMacroStack;
229
230  /// Callbacks - These are actions invoked when some preprocessor activity is
231  /// encountered (e.g. a file is #included, etc).
232  PPCallbacks *Callbacks;
233
234  /// Macros - For each IdentifierInfo with 'HasMacro' set, we keep a mapping
235  /// to the actual definition of the macro.
236  llvm::DenseMap<IdentifierInfo*, MacroInfo*> Macros;
237
238  /// \brief Macros that we want to warn because they are not used at the end
239  /// of the translation unit; we store just their SourceLocations instead
240  /// something like MacroInfo*. The benefit of this is that when we are
241  /// deserializing from PCH, we don't need to deserialize identifier & macros
242  /// just so that we can report that they are unused, we just warn using
243  /// the SourceLocations of this set (that will be filled by the ASTReader).
244  /// We are using SmallPtrSet instead of a vector for faster removal.
245  typedef llvm::SmallPtrSet<SourceLocation, 32> WarnUnusedMacroLocsTy;
246  WarnUnusedMacroLocsTy WarnUnusedMacroLocs;
247
248  /// MacroArgCache - This is a "freelist" of MacroArg objects that can be
249  /// reused for quick allocation.
250  MacroArgs *MacroArgCache;
251  friend class MacroArgs;
252
253  /// PragmaPushMacroInfo - For each IdentifierInfo used in a #pragma
254  /// push_macro directive, we keep a MacroInfo stack used to restore
255  /// previous macro value.
256  llvm::DenseMap<IdentifierInfo*, std::vector<MacroInfo*> > PragmaPushMacroInfo;
257
258  // Various statistics we track for performance analysis.
259  unsigned NumDirectives, NumIncluded, NumDefined, NumUndefined, NumPragma;
260  unsigned NumIf, NumElse, NumEndif;
261  unsigned NumEnteredSourceFiles, MaxIncludeStackDepth;
262  unsigned NumMacroExpanded, NumFnMacroExpanded, NumBuiltinMacroExpanded;
263  unsigned NumFastMacroExpanded, NumTokenPaste, NumFastTokenPaste;
264  unsigned NumSkipped;
265
266  /// Predefines - This string is the predefined macros that preprocessor
267  /// should use from the command line etc.
268  std::string Predefines;
269
270  /// TokenLexerCache - Cache macro expanders to reduce malloc traffic.
271  enum { TokenLexerCacheSize = 8 };
272  unsigned NumCachedTokenLexers;
273  TokenLexer *TokenLexerCache[TokenLexerCacheSize];
274
275  /// \brief Keeps macro expanded tokens for TokenLexers.
276  //
277  /// Works like a stack; a TokenLexer adds the macro expanded tokens that is
278  /// going to lex in the cache and when it finishes the tokens are removed
279  /// from the end of the cache.
280  SmallVector<Token, 16> MacroExpandedTokens;
281  std::vector<std::pair<TokenLexer *, size_t> > MacroExpandingLexersStack;
282
283  /// \brief A record of the macro definitions and expansions that
284  /// occurred during preprocessing.
285  ///
286  /// This is an optional side structure that can be enabled with
287  /// \c createPreprocessingRecord() prior to preprocessing.
288  PreprocessingRecord *Record;
289
290private:  // Cached tokens state.
291  typedef SmallVector<Token, 1> CachedTokensTy;
292
293  /// CachedTokens - Cached tokens are stored here when we do backtracking or
294  /// lookahead. They are "lexed" by the CachingLex() method.
295  CachedTokensTy CachedTokens;
296
297  /// CachedLexPos - The position of the cached token that CachingLex() should
298  /// "lex" next. If it points beyond the CachedTokens vector, it means that
299  /// a normal Lex() should be invoked.
300  CachedTokensTy::size_type CachedLexPos;
301
302  /// BacktrackPositions - Stack of backtrack positions, allowing nested
303  /// backtracks. The EnableBacktrackAtThisPos() method pushes a position to
304  /// indicate where CachedLexPos should be set when the BackTrack() method is
305  /// invoked (at which point the last position is popped).
306  std::vector<CachedTokensTy::size_type> BacktrackPositions;
307
308  struct MacroInfoChain {
309    MacroInfo MI;
310    MacroInfoChain *Next;
311    MacroInfoChain *Prev;
312  };
313
314  /// MacroInfos are managed as a chain for easy disposal.  This is the head
315  /// of that list.
316  MacroInfoChain *MIChainHead;
317
318  /// MICache - A "freelist" of MacroInfo objects that can be reused for quick
319  /// allocation.
320  MacroInfoChain *MICache;
321
322  MacroInfo *getInfoForMacro(IdentifierInfo *II) const;
323
324public:
325  Preprocessor(Diagnostic &diags, LangOptions &opts,
326               const TargetInfo *target,
327               SourceManager &SM, HeaderSearch &Headers,
328               ModuleLoader &TheModuleLoader,
329               IdentifierInfoLookup *IILookup = 0,
330               bool OwnsHeaderSearch = false,
331               bool DelayInitialization = false);
332
333  ~Preprocessor();
334
335  /// \brief Initialize the preprocessor, if the constructor did not already
336  /// perform the initialization.
337  ///
338  /// \param Target Information about the target.
339  void Initialize(const TargetInfo &Target);
340
341  Diagnostic &getDiagnostics() const { return *Diags; }
342  void setDiagnostics(Diagnostic &D) { Diags = &D; }
343
344  const LangOptions &getLangOptions() const { return Features; }
345  const TargetInfo &getTargetInfo() const { return *Target; }
346  FileManager &getFileManager() const { return FileMgr; }
347  SourceManager &getSourceManager() const { return SourceMgr; }
348  HeaderSearch &getHeaderSearchInfo() const { return HeaderInfo; }
349
350  IdentifierTable &getIdentifierTable() { return Identifiers; }
351  SelectorTable &getSelectorTable() { return Selectors; }
352  Builtin::Context &getBuiltinInfo() { return BuiltinInfo; }
353  llvm::BumpPtrAllocator &getPreprocessorAllocator() { return BP; }
354
355  void setPTHManager(PTHManager* pm);
356
357  PTHManager *getPTHManager() { return PTH.get(); }
358
359  void setExternalSource(ExternalPreprocessorSource *Source) {
360    ExternalSource = Source;
361  }
362
363  ExternalPreprocessorSource *getExternalSource() const {
364    return ExternalSource;
365  }
366
367  /// \brief Retrieve the module loader associated with this preprocessor.
368  ModuleLoader &getModuleLoader() const { return TheModuleLoader; }
369
370  /// SetCommentRetentionState - Control whether or not the preprocessor retains
371  /// comments in output.
372  void SetCommentRetentionState(bool KeepComments, bool KeepMacroComments) {
373    this->KeepComments = KeepComments | KeepMacroComments;
374    this->KeepMacroComments = KeepMacroComments;
375  }
376
377  bool getCommentRetentionState() const { return KeepComments; }
378
379  void SetSuppressIncludeNotFoundError(bool Suppress) {
380    SuppressIncludeNotFoundError = Suppress;
381  }
382
383  bool GetSuppressIncludeNotFoundError() {
384    return SuppressIncludeNotFoundError;
385  }
386
387  /// \brief Specify whether automatic module imports are enabled.
388  void setAutoModuleImport(bool AutoModuleImport = true) {
389    this->AutoModuleImport = AutoModuleImport;
390  }
391
392  /// isCurrentLexer - Return true if we are lexing directly from the specified
393  /// lexer.
394  bool isCurrentLexer(const PreprocessorLexer *L) const {
395    return CurPPLexer == L;
396  }
397
398  /// getCurrentLexer - Return the current lexer being lexed from.  Note
399  /// that this ignores any potentially active macro expansions and _Pragma
400  /// expansions going on at the time.
401  PreprocessorLexer *getCurrentLexer() const { return CurPPLexer; }
402
403  /// getCurrentFileLexer - Return the current file lexer being lexed from.
404  /// Note that this ignores any potentially active macro expansions and _Pragma
405  /// expansions going on at the time.
406  PreprocessorLexer *getCurrentFileLexer() const;
407
408  /// getPPCallbacks/addPPCallbacks - Accessors for preprocessor callbacks.
409  /// Note that this class takes ownership of any PPCallbacks object given to
410  /// it.
411  PPCallbacks *getPPCallbacks() const { return Callbacks; }
412  void addPPCallbacks(PPCallbacks *C) {
413    if (Callbacks)
414      C = new PPChainedCallbacks(C, Callbacks);
415    Callbacks = C;
416  }
417
418  /// getMacroInfo - Given an identifier, return the MacroInfo it is #defined to
419  /// or null if it isn't #define'd.
420  MacroInfo *getMacroInfo(IdentifierInfo *II) const {
421    if (!II->hasMacroDefinition())
422      return 0;
423
424    return getInfoForMacro(II);
425  }
426
427  /// setMacroInfo - Specify a macro for this identifier.
428  ///
429  void setMacroInfo(IdentifierInfo *II, MacroInfo *MI);
430
431  /// macro_iterator/macro_begin/macro_end - This allows you to walk the current
432  /// state of the macro table.  This visits every currently-defined macro.
433  typedef llvm::DenseMap<IdentifierInfo*,
434                         MacroInfo*>::const_iterator macro_iterator;
435  macro_iterator macro_begin(bool IncludeExternalMacros = true) const;
436  macro_iterator macro_end(bool IncludeExternalMacros = true) const;
437
438  const std::string &getPredefines() const { return Predefines; }
439  /// setPredefines - Set the predefines for this Preprocessor.  These
440  /// predefines are automatically injected when parsing the main file.
441  void setPredefines(const char *P) { Predefines = P; }
442  void setPredefines(const std::string &P) { Predefines = P; }
443
444  /// getIdentifierInfo - Return information about the specified preprocessor
445  /// identifier token.  The version of this method that takes two character
446  /// pointers is preferred unless the identifier is already available as a
447  /// string (this avoids allocation and copying of memory to construct an
448  /// std::string).
449  IdentifierInfo *getIdentifierInfo(StringRef Name) const {
450    return &Identifiers.get(Name);
451  }
452
453  /// AddPragmaHandler - Add the specified pragma handler to the preprocessor.
454  /// If 'Namespace' is non-null, then it is a token required to exist on the
455  /// pragma line before the pragma string starts, e.g. "STDC" or "GCC".
456  void AddPragmaHandler(StringRef Namespace, PragmaHandler *Handler);
457  void AddPragmaHandler(PragmaHandler *Handler) {
458    AddPragmaHandler(StringRef(), Handler);
459  }
460
461  /// RemovePragmaHandler - Remove the specific pragma handler from
462  /// the preprocessor. If \arg Namespace is non-null, then it should
463  /// be the namespace that \arg Handler was added to. It is an error
464  /// to remove a handler that has not been registered.
465  void RemovePragmaHandler(StringRef Namespace, PragmaHandler *Handler);
466  void RemovePragmaHandler(PragmaHandler *Handler) {
467    RemovePragmaHandler(StringRef(), Handler);
468  }
469
470  /// \brief Add the specified comment handler to the preprocessor.
471  void AddCommentHandler(CommentHandler *Handler);
472
473  /// \brief Remove the specified comment handler.
474  ///
475  /// It is an error to remove a handler that has not been registered.
476  void RemoveCommentHandler(CommentHandler *Handler);
477
478  /// \brief Set the code completion handler to the given object.
479  void setCodeCompletionHandler(CodeCompletionHandler &Handler) {
480    CodeComplete = &Handler;
481  }
482
483  /// \brief Retrieve the current code-completion handler.
484  CodeCompletionHandler *getCodeCompletionHandler() const {
485    return CodeComplete;
486  }
487
488  /// \brief Clear out the code completion handler.
489  void clearCodeCompletionHandler() {
490    CodeComplete = 0;
491  }
492
493  /// \brief Hook used by the lexer to invoke the "natural language" code
494  /// completion point.
495  void CodeCompleteNaturalLanguage();
496
497  /// \brief Retrieve the preprocessing record, or NULL if there is no
498  /// preprocessing record.
499  PreprocessingRecord *getPreprocessingRecord() const { return Record; }
500
501  /// \brief Create a new preprocessing record, which will keep track of
502  /// all macro expansions, macro definitions, etc.
503  void createPreprocessingRecord(bool IncludeNestedMacroExpansions);
504
505  /// EnterMainSourceFile - Enter the specified FileID as the main source file,
506  /// which implicitly adds the builtin defines etc.
507  void EnterMainSourceFile();
508
509  /// EndSourceFile - Inform the preprocessor callbacks that processing is
510  /// complete.
511  void EndSourceFile();
512
513  /// EnterSourceFile - Add a source file to the top of the include stack and
514  /// start lexing tokens from it instead of the current buffer.  Emit an error
515  /// and don't enter the file on error.
516  void EnterSourceFile(FileID CurFileID, const DirectoryLookup *Dir,
517                       SourceLocation Loc);
518
519  /// EnterMacro - Add a Macro to the top of the include stack and start lexing
520  /// tokens from it instead of the current buffer.  Args specifies the
521  /// tokens input to a function-like macro.
522  ///
523  /// ILEnd specifies the location of the ')' for a function-like macro or the
524  /// identifier for an object-like macro.
525  void EnterMacro(Token &Identifier, SourceLocation ILEnd, MacroArgs *Args);
526
527  /// EnterTokenStream - Add a "macro" context to the top of the include stack,
528  /// which will cause the lexer to start returning the specified tokens.
529  ///
530  /// If DisableMacroExpansion is true, tokens lexed from the token stream will
531  /// not be subject to further macro expansion.  Otherwise, these tokens will
532  /// be re-macro-expanded when/if expansion is enabled.
533  ///
534  /// If OwnsTokens is false, this method assumes that the specified stream of
535  /// tokens has a permanent owner somewhere, so they do not need to be copied.
536  /// If it is true, it assumes the array of tokens is allocated with new[] and
537  /// must be freed.
538  ///
539  void EnterTokenStream(const Token *Toks, unsigned NumToks,
540                        bool DisableMacroExpansion, bool OwnsTokens);
541
542  /// RemoveTopOfLexerStack - Pop the current lexer/macro exp off the top of the
543  /// lexer stack.  This should only be used in situations where the current
544  /// state of the top-of-stack lexer is known.
545  void RemoveTopOfLexerStack();
546
547  /// EnableBacktrackAtThisPos - From the point that this method is called, and
548  /// until CommitBacktrackedTokens() or Backtrack() is called, the Preprocessor
549  /// keeps track of the lexed tokens so that a subsequent Backtrack() call will
550  /// make the Preprocessor re-lex the same tokens.
551  ///
552  /// Nested backtracks are allowed, meaning that EnableBacktrackAtThisPos can
553  /// be called multiple times and CommitBacktrackedTokens/Backtrack calls will
554  /// be combined with the EnableBacktrackAtThisPos calls in reverse order.
555  ///
556  /// NOTE: *DO NOT* forget to call either CommitBacktrackedTokens or Backtrack
557  /// at some point after EnableBacktrackAtThisPos. If you don't, caching of
558  /// tokens will continue indefinitely.
559  ///
560  void EnableBacktrackAtThisPos();
561
562  /// CommitBacktrackedTokens - Disable the last EnableBacktrackAtThisPos call.
563  void CommitBacktrackedTokens();
564
565  /// Backtrack - Make Preprocessor re-lex the tokens that were lexed since
566  /// EnableBacktrackAtThisPos() was previously called.
567  void Backtrack();
568
569  /// isBacktrackEnabled - True if EnableBacktrackAtThisPos() was called and
570  /// caching of tokens is on.
571  bool isBacktrackEnabled() const { return !BacktrackPositions.empty(); }
572
573  /// Lex - To lex a token from the preprocessor, just pull a token from the
574  /// current lexer or macro object.
575  void Lex(Token &Result) {
576    switch (CurLexerKind) {
577    case CLK_Lexer: CurLexer->Lex(Result); break;
578    case CLK_PTHLexer: CurPTHLexer->Lex(Result); break;
579    case CLK_TokenLexer: CurTokenLexer->Lex(Result); break;
580    case CLK_CachingLexer: CachingLex(Result); break;
581    case CLK_LexAfterModuleImport: LexAfterModuleImport(Result); break;
582    }
583  }
584
585  void LexAfterModuleImport(Token &Result);
586
587  /// LexNonComment - Lex a token.  If it's a comment, keep lexing until we get
588  /// something not a comment.  This is useful in -E -C mode where comments
589  /// would foul up preprocessor directive handling.
590  void LexNonComment(Token &Result) {
591    do
592      Lex(Result);
593    while (Result.getKind() == tok::comment);
594  }
595
596  /// LexUnexpandedToken - This is just like Lex, but this disables macro
597  /// expansion of identifier tokens.
598  void LexUnexpandedToken(Token &Result) {
599    // Disable macro expansion.
600    bool OldVal = DisableMacroExpansion;
601    DisableMacroExpansion = true;
602    // Lex the token.
603    Lex(Result);
604
605    // Reenable it.
606    DisableMacroExpansion = OldVal;
607  }
608
609  /// LexUnexpandedNonComment - Like LexNonComment, but this disables macro
610  /// expansion of identifier tokens.
611  void LexUnexpandedNonComment(Token &Result) {
612    do
613      LexUnexpandedToken(Result);
614    while (Result.getKind() == tok::comment);
615  }
616
617  /// LookAhead - This peeks ahead N tokens and returns that token without
618  /// consuming any tokens.  LookAhead(0) returns the next token that would be
619  /// returned by Lex(), LookAhead(1) returns the token after it, etc.  This
620  /// returns normal tokens after phase 5.  As such, it is equivalent to using
621  /// 'Lex', not 'LexUnexpandedToken'.
622  const Token &LookAhead(unsigned N) {
623    if (CachedLexPos + N < CachedTokens.size())
624      return CachedTokens[CachedLexPos+N];
625    else
626      return PeekAhead(N+1);
627  }
628
629  /// RevertCachedTokens - When backtracking is enabled and tokens are cached,
630  /// this allows to revert a specific number of tokens.
631  /// Note that the number of tokens being reverted should be up to the last
632  /// backtrack position, not more.
633  void RevertCachedTokens(unsigned N) {
634    assert(isBacktrackEnabled() &&
635           "Should only be called when tokens are cached for backtracking");
636    assert(signed(CachedLexPos) - signed(N) >= signed(BacktrackPositions.back())
637         && "Should revert tokens up to the last backtrack position, not more");
638    assert(signed(CachedLexPos) - signed(N) >= 0 &&
639           "Corrupted backtrack positions ?");
640    CachedLexPos -= N;
641  }
642
643  /// EnterToken - Enters a token in the token stream to be lexed next. If
644  /// BackTrack() is called afterwards, the token will remain at the insertion
645  /// point.
646  void EnterToken(const Token &Tok) {
647    EnterCachingLexMode();
648    CachedTokens.insert(CachedTokens.begin()+CachedLexPos, Tok);
649  }
650
651  /// AnnotateCachedTokens - We notify the Preprocessor that if it is caching
652  /// tokens (because backtrack is enabled) it should replace the most recent
653  /// cached tokens with the given annotation token. This function has no effect
654  /// if backtracking is not enabled.
655  ///
656  /// Note that the use of this function is just for optimization; so that the
657  /// cached tokens doesn't get re-parsed and re-resolved after a backtrack is
658  /// invoked.
659  void AnnotateCachedTokens(const Token &Tok) {
660    assert(Tok.isAnnotation() && "Expected annotation token");
661    if (CachedLexPos != 0 && isBacktrackEnabled())
662      AnnotatePreviousCachedTokens(Tok);
663  }
664
665  /// \brief Replace the last token with an annotation token.
666  ///
667  /// Like AnnotateCachedTokens(), this routine replaces an
668  /// already-parsed (and resolved) token with an annotation
669  /// token. However, this routine only replaces the last token with
670  /// the annotation token; it does not affect any other cached
671  /// tokens. This function has no effect if backtracking is not
672  /// enabled.
673  void ReplaceLastTokenWithAnnotation(const Token &Tok) {
674    assert(Tok.isAnnotation() && "Expected annotation token");
675    if (CachedLexPos != 0 && isBacktrackEnabled())
676      CachedTokens[CachedLexPos-1] = Tok;
677  }
678
679  /// \brief Specify the point at which code-completion will be performed.
680  ///
681  /// \param File the file in which code completion should occur. If
682  /// this file is included multiple times, code-completion will
683  /// perform completion the first time it is included. If NULL, this
684  /// function clears out the code-completion point.
685  ///
686  /// \param Line the line at which code completion should occur
687  /// (1-based).
688  ///
689  /// \param Column the column at which code completion should occur
690  /// (1-based).
691  ///
692  /// \returns true if an error occurred, false otherwise.
693  bool SetCodeCompletionPoint(const FileEntry *File,
694                              unsigned Line, unsigned Column);
695
696  /// \brief Determine if we are performing code completion.
697  bool isCodeCompletionEnabled() const { return CodeCompletionFile != 0; }
698
699  /// \brief Returns the location of the code-completion point.
700  /// Returns an invalid location if code-completion is not enabled or the file
701  /// containing the code-completion point has not been lexed yet.
702  SourceLocation getCodeCompletionLoc() const { return CodeCompletionLoc; }
703
704  /// \brief Returns the start location of the file of 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 getCodeCompletionFileLoc() const {
708    return CodeCompletionFileLoc;
709  }
710
711  /// \brief Returns true if code-completion is enabled and we have hit the
712  /// code-completion point.
713  bool isCodeCompletionReached() const { return CodeCompletionReached; }
714
715  /// \brief Note that we hit the code-completion point.
716  void setCodeCompletionReached() {
717    assert(isCodeCompletionEnabled() && "Code-completion not enabled!");
718    CodeCompletionReached = true;
719    // Silence any diagnostics that occur after we hit the code-completion.
720    getDiagnostics().setSuppressAllDiagnostics(true);
721  }
722
723  /// \brief Instruct the preprocessor to skip part of the main
724  /// the main source file.
725  ///
726  /// \brief Bytes The number of bytes in the preamble to skip.
727  ///
728  /// \brief StartOfLine Whether skipping these bytes puts the lexer at the
729  /// start of a line.
730  void setSkipMainFilePreamble(unsigned Bytes, bool StartOfLine) {
731    SkipMainFilePreamble.first = Bytes;
732    SkipMainFilePreamble.second = StartOfLine;
733  }
734
735  /// Diag - Forwarding function for diagnostics.  This emits a diagnostic at
736  /// the specified Token's location, translating the token's start
737  /// position in the current buffer into a SourcePosition object for rendering.
738  DiagnosticBuilder Diag(SourceLocation Loc, unsigned DiagID) {
739    return Diags->Report(Loc, DiagID);
740  }
741
742  DiagnosticBuilder Diag(const Token &Tok, unsigned DiagID) {
743    return Diags->Report(Tok.getLocation(), DiagID);
744  }
745
746  /// getSpelling() - Return the 'spelling' of the token at the given
747  /// location; does not go up to the spelling location or down to the
748  /// expansion location.
749  ///
750  /// \param buffer A buffer which will be used only if the token requires
751  ///   "cleaning", e.g. if it contains trigraphs or escaped newlines
752  /// \param invalid If non-null, will be set \c true if an error occurs.
753  StringRef getSpelling(SourceLocation loc,
754                              SmallVectorImpl<char> &buffer,
755                              bool *invalid = 0) const {
756    return Lexer::getSpelling(loc, buffer, SourceMgr, Features, invalid);
757  }
758
759  /// getSpelling() - Return the 'spelling' of the Tok token.  The spelling of a
760  /// token is the characters used to represent the token in the source file
761  /// after trigraph expansion and escaped-newline folding.  In particular, this
762  /// wants to get the true, uncanonicalized, spelling of things like digraphs
763  /// UCNs, etc.
764  ///
765  /// \param Invalid If non-null, will be set \c true if an error occurs.
766  std::string getSpelling(const Token &Tok, bool *Invalid = 0) const {
767    return Lexer::getSpelling(Tok, SourceMgr, Features, Invalid);
768  }
769
770  /// getSpelling - This method is used to get the spelling of a token into a
771  /// preallocated buffer, instead of as an std::string.  The caller is required
772  /// to allocate enough space for the token, which is guaranteed to be at least
773  /// Tok.getLength() bytes long.  The length of the actual result is returned.
774  ///
775  /// Note that this method may do two possible things: it may either fill in
776  /// the buffer specified with characters, or it may *change the input pointer*
777  /// to point to a constant buffer with the data already in it (avoiding a
778  /// copy).  The caller is not allowed to modify the returned buffer pointer
779  /// if an internal buffer is returned.
780  unsigned getSpelling(const Token &Tok, const char *&Buffer,
781                       bool *Invalid = 0) const {
782    return Lexer::getSpelling(Tok, Buffer, SourceMgr, Features, Invalid);
783  }
784
785  /// getSpelling - This method is used to get the spelling of a token into a
786  /// SmallVector. Note that the returned StringRef may not point to the
787  /// supplied buffer if a copy can be avoided.
788  StringRef getSpelling(const Token &Tok,
789                              SmallVectorImpl<char> &Buffer,
790                              bool *Invalid = 0) const;
791
792  /// getSpellingOfSingleCharacterNumericConstant - Tok is a numeric constant
793  /// with length 1, return the character.
794  char getSpellingOfSingleCharacterNumericConstant(const Token &Tok,
795                                                   bool *Invalid = 0) const {
796    assert(Tok.is(tok::numeric_constant) &&
797           Tok.getLength() == 1 && "Called on unsupported token");
798    assert(!Tok.needsCleaning() && "Token can't need cleaning with length 1");
799
800    // If the token is carrying a literal data pointer, just use it.
801    if (const char *D = Tok.getLiteralData())
802      return *D;
803
804    // Otherwise, fall back on getCharacterData, which is slower, but always
805    // works.
806    return *SourceMgr.getCharacterData(Tok.getLocation(), Invalid);
807  }
808
809  /// CreateString - Plop the specified string into a scratch buffer and set the
810  /// specified token's location and length to it.  If specified, the source
811  /// location provides a location of the expansion point of the token.
812  void CreateString(const char *Buf, unsigned Len,
813                    Token &Tok, SourceLocation SourceLoc = SourceLocation());
814
815  /// \brief Computes the source location just past the end of the
816  /// token at this source location.
817  ///
818  /// This routine can be used to produce a source location that
819  /// points just past the end of the token referenced by \p Loc, and
820  /// is generally used when a diagnostic needs to point just after a
821  /// token where it expected something different that it received. If
822  /// the returned source location would not be meaningful (e.g., if
823  /// it points into a macro), this routine returns an invalid
824  /// source location.
825  ///
826  /// \param Offset an offset from the end of the token, where the source
827  /// location should refer to. The default offset (0) produces a source
828  /// location pointing just past the end of the token; an offset of 1 produces
829  /// a source location pointing to the last character in the token, etc.
830  SourceLocation getLocForEndOfToken(SourceLocation Loc, unsigned Offset = 0) {
831    return Lexer::getLocForEndOfToken(Loc, Offset, SourceMgr, Features);
832  }
833
834  /// \brief Returns true if the given MacroID location points at the first
835  /// token of the macro expansion.
836  bool isAtStartOfMacroExpansion(SourceLocation loc) const {
837    return Lexer::isAtStartOfMacroExpansion(loc, SourceMgr, Features);
838  }
839
840  /// \brief Returns true if the given MacroID location points at the last
841  /// token of the macro expansion.
842  bool isAtEndOfMacroExpansion(SourceLocation loc) const {
843    return Lexer::isAtEndOfMacroExpansion(loc, SourceMgr, Features);
844  }
845
846  /// DumpToken - Print the token to stderr, used for debugging.
847  ///
848  void DumpToken(const Token &Tok, bool DumpFlags = false) const;
849  void DumpLocation(SourceLocation Loc) const;
850  void DumpMacro(const MacroInfo &MI) const;
851
852  /// AdvanceToTokenCharacter - Given a location that specifies the start of a
853  /// token, return a new location that specifies a character within the token.
854  SourceLocation AdvanceToTokenCharacter(SourceLocation TokStart,
855                                         unsigned Char) const {
856    return Lexer::AdvanceToTokenCharacter(TokStart, Char, SourceMgr, Features);
857  }
858
859  /// IncrementPasteCounter - Increment the counters for the number of token
860  /// paste operations performed.  If fast was specified, this is a 'fast paste'
861  /// case we handled.
862  ///
863  void IncrementPasteCounter(bool isFast) {
864    if (isFast)
865      ++NumFastTokenPaste;
866    else
867      ++NumTokenPaste;
868  }
869
870  void PrintStats();
871
872  size_t getTotalMemory() const;
873
874  /// HandleMicrosoftCommentPaste - When the macro expander pastes together a
875  /// comment (/##/) in microsoft mode, this method handles updating the current
876  /// state, returning the token on the next source line.
877  void HandleMicrosoftCommentPaste(Token &Tok);
878
879  //===--------------------------------------------------------------------===//
880  // Preprocessor callback methods.  These are invoked by a lexer as various
881  // directives and events are found.
882
883  /// LookUpIdentifierInfo - Given a tok::raw_identifier token, look up the
884  /// identifier information for the token and install it into the token,
885  /// updating the token kind accordingly.
886  IdentifierInfo *LookUpIdentifierInfo(Token &Identifier) const;
887
888private:
889  llvm::DenseMap<IdentifierInfo*,unsigned> PoisonReasons;
890
891public:
892
893  // SetPoisonReason - Call this function to indicate the reason for
894  // poisoning an identifier. If that identifier is accessed while
895  // poisoned, then this reason will be used instead of the default
896  // "poisoned" diagnostic.
897  void SetPoisonReason(IdentifierInfo *II, unsigned DiagID);
898
899  // HandlePoisonedIdentifier - Display reason for poisoned
900  // identifier.
901  void HandlePoisonedIdentifier(Token & Tok);
902
903  void MaybeHandlePoisonedIdentifier(Token & Identifier) {
904    if(IdentifierInfo * II = Identifier.getIdentifierInfo()) {
905      if(II->isPoisoned()) {
906        HandlePoisonedIdentifier(Identifier);
907      }
908    }
909  }
910
911private:
912  /// Identifiers used for SEH handling in Borland. These are only
913  /// allowed in particular circumstances
914  IdentifierInfo *Ident__exception_code, *Ident___exception_code, *Ident_GetExceptionCode; // __except block
915  IdentifierInfo *Ident__exception_info, *Ident___exception_info, *Ident_GetExceptionInfo; // __except filter expression
916  IdentifierInfo *Ident__abnormal_termination, *Ident___abnormal_termination, *Ident_AbnormalTermination; // __finally
917public:
918  void PoisonSEHIdentifiers(bool Poison = true); // Borland
919
920  /// HandleIdentifier - This callback is invoked when the lexer reads an
921  /// identifier and has filled in the tokens IdentifierInfo member.  This
922  /// callback potentially macro expands it or turns it into a named token (like
923  /// 'for').
924  void HandleIdentifier(Token &Identifier);
925
926
927  /// HandleEndOfFile - This callback is invoked when the lexer hits the end of
928  /// the current file.  This either returns the EOF token and returns true, or
929  /// pops a level off the include stack and returns false, at which point the
930  /// client should call lex again.
931  bool HandleEndOfFile(Token &Result, bool isEndOfMacro = false);
932
933  /// HandleEndOfTokenLexer - This callback is invoked when the current
934  /// TokenLexer hits the end of its token stream.
935  bool HandleEndOfTokenLexer(Token &Result);
936
937  /// HandleDirective - This callback is invoked when the lexer sees a # token
938  /// at the start of a line.  This consumes the directive, modifies the
939  /// lexer/preprocessor state, and advances the lexer(s) so that the next token
940  /// read is the correct one.
941  void HandleDirective(Token &Result);
942
943  /// CheckEndOfDirective - Ensure that the next token is a tok::eod token.  If
944  /// not, emit a diagnostic and consume up until the eod.  If EnableMacros is
945  /// true, then we consider macros that expand to zero tokens as being ok.
946  void CheckEndOfDirective(const char *Directive, bool EnableMacros = false);
947
948  /// DiscardUntilEndOfDirective - Read and discard all tokens remaining on the
949  /// current line until the tok::eod token is found.
950  void DiscardUntilEndOfDirective();
951
952  /// SawDateOrTime - This returns true if the preprocessor has seen a use of
953  /// __DATE__ or __TIME__ in the file so far.
954  bool SawDateOrTime() const {
955    return DATELoc != SourceLocation() || TIMELoc != SourceLocation();
956  }
957  unsigned getCounterValue() const { return CounterValue; }
958  void setCounterValue(unsigned V) { CounterValue = V; }
959
960  /// AllocateMacroInfo - Allocate a new MacroInfo object with the provide
961  ///  SourceLocation.
962  MacroInfo *AllocateMacroInfo(SourceLocation L);
963
964  /// CloneMacroInfo - Allocate a new MacroInfo object which is clone of MI.
965  MacroInfo *CloneMacroInfo(const MacroInfo &MI);
966
967  /// GetIncludeFilenameSpelling - Turn the specified lexer token into a fully
968  /// checked and spelled filename, e.g. as an operand of #include. This returns
969  /// true if the input filename was in <>'s or false if it were in ""'s.  The
970  /// caller is expected to provide a buffer that is large enough to hold the
971  /// spelling of the filename, but is also expected to handle the case when
972  /// this method decides to use a different buffer.
973  bool GetIncludeFilenameSpelling(SourceLocation Loc,StringRef &Filename);
974
975  /// LookupFile - Given a "foo" or <foo> reference, look up the indicated file,
976  /// return null on failure.  isAngled indicates whether the file reference is
977  /// for system #include's or not (i.e. using <> instead of "").
978  const FileEntry *LookupFile(StringRef Filename,
979                              bool isAngled, const DirectoryLookup *FromDir,
980                              const DirectoryLookup *&CurDir,
981                              SmallVectorImpl<char> *SearchPath,
982                              SmallVectorImpl<char> *RelativePath,
983                              StringRef *SuggestedModule);
984
985  /// GetCurLookup - The DirectoryLookup structure used to find the current
986  /// FileEntry, if CurLexer is non-null and if applicable.  This allows us to
987  /// implement #include_next and find directory-specific properties.
988  const DirectoryLookup *GetCurDirLookup() { return CurDirLookup; }
989
990  /// isInPrimaryFile - Return true if we're in the top-level file, not in a
991  /// #include.
992  bool isInPrimaryFile() const;
993
994  /// ConcatenateIncludeName - Handle cases where the #include name is expanded
995  /// from a macro as multiple tokens, which need to be glued together.  This
996  /// occurs for code like:
997  ///    #define FOO <a/b.h>
998  ///    #include FOO
999  /// because in this case, "<a/b.h>" is returned as 7 tokens, not one.
1000  ///
1001  /// This code concatenates and consumes tokens up to the '>' token.  It
1002  /// returns false if the > was found, otherwise it returns true if it finds
1003  /// and consumes the EOD marker.
1004  bool ConcatenateIncludeName(llvm::SmallString<128> &FilenameBuffer,
1005                              SourceLocation &End);
1006
1007  /// LexOnOffSwitch - Lex an on-off-switch (C99 6.10.6p2) and verify that it is
1008  /// followed by EOD.  Return true if the token is not a valid on-off-switch.
1009  bool LexOnOffSwitch(tok::OnOffSwitch &OOS);
1010
1011private:
1012
1013  void PushIncludeMacroStack() {
1014    IncludeMacroStack.push_back(IncludeStackInfo(CurLexerKind,
1015                                                 CurLexer.take(),
1016                                                 CurPTHLexer.take(),
1017                                                 CurPPLexer,
1018                                                 CurTokenLexer.take(),
1019                                                 CurDirLookup));
1020    CurPPLexer = 0;
1021  }
1022
1023  void PopIncludeMacroStack() {
1024    CurLexer.reset(IncludeMacroStack.back().TheLexer);
1025    CurPTHLexer.reset(IncludeMacroStack.back().ThePTHLexer);
1026    CurPPLexer = IncludeMacroStack.back().ThePPLexer;
1027    CurTokenLexer.reset(IncludeMacroStack.back().TheTokenLexer);
1028    CurDirLookup  = IncludeMacroStack.back().TheDirLookup;
1029    CurLexerKind = IncludeMacroStack.back().CurLexerKind;
1030    IncludeMacroStack.pop_back();
1031  }
1032
1033  /// AllocateMacroInfo - Allocate a new MacroInfo object.
1034  MacroInfo *AllocateMacroInfo();
1035
1036  /// ReleaseMacroInfo - Release the specified MacroInfo.  This memory will
1037  ///  be reused for allocating new MacroInfo objects.
1038  void ReleaseMacroInfo(MacroInfo* MI);
1039
1040  /// ReadMacroName - Lex and validate a macro name, which occurs after a
1041  /// #define or #undef.  This emits a diagnostic, sets the token kind to eod,
1042  /// and discards the rest of the macro line if the macro name is invalid.
1043  void ReadMacroName(Token &MacroNameTok, char isDefineUndef = 0);
1044
1045  /// ReadMacroDefinitionArgList - The ( starting an argument list of a macro
1046  /// definition has just been read.  Lex the rest of the arguments and the
1047  /// closing ), updating MI with what we learn.  Return true if an error occurs
1048  /// parsing the arg list.
1049  bool ReadMacroDefinitionArgList(MacroInfo *MI);
1050
1051  /// SkipExcludedConditionalBlock - We just read a #if or related directive and
1052  /// decided that the subsequent tokens are in the #if'd out portion of the
1053  /// file.  Lex the rest of the file, until we see an #endif.  If
1054  /// FoundNonSkipPortion is true, then we have already emitted code for part of
1055  /// this #if directive, so #else/#elif blocks should never be entered. If
1056  /// FoundElse is false, then #else directives are ok, if not, then we have
1057  /// already seen one so a #else directive is a duplicate.  When this returns,
1058  /// the caller can lex the first valid token.
1059  void SkipExcludedConditionalBlock(SourceLocation IfTokenLoc,
1060                                    bool FoundNonSkipPortion, bool FoundElse);
1061
1062  /// PTHSkipExcludedConditionalBlock - A fast PTH version of
1063  ///  SkipExcludedConditionalBlock.
1064  void PTHSkipExcludedConditionalBlock();
1065
1066  /// EvaluateDirectiveExpression - Evaluate an integer constant expression that
1067  /// may occur after a #if or #elif directive and return it as a bool.  If the
1068  /// expression is equivalent to "!defined(X)" return X in IfNDefMacro.
1069  bool EvaluateDirectiveExpression(IdentifierInfo *&IfNDefMacro);
1070
1071  /// RegisterBuiltinPragmas - Install the standard preprocessor pragmas:
1072  /// #pragma GCC poison/system_header/dependency and #pragma once.
1073  void RegisterBuiltinPragmas();
1074
1075  /// RegisterBuiltinMacros - Register builtin macros, such as __LINE__ with the
1076  /// identifier table.
1077  void RegisterBuiltinMacros();
1078
1079  /// HandleMacroExpandedIdentifier - If an identifier token is read that is to
1080  /// be expanded as a macro, handle it and return the next token as 'Tok'.  If
1081  /// the macro should not be expanded return true, otherwise return false.
1082  bool HandleMacroExpandedIdentifier(Token &Tok, MacroInfo *MI);
1083
1084  /// \brief Cache macro expanded tokens for TokenLexers.
1085  //
1086  /// Works like a stack; a TokenLexer adds the macro expanded tokens that is
1087  /// going to lex in the cache and when it finishes the tokens are removed
1088  /// from the end of the cache.
1089  Token *cacheMacroExpandedTokens(TokenLexer *tokLexer,
1090                                  ArrayRef<Token> tokens);
1091  void removeCachedMacroExpandedTokensOfLastLexer();
1092  friend void TokenLexer::ExpandFunctionArguments();
1093
1094  /// isNextPPTokenLParen - Determine whether the next preprocessor token to be
1095  /// lexed is a '('.  If so, consume the token and return true, if not, this
1096  /// method should have no observable side-effect on the lexed tokens.
1097  bool isNextPPTokenLParen();
1098
1099  /// ReadFunctionLikeMacroArgs - After reading "MACRO(", this method is
1100  /// invoked to read all of the formal arguments specified for the macro
1101  /// invocation.  This returns null on error.
1102  MacroArgs *ReadFunctionLikeMacroArgs(Token &MacroName, MacroInfo *MI,
1103                                       SourceLocation &ExpansionEnd);
1104
1105  /// ExpandBuiltinMacro - If an identifier token is read that is to be expanded
1106  /// as a builtin macro, handle it and return the next token as 'Tok'.
1107  void ExpandBuiltinMacro(Token &Tok);
1108
1109  /// Handle_Pragma - Read a _Pragma directive, slice it up, process it, then
1110  /// return the first token after the directive.  The _Pragma token has just
1111  /// been read into 'Tok'.
1112  void Handle_Pragma(Token &Tok);
1113
1114  /// HandleMicrosoft__pragma - Like Handle_Pragma except the pragma text
1115  /// is not enclosed within a string literal.
1116  void HandleMicrosoft__pragma(Token &Tok);
1117
1118  /// EnterSourceFileWithLexer - Add a lexer to the top of the include stack and
1119  /// start lexing tokens from it instead of the current buffer.
1120  void EnterSourceFileWithLexer(Lexer *TheLexer, const DirectoryLookup *Dir);
1121
1122  /// EnterSourceFileWithPTH - Add a lexer to the top of the include stack and
1123  /// start getting tokens from it using the PTH cache.
1124  void EnterSourceFileWithPTH(PTHLexer *PL, const DirectoryLookup *Dir);
1125
1126  /// IsFileLexer - Returns true if we are lexing from a file and not a
1127  ///  pragma or a macro.
1128  static bool IsFileLexer(const Lexer* L, const PreprocessorLexer* P) {
1129    return L ? !L->isPragmaLexer() : P != 0;
1130  }
1131
1132  static bool IsFileLexer(const IncludeStackInfo& I) {
1133    return IsFileLexer(I.TheLexer, I.ThePPLexer);
1134  }
1135
1136  bool IsFileLexer() const {
1137    return IsFileLexer(CurLexer.get(), CurPPLexer);
1138  }
1139
1140  //===--------------------------------------------------------------------===//
1141  // Caching stuff.
1142  void CachingLex(Token &Result);
1143  bool InCachingLexMode() const {
1144    // If the Lexer pointers are 0 and IncludeMacroStack is empty, it means
1145    // that we are past EOF, not that we are in CachingLex mode.
1146    return CurPPLexer == 0 && CurTokenLexer == 0 && CurPTHLexer == 0 &&
1147           !IncludeMacroStack.empty();
1148  }
1149  void EnterCachingLexMode();
1150  void ExitCachingLexMode() {
1151    if (InCachingLexMode())
1152      RemoveTopOfLexerStack();
1153  }
1154  const Token &PeekAhead(unsigned N);
1155  void AnnotatePreviousCachedTokens(const Token &Tok);
1156
1157  //===--------------------------------------------------------------------===//
1158  /// Handle*Directive - implement the various preprocessor directives.  These
1159  /// should side-effect the current preprocessor object so that the next call
1160  /// to Lex() will return the appropriate token next.
1161  void HandleLineDirective(Token &Tok);
1162  void HandleDigitDirective(Token &Tok);
1163  void HandleUserDiagnosticDirective(Token &Tok, bool isWarning);
1164  void HandleIdentSCCSDirective(Token &Tok);
1165  void HandleMacroExportDirective(Token &Tok);
1166
1167  // File inclusion.
1168  void HandleIncludeDirective(SourceLocation HashLoc,
1169                              Token &Tok,
1170                              const DirectoryLookup *LookupFrom = 0,
1171                              bool isImport = false);
1172  void HandleIncludeNextDirective(SourceLocation HashLoc, Token &Tok);
1173  void HandleIncludeMacrosDirective(SourceLocation HashLoc, Token &Tok);
1174  void HandleImportDirective(SourceLocation HashLoc, Token &Tok);
1175
1176  // Macro handling.
1177  void HandleDefineDirective(Token &Tok);
1178  void HandleUndefDirective(Token &Tok);
1179
1180  // Conditional Inclusion.
1181  void HandleIfdefDirective(Token &Tok, bool isIfndef,
1182                            bool ReadAnyTokensBeforeDirective);
1183  void HandleIfDirective(Token &Tok, bool ReadAnyTokensBeforeDirective);
1184  void HandleEndifDirective(Token &Tok);
1185  void HandleElseDirective(Token &Tok);
1186  void HandleElifDirective(Token &Tok);
1187
1188  // Pragmas.
1189  void HandlePragmaDirective(unsigned Introducer);
1190public:
1191  void HandlePragmaOnce(Token &OnceTok);
1192  void HandlePragmaMark();
1193  void HandlePragmaPoison(Token &PoisonTok);
1194  void HandlePragmaSystemHeader(Token &SysHeaderTok);
1195  void HandlePragmaDependency(Token &DependencyTok);
1196  void HandlePragmaComment(Token &CommentTok);
1197  void HandlePragmaMessage(Token &MessageTok);
1198  void HandlePragmaPushMacro(Token &Tok);
1199  void HandlePragmaPopMacro(Token &Tok);
1200  IdentifierInfo *ParsePragmaPushOrPopMacro(Token &Tok);
1201
1202  // Return true and store the first token only if any CommentHandler
1203  // has inserted some tokens and getCommentRetentionState() is false.
1204  bool HandleComment(Token &Token, SourceRange Comment);
1205
1206  /// \brief A macro is used, update information about macros that need unused
1207  /// warnings.
1208  void markMacroAsUsed(MacroInfo *MI);
1209};
1210
1211/// \brief Abstract base class that describes a handler that will receive
1212/// source ranges for each of the comments encountered in the source file.
1213class CommentHandler {
1214public:
1215  virtual ~CommentHandler();
1216
1217  // The handler shall return true if it has pushed any tokens
1218  // to be read using e.g. EnterToken or EnterTokenStream.
1219  virtual bool HandleComment(Preprocessor &PP, SourceRange Comment) = 0;
1220};
1221
1222}  // end namespace clang
1223
1224#endif
1225