xref: /NextBSD/contrib/llvm/tools/clang/include/clang/Lex/Preprocessor.h (revision 84d351007654069f9643c8e4b4802a7f5f08ee42)
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 /// \file
11 /// \brief Defines the clang::Preprocessor interface.
12 ///
13 //===----------------------------------------------------------------------===//
14 
15 #ifndef LLVM_CLANG_LEX_PREPROCESSOR_H
16 #define LLVM_CLANG_LEX_PREPROCESSOR_H
17 
18 #include "clang/Basic/Builtins.h"
19 #include "clang/Basic/Diagnostic.h"
20 #include "clang/Basic/IdentifierTable.h"
21 #include "clang/Basic/SourceLocation.h"
22 #include "clang/Lex/Lexer.h"
23 #include "clang/Lex/MacroInfo.h"
24 #include "clang/Lex/ModuleMap.h"
25 #include "clang/Lex/PPCallbacks.h"
26 #include "clang/Lex/PTHLexer.h"
27 #include "clang/Lex/PTHManager.h"
28 #include "clang/Lex/TokenLexer.h"
29 #include "llvm/ADT/ArrayRef.h"
30 #include "llvm/ADT/DenseMap.h"
31 #include "llvm/ADT/IntrusiveRefCntPtr.h"
32 #include "llvm/ADT/SmallPtrSet.h"
33 #include "llvm/ADT/SmallVector.h"
34 #include "llvm/ADT/TinyPtrVector.h"
35 #include "llvm/Support/Allocator.h"
36 #include <memory>
37 #include <vector>
38 
39 namespace llvm {
40   template<unsigned InternalLen> class SmallString;
41 }
42 
43 namespace clang {
44 
45 class SourceManager;
46 class ExternalPreprocessorSource;
47 class FileManager;
48 class FileEntry;
49 class HeaderSearch;
50 class PragmaNamespace;
51 class PragmaHandler;
52 class CommentHandler;
53 class ScratchBuffer;
54 class TargetInfo;
55 class PPCallbacks;
56 class CodeCompletionHandler;
57 class DirectoryLookup;
58 class PreprocessingRecord;
59 class ModuleLoader;
60 class PreprocessorOptions;
61 
62 /// \brief Stores token information for comparing actual tokens with
63 /// predefined values.  Only handles simple tokens and identifiers.
64 class TokenValue {
65   tok::TokenKind Kind;
66   IdentifierInfo *II;
67 
68 public:
TokenValue(tok::TokenKind Kind)69   TokenValue(tok::TokenKind Kind) : Kind(Kind), II(nullptr) {
70     assert(Kind != tok::raw_identifier && "Raw identifiers are not supported.");
71     assert(Kind != tok::identifier &&
72            "Identifiers should be created by TokenValue(IdentifierInfo *)");
73     assert(!tok::isLiteral(Kind) && "Literals are not supported.");
74     assert(!tok::isAnnotation(Kind) && "Annotations are not supported.");
75   }
TokenValue(IdentifierInfo * II)76   TokenValue(IdentifierInfo *II) : Kind(tok::identifier), II(II) {}
77   bool operator==(const Token &Tok) const {
78     return Tok.getKind() == Kind &&
79         (!II || II == Tok.getIdentifierInfo());
80   }
81 };
82 
83 /// \brief Context in which macro name is used.
84 enum MacroUse {
85   MU_Other  = 0,  // other than #define or #undef
86   MU_Define = 1,  // macro name specified in #define
87   MU_Undef  = 2   // macro name specified in #undef
88 };
89 
90 /// \brief Engages in a tight little dance with the lexer to efficiently
91 /// preprocess tokens.
92 ///
93 /// Lexers know only about tokens within a single source file, and don't
94 /// know anything about preprocessor-level issues like the \#include stack,
95 /// token expansion, etc.
96 class Preprocessor : public RefCountedBase<Preprocessor> {
97   IntrusiveRefCntPtr<PreprocessorOptions> PPOpts;
98   DiagnosticsEngine        *Diags;
99   LangOptions       &LangOpts;
100   const TargetInfo  *Target;
101   FileManager       &FileMgr;
102   SourceManager     &SourceMgr;
103   std::unique_ptr<ScratchBuffer> ScratchBuf;
104   HeaderSearch      &HeaderInfo;
105   ModuleLoader      &TheModuleLoader;
106 
107   /// \brief External source of macros.
108   ExternalPreprocessorSource *ExternalSource;
109 
110 
111   /// An optional PTHManager object used for getting tokens from
112   /// a token cache rather than lexing the original source file.
113   std::unique_ptr<PTHManager> PTH;
114 
115   /// A BumpPtrAllocator object used to quickly allocate and release
116   /// objects internal to the Preprocessor.
117   llvm::BumpPtrAllocator BP;
118 
119   /// Identifiers for builtin macros and other builtins.
120   IdentifierInfo *Ident__LINE__, *Ident__FILE__;   // __LINE__, __FILE__
121   IdentifierInfo *Ident__DATE__, *Ident__TIME__;   // __DATE__, __TIME__
122   IdentifierInfo *Ident__INCLUDE_LEVEL__;          // __INCLUDE_LEVEL__
123   IdentifierInfo *Ident__BASE_FILE__;              // __BASE_FILE__
124   IdentifierInfo *Ident__TIMESTAMP__;              // __TIMESTAMP__
125   IdentifierInfo *Ident__COUNTER__;                // __COUNTER__
126   IdentifierInfo *Ident_Pragma, *Ident__pragma;    // _Pragma, __pragma
127   IdentifierInfo *Ident__identifier;               // __identifier
128   IdentifierInfo *Ident__VA_ARGS__;                // __VA_ARGS__
129   IdentifierInfo *Ident__has_feature;              // __has_feature
130   IdentifierInfo *Ident__has_extension;            // __has_extension
131   IdentifierInfo *Ident__has_builtin;              // __has_builtin
132   IdentifierInfo *Ident__has_attribute;            // __has_attribute
133   IdentifierInfo *Ident__has_include;              // __has_include
134   IdentifierInfo *Ident__has_include_next;         // __has_include_next
135   IdentifierInfo *Ident__has_warning;              // __has_warning
136   IdentifierInfo *Ident__is_identifier;            // __is_identifier
137   IdentifierInfo *Ident__building_module;          // __building_module
138   IdentifierInfo *Ident__MODULE__;                 // __MODULE__
139   IdentifierInfo *Ident__has_cpp_attribute;        // __has_cpp_attribute
140   IdentifierInfo *Ident__has_declspec;             // __has_declspec_attribute
141 
142   SourceLocation DATELoc, TIMELoc;
143   unsigned CounterValue;  // Next __COUNTER__ value.
144 
145   enum {
146     /// \brief Maximum depth of \#includes.
147     MaxAllowedIncludeStackDepth = 200
148   };
149 
150   // State that is set before the preprocessor begins.
151   bool KeepComments : 1;
152   bool KeepMacroComments : 1;
153   bool SuppressIncludeNotFoundError : 1;
154 
155   // State that changes while the preprocessor runs:
156   bool InMacroArgs : 1;            // True if parsing fn macro invocation args.
157 
158   /// Whether the preprocessor owns the header search object.
159   bool OwnsHeaderSearch : 1;
160 
161   /// True if macro expansion is disabled.
162   bool DisableMacroExpansion : 1;
163 
164   /// Temporarily disables DisableMacroExpansion (i.e. enables expansion)
165   /// when parsing preprocessor directives.
166   bool MacroExpansionInDirectivesOverride : 1;
167 
168   class ResetMacroExpansionHelper;
169 
170   /// \brief Whether we have already loaded macros from the external source.
171   mutable bool ReadMacrosFromExternalSource : 1;
172 
173   /// \brief True if pragmas are enabled.
174   bool PragmasEnabled : 1;
175 
176   /// \brief True if the current build action is a preprocessing action.
177   bool PreprocessedOutput : 1;
178 
179   /// \brief True if we are currently preprocessing a #if or #elif directive
180   bool ParsingIfOrElifDirective;
181 
182   /// \brief True if we are pre-expanding macro arguments.
183   bool InMacroArgPreExpansion;
184 
185   /// \brief Mapping/lookup information for all identifiers in
186   /// the program, including program keywords.
187   mutable IdentifierTable Identifiers;
188 
189   /// \brief This table contains all the selectors in the program.
190   ///
191   /// Unlike IdentifierTable above, this table *isn't* populated by the
192   /// preprocessor. It is declared/expanded here because its role/lifetime is
193   /// conceptually similar to the IdentifierTable. In addition, the current
194   /// control flow (in clang::ParseAST()), make it convenient to put here.
195   ///
196   /// FIXME: Make sure the lifetime of Identifiers/Selectors *isn't* tied to
197   /// the lifetime of the preprocessor.
198   SelectorTable Selectors;
199 
200   /// \brief Information about builtins.
201   Builtin::Context BuiltinInfo;
202 
203   /// \brief Tracks all of the pragmas that the client registered
204   /// with this preprocessor.
205   std::unique_ptr<PragmaNamespace> PragmaHandlers;
206 
207   /// \brief Pragma handlers of the original source is stored here during the
208   /// parsing of a model file.
209   std::unique_ptr<PragmaNamespace> PragmaHandlersBackup;
210 
211   /// \brief Tracks all of the comment handlers that the client registered
212   /// with this preprocessor.
213   std::vector<CommentHandler *> CommentHandlers;
214 
215   /// \brief True if we want to ignore EOF token and continue later on (thus
216   /// avoid tearing the Lexer and etc. down).
217   bool IncrementalProcessing;
218 
219   /// The kind of translation unit we are processing.
220   TranslationUnitKind TUKind;
221 
222   /// \brief The code-completion handler.
223   CodeCompletionHandler *CodeComplete;
224 
225   /// \brief The file that we're performing code-completion for, if any.
226   const FileEntry *CodeCompletionFile;
227 
228   /// \brief The offset in file for the code-completion point.
229   unsigned CodeCompletionOffset;
230 
231   /// \brief The location for the code-completion point. This gets instantiated
232   /// when the CodeCompletionFile gets \#include'ed for preprocessing.
233   SourceLocation CodeCompletionLoc;
234 
235   /// \brief The start location for the file of the code-completion point.
236   ///
237   /// This gets instantiated when the CodeCompletionFile gets \#include'ed
238   /// for preprocessing.
239   SourceLocation CodeCompletionFileLoc;
240 
241   /// \brief The source location of the \c import contextual keyword we just
242   /// lexed, if any.
243   SourceLocation ModuleImportLoc;
244 
245   /// \brief The module import path that we're currently processing.
246   SmallVector<std::pair<IdentifierInfo *, SourceLocation>, 2> ModuleImportPath;
247 
248   /// \brief Whether the last token we lexed was an '@'.
249   bool LastTokenWasAt;
250 
251   /// \brief Whether the module import expects an identifier next. Otherwise,
252   /// it expects a '.' or ';'.
253   bool ModuleImportExpectsIdentifier;
254 
255   /// \brief The source location of the currently-active
256   /// \#pragma clang arc_cf_code_audited begin.
257   SourceLocation PragmaARCCFCodeAuditedLoc;
258 
259   /// \brief The source location of the currently-active
260   /// \#pragma clang assume_nonnull begin.
261   SourceLocation PragmaAssumeNonNullLoc;
262 
263   /// \brief True if we hit the code-completion point.
264   bool CodeCompletionReached;
265 
266   /// \brief The directory that the main file should be considered to occupy,
267   /// if it does not correspond to a real file (as happens when building a
268   /// module).
269   const DirectoryEntry *MainFileDir;
270 
271   /// \brief The number of bytes that we will initially skip when entering the
272   /// main file, along with a flag that indicates whether skipping this number
273   /// of bytes will place the lexer at the start of a line.
274   ///
275   /// This is used when loading a precompiled preamble.
276   std::pair<int, bool> SkipMainFilePreamble;
277 
278   /// \brief The current top of the stack that we're lexing from if
279   /// not expanding a macro and we are lexing directly from source code.
280   ///
281   /// Only one of CurLexer, CurPTHLexer, or CurTokenLexer will be non-null.
282   std::unique_ptr<Lexer> CurLexer;
283 
284   /// \brief The current top of stack that we're lexing from if
285   /// not expanding from a macro and we are lexing from a PTH cache.
286   ///
287   /// Only one of CurLexer, CurPTHLexer, or CurTokenLexer will be non-null.
288   std::unique_ptr<PTHLexer> CurPTHLexer;
289 
290   /// \brief The current top of the stack what we're lexing from
291   /// if not expanding a macro.
292   ///
293   /// This is an alias for either CurLexer or  CurPTHLexer.
294   PreprocessorLexer *CurPPLexer;
295 
296   /// \brief Used to find the current FileEntry, if CurLexer is non-null
297   /// and if applicable.
298   ///
299   /// This allows us to implement \#include_next and find directory-specific
300   /// properties.
301   const DirectoryLookup *CurDirLookup;
302 
303   /// \brief The current macro we are expanding, if we are expanding a macro.
304   ///
305   /// One of CurLexer and CurTokenLexer must be null.
306   std::unique_ptr<TokenLexer> CurTokenLexer;
307 
308   /// \brief The kind of lexer we're currently working with.
309   enum CurLexerKind {
310     CLK_Lexer,
311     CLK_PTHLexer,
312     CLK_TokenLexer,
313     CLK_CachingLexer,
314     CLK_LexAfterModuleImport
315   } CurLexerKind;
316 
317   /// \brief If the current lexer is for a submodule that is being built, this
318   /// is that submodule.
319   Module *CurSubmodule;
320 
321   /// \brief Keeps track of the stack of files currently
322   /// \#included, and macros currently being expanded from, not counting
323   /// CurLexer/CurTokenLexer.
324   struct IncludeStackInfo {
325     enum CurLexerKind           CurLexerKind;
326     Module                     *TheSubmodule;
327     std::unique_ptr<Lexer>      TheLexer;
328     std::unique_ptr<PTHLexer>   ThePTHLexer;
329     PreprocessorLexer          *ThePPLexer;
330     std::unique_ptr<TokenLexer> TheTokenLexer;
331     const DirectoryLookup      *TheDirLookup;
332 
333     // The following constructors are completely useless copies of the default
334     // versions, only needed to pacify MSVC.
IncludeStackInfoIncludeStackInfo335     IncludeStackInfo(enum CurLexerKind CurLexerKind, Module *TheSubmodule,
336                      std::unique_ptr<Lexer> &&TheLexer,
337                      std::unique_ptr<PTHLexer> &&ThePTHLexer,
338                      PreprocessorLexer *ThePPLexer,
339                      std::unique_ptr<TokenLexer> &&TheTokenLexer,
340                      const DirectoryLookup *TheDirLookup)
341         : CurLexerKind(std::move(CurLexerKind)),
342           TheSubmodule(std::move(TheSubmodule)), TheLexer(std::move(TheLexer)),
343           ThePTHLexer(std::move(ThePTHLexer)),
344           ThePPLexer(std::move(ThePPLexer)),
345           TheTokenLexer(std::move(TheTokenLexer)),
346           TheDirLookup(std::move(TheDirLookup)) {}
IncludeStackInfoIncludeStackInfo347     IncludeStackInfo(IncludeStackInfo &&RHS)
348         : CurLexerKind(std::move(RHS.CurLexerKind)),
349           TheSubmodule(std::move(RHS.TheSubmodule)),
350           TheLexer(std::move(RHS.TheLexer)),
351           ThePTHLexer(std::move(RHS.ThePTHLexer)),
352           ThePPLexer(std::move(RHS.ThePPLexer)),
353           TheTokenLexer(std::move(RHS.TheTokenLexer)),
354           TheDirLookup(std::move(RHS.TheDirLookup)) {}
355   };
356   std::vector<IncludeStackInfo> IncludeMacroStack;
357 
358   /// \brief Actions invoked when some preprocessor activity is
359   /// encountered (e.g. a file is \#included, etc).
360   std::unique_ptr<PPCallbacks> Callbacks;
361 
362   struct MacroExpandsInfo {
363     Token Tok;
364     MacroDefinition MD;
365     SourceRange Range;
MacroExpandsInfoMacroExpandsInfo366     MacroExpandsInfo(Token Tok, MacroDefinition MD, SourceRange Range)
367       : Tok(Tok), MD(MD), Range(Range) { }
368   };
369   SmallVector<MacroExpandsInfo, 2> DelayedMacroExpandsCallbacks;
370 
371   /// Information about a name that has been used to define a module macro.
372   struct ModuleMacroInfo {
ModuleMacroInfoModuleMacroInfo373     ModuleMacroInfo(MacroDirective *MD)
374         : MD(MD), ActiveModuleMacrosGeneration(0), IsAmbiguous(false) {}
375 
376     /// The most recent macro directive for this identifier.
377     MacroDirective *MD;
378     /// The active module macros for this identifier.
379     llvm::TinyPtrVector<ModuleMacro*> ActiveModuleMacros;
380     /// The generation number at which we last updated ActiveModuleMacros.
381     /// \see Preprocessor::VisibleModules.
382     unsigned ActiveModuleMacrosGeneration;
383     /// Whether this macro name is ambiguous.
384     bool IsAmbiguous;
385     /// The module macros that are overridden by this macro.
386     llvm::TinyPtrVector<ModuleMacro*> OverriddenMacros;
387   };
388 
389   /// The state of a macro for an identifier.
390   class MacroState {
391     mutable llvm::PointerUnion<MacroDirective *, ModuleMacroInfo *> State;
392 
getModuleInfo(Preprocessor & PP,const IdentifierInfo * II)393     ModuleMacroInfo *getModuleInfo(Preprocessor &PP,
394                                    const IdentifierInfo *II) const {
395       // FIXME: Find a spare bit on IdentifierInfo and store a
396       //        HasModuleMacros flag.
397       if (!II->hasMacroDefinition() ||
398           (!PP.getLangOpts().Modules &&
399            !PP.getLangOpts().ModulesLocalVisibility) ||
400           !PP.CurSubmoduleState->VisibleModules.getGeneration())
401         return nullptr;
402 
403       auto *Info = State.dyn_cast<ModuleMacroInfo*>();
404       if (!Info) {
405         Info = new (PP.getPreprocessorAllocator())
406             ModuleMacroInfo(State.get<MacroDirective *>());
407         State = Info;
408       }
409 
410       if (PP.CurSubmoduleState->VisibleModules.getGeneration() !=
411           Info->ActiveModuleMacrosGeneration)
412         PP.updateModuleMacroInfo(II, *Info);
413       return Info;
414     }
415 
416   public:
MacroState()417     MacroState() : MacroState(nullptr) {}
MacroState(MacroDirective * MD)418     MacroState(MacroDirective *MD) : State(MD) {}
MacroState(MacroState && O)419     MacroState(MacroState &&O) LLVM_NOEXCEPT : State(O.State) {
420       O.State = (MacroDirective *)nullptr;
421     }
422     MacroState &operator=(MacroState &&O) LLVM_NOEXCEPT {
423       auto S = O.State;
424       O.State = (MacroDirective *)nullptr;
425       State = S;
426       return *this;
427     }
~MacroState()428     ~MacroState() {
429       if (auto *Info = State.dyn_cast<ModuleMacroInfo*>())
430         Info->~ModuleMacroInfo();
431     }
432 
getLatest()433     MacroDirective *getLatest() const {
434       if (auto *Info = State.dyn_cast<ModuleMacroInfo*>())
435         return Info->MD;
436       return State.get<MacroDirective*>();
437     }
setLatest(MacroDirective * MD)438     void setLatest(MacroDirective *MD) {
439       if (auto *Info = State.dyn_cast<ModuleMacroInfo*>())
440         Info->MD = MD;
441       else
442         State = MD;
443     }
444 
isAmbiguous(Preprocessor & PP,const IdentifierInfo * II)445     bool isAmbiguous(Preprocessor &PP, const IdentifierInfo *II) const {
446       auto *Info = getModuleInfo(PP, II);
447       return Info ? Info->IsAmbiguous : false;
448     }
449     ArrayRef<ModuleMacro *>
getActiveModuleMacros(Preprocessor & PP,const IdentifierInfo * II)450     getActiveModuleMacros(Preprocessor &PP, const IdentifierInfo *II) const {
451       if (auto *Info = getModuleInfo(PP, II))
452         return Info->ActiveModuleMacros;
453       return None;
454     }
455 
findDirectiveAtLoc(SourceLocation Loc,SourceManager & SourceMgr)456     MacroDirective::DefInfo findDirectiveAtLoc(SourceLocation Loc,
457                                                SourceManager &SourceMgr) const {
458       // FIXME: Incorporate module macros into the result of this.
459       if (auto *Latest = getLatest())
460         return Latest->findDirectiveAtLoc(Loc, SourceMgr);
461       return MacroDirective::DefInfo();
462     }
463 
overrideActiveModuleMacros(Preprocessor & PP,IdentifierInfo * II)464     void overrideActiveModuleMacros(Preprocessor &PP, IdentifierInfo *II) {
465       if (auto *Info = getModuleInfo(PP, II)) {
466         Info->OverriddenMacros.insert(Info->OverriddenMacros.end(),
467                                       Info->ActiveModuleMacros.begin(),
468                                       Info->ActiveModuleMacros.end());
469         Info->ActiveModuleMacros.clear();
470         Info->IsAmbiguous = false;
471       }
472     }
getOverriddenMacros()473     ArrayRef<ModuleMacro*> getOverriddenMacros() const {
474       if (auto *Info = State.dyn_cast<ModuleMacroInfo*>())
475         return Info->OverriddenMacros;
476       return None;
477     }
setOverriddenMacros(Preprocessor & PP,ArrayRef<ModuleMacro * > Overrides)478     void setOverriddenMacros(Preprocessor &PP,
479                              ArrayRef<ModuleMacro *> Overrides) {
480       auto *Info = State.dyn_cast<ModuleMacroInfo*>();
481       if (!Info) {
482         if (Overrides.empty())
483           return;
484         Info = new (PP.getPreprocessorAllocator())
485             ModuleMacroInfo(State.get<MacroDirective *>());
486         State = Info;
487       }
488       Info->OverriddenMacros.clear();
489       Info->OverriddenMacros.insert(Info->OverriddenMacros.end(),
490                                     Overrides.begin(), Overrides.end());
491       Info->ActiveModuleMacrosGeneration = 0;
492     }
493   };
494 
495   /// For each IdentifierInfo that was associated with a macro, we
496   /// keep a mapping to the history of all macro definitions and #undefs in
497   /// the reverse order (the latest one is in the head of the list).
498   ///
499   /// This mapping lives within the \p CurSubmoduleState.
500   typedef llvm::DenseMap<const IdentifierInfo *, MacroState> MacroMap;
501 
502   friend class ASTReader;
503 
504   struct SubmoduleState;
505 
506   /// \brief Information about a submodule that we're currently building.
507   struct BuildingSubmoduleInfo {
BuildingSubmoduleInfoBuildingSubmoduleInfo508     BuildingSubmoduleInfo(Module *M, SourceLocation ImportLoc,
509                           SubmoduleState *OuterSubmoduleState)
510         : M(M), ImportLoc(ImportLoc), OuterSubmoduleState(OuterSubmoduleState) {
511     }
512 
513     /// The module that we are building.
514     Module *M;
515     /// The location at which the module was included.
516     SourceLocation ImportLoc;
517     /// The previous SubmoduleState.
518     SubmoduleState *OuterSubmoduleState;
519   };
520   SmallVector<BuildingSubmoduleInfo, 8> BuildingSubmoduleStack;
521 
522   /// \brief Information about a submodule's preprocessor state.
523   struct SubmoduleState {
524     /// The macros for the submodule.
525     MacroMap Macros;
526     /// The set of modules that are visible within the submodule.
527     VisibleModuleSet VisibleModules;
528     // FIXME: CounterValue?
529     // FIXME: PragmaPushMacroInfo?
530   };
531   std::map<Module*, SubmoduleState> Submodules;
532 
533   /// The preprocessor state for preprocessing outside of any submodule.
534   SubmoduleState NullSubmoduleState;
535 
536   /// The current submodule state. Will be \p NullSubmoduleState if we're not
537   /// in a submodule.
538   SubmoduleState *CurSubmoduleState;
539 
540   /// The set of known macros exported from modules.
541   llvm::FoldingSet<ModuleMacro> ModuleMacros;
542 
543   /// The list of module macros, for each identifier, that are not overridden by
544   /// any other module macro.
545   llvm::DenseMap<const IdentifierInfo *, llvm::TinyPtrVector<ModuleMacro*>>
546       LeafModuleMacros;
547 
548   /// \brief Macros that we want to warn because they are not used at the end
549   /// of the translation unit.
550   ///
551   /// We store just their SourceLocations instead of
552   /// something like MacroInfo*. The benefit of this is that when we are
553   /// deserializing from PCH, we don't need to deserialize identifier & macros
554   /// just so that we can report that they are unused, we just warn using
555   /// the SourceLocations of this set (that will be filled by the ASTReader).
556   /// We are using SmallPtrSet instead of a vector for faster removal.
557   typedef llvm::SmallPtrSet<SourceLocation, 32> WarnUnusedMacroLocsTy;
558   WarnUnusedMacroLocsTy WarnUnusedMacroLocs;
559 
560   /// \brief A "freelist" of MacroArg objects that can be
561   /// reused for quick allocation.
562   MacroArgs *MacroArgCache;
563   friend class MacroArgs;
564 
565   /// For each IdentifierInfo used in a \#pragma push_macro directive,
566   /// we keep a MacroInfo stack used to restore the previous macro value.
567   llvm::DenseMap<IdentifierInfo*, std::vector<MacroInfo*> > PragmaPushMacroInfo;
568 
569   // Various statistics we track for performance analysis.
570   unsigned NumDirectives, NumDefined, NumUndefined, NumPragma;
571   unsigned NumIf, NumElse, NumEndif;
572   unsigned NumEnteredSourceFiles, MaxIncludeStackDepth;
573   unsigned NumMacroExpanded, NumFnMacroExpanded, NumBuiltinMacroExpanded;
574   unsigned NumFastMacroExpanded, NumTokenPaste, NumFastTokenPaste;
575   unsigned NumSkipped;
576 
577   /// \brief The predefined macros that preprocessor should use from the
578   /// command line etc.
579   std::string Predefines;
580 
581   /// \brief The file ID for the preprocessor predefines.
582   FileID PredefinesFileID;
583 
584   /// \{
585   /// \brief Cache of macro expanders to reduce malloc traffic.
586   enum { TokenLexerCacheSize = 8 };
587   unsigned NumCachedTokenLexers;
588   std::unique_ptr<TokenLexer> TokenLexerCache[TokenLexerCacheSize];
589   /// \}
590 
591   /// \brief Keeps macro expanded tokens for TokenLexers.
592   //
593   /// Works like a stack; a TokenLexer adds the macro expanded tokens that is
594   /// going to lex in the cache and when it finishes the tokens are removed
595   /// from the end of the cache.
596   SmallVector<Token, 16> MacroExpandedTokens;
597   std::vector<std::pair<TokenLexer *, size_t> > MacroExpandingLexersStack;
598 
599   /// \brief A record of the macro definitions and expansions that
600   /// occurred during preprocessing.
601   ///
602   /// This is an optional side structure that can be enabled with
603   /// \c createPreprocessingRecord() prior to preprocessing.
604   PreprocessingRecord *Record;
605 
606   /// Cached tokens state.
607   typedef SmallVector<Token, 1> CachedTokensTy;
608 
609   /// \brief Cached tokens are stored here when we do backtracking or
610   /// lookahead. They are "lexed" by the CachingLex() method.
611   CachedTokensTy CachedTokens;
612 
613   /// \brief The position of the cached token that CachingLex() should
614   /// "lex" next.
615   ///
616   /// If it points beyond the CachedTokens vector, it means that a normal
617   /// Lex() should be invoked.
618   CachedTokensTy::size_type CachedLexPos;
619 
620   /// \brief Stack of backtrack positions, allowing nested backtracks.
621   ///
622   /// The EnableBacktrackAtThisPos() method pushes a position to
623   /// indicate where CachedLexPos should be set when the BackTrack() method is
624   /// invoked (at which point the last position is popped).
625   std::vector<CachedTokensTy::size_type> BacktrackPositions;
626 
627   struct MacroInfoChain {
628     MacroInfo MI;
629     MacroInfoChain *Next;
630   };
631 
632   /// MacroInfos are managed as a chain for easy disposal.  This is the head
633   /// of that list.
634   MacroInfoChain *MIChainHead;
635 
636   struct DeserializedMacroInfoChain {
637     MacroInfo MI;
638     unsigned OwningModuleID; // MUST be immediately after the MacroInfo object
639                      // so it can be accessed by MacroInfo::getOwningModuleID().
640     DeserializedMacroInfoChain *Next;
641   };
642   DeserializedMacroInfoChain *DeserialMIChainHead;
643 
644 public:
645   Preprocessor(IntrusiveRefCntPtr<PreprocessorOptions> PPOpts,
646                DiagnosticsEngine &diags, LangOptions &opts,
647                SourceManager &SM, HeaderSearch &Headers,
648                ModuleLoader &TheModuleLoader,
649                IdentifierInfoLookup *IILookup = nullptr,
650                bool OwnsHeaderSearch = false,
651                TranslationUnitKind TUKind = TU_Complete);
652 
653   ~Preprocessor();
654 
655   /// \brief Initialize the preprocessor using information about the target.
656   ///
657   /// \param Target is owned by the caller and must remain valid for the
658   /// lifetime of the preprocessor.
659   void Initialize(const TargetInfo &Target);
660 
661   /// \brief Initialize the preprocessor to parse a model file
662   ///
663   /// To parse model files the preprocessor of the original source is reused to
664   /// preserver the identifier table. However to avoid some duplicate
665   /// information in the preprocessor some cleanup is needed before it is used
666   /// to parse model files. This method does that cleanup.
667   void InitializeForModelFile();
668 
669   /// \brief Cleanup after model file parsing
670   void FinalizeForModelFile();
671 
672   /// \brief Retrieve the preprocessor options used to initialize this
673   /// preprocessor.
getPreprocessorOpts()674   PreprocessorOptions &getPreprocessorOpts() const { return *PPOpts; }
675 
getDiagnostics()676   DiagnosticsEngine &getDiagnostics() const { return *Diags; }
setDiagnostics(DiagnosticsEngine & D)677   void setDiagnostics(DiagnosticsEngine &D) { Diags = &D; }
678 
getLangOpts()679   const LangOptions &getLangOpts() const { return LangOpts; }
getTargetInfo()680   const TargetInfo &getTargetInfo() const { return *Target; }
getFileManager()681   FileManager &getFileManager() const { return FileMgr; }
getSourceManager()682   SourceManager &getSourceManager() const { return SourceMgr; }
getHeaderSearchInfo()683   HeaderSearch &getHeaderSearchInfo() const { return HeaderInfo; }
684 
getIdentifierTable()685   IdentifierTable &getIdentifierTable() { return Identifiers; }
getIdentifierTable()686   const IdentifierTable &getIdentifierTable() const { return Identifiers; }
getSelectorTable()687   SelectorTable &getSelectorTable() { return Selectors; }
getBuiltinInfo()688   Builtin::Context &getBuiltinInfo() { return BuiltinInfo; }
getPreprocessorAllocator()689   llvm::BumpPtrAllocator &getPreprocessorAllocator() { return BP; }
690 
691   void setPTHManager(PTHManager* pm);
692 
getPTHManager()693   PTHManager *getPTHManager() { return PTH.get(); }
694 
setExternalSource(ExternalPreprocessorSource * Source)695   void setExternalSource(ExternalPreprocessorSource *Source) {
696     ExternalSource = Source;
697   }
698 
getExternalSource()699   ExternalPreprocessorSource *getExternalSource() const {
700     return ExternalSource;
701   }
702 
703   /// \brief Retrieve the module loader associated with this preprocessor.
getModuleLoader()704   ModuleLoader &getModuleLoader() const { return TheModuleLoader; }
705 
hadModuleLoaderFatalFailure()706   bool hadModuleLoaderFatalFailure() const {
707     return TheModuleLoader.HadFatalFailure;
708   }
709 
710   /// \brief True if we are currently preprocessing a #if or #elif directive
isParsingIfOrElifDirective()711   bool isParsingIfOrElifDirective() const {
712     return ParsingIfOrElifDirective;
713   }
714 
715   /// \brief Control whether the preprocessor retains comments in output.
SetCommentRetentionState(bool KeepComments,bool KeepMacroComments)716   void SetCommentRetentionState(bool KeepComments, bool KeepMacroComments) {
717     this->KeepComments = KeepComments | KeepMacroComments;
718     this->KeepMacroComments = KeepMacroComments;
719   }
720 
getCommentRetentionState()721   bool getCommentRetentionState() const { return KeepComments; }
722 
setPragmasEnabled(bool Enabled)723   void setPragmasEnabled(bool Enabled) { PragmasEnabled = Enabled; }
getPragmasEnabled()724   bool getPragmasEnabled() const { return PragmasEnabled; }
725 
SetSuppressIncludeNotFoundError(bool Suppress)726   void SetSuppressIncludeNotFoundError(bool Suppress) {
727     SuppressIncludeNotFoundError = Suppress;
728   }
729 
GetSuppressIncludeNotFoundError()730   bool GetSuppressIncludeNotFoundError() {
731     return SuppressIncludeNotFoundError;
732   }
733 
734   /// Sets whether the preprocessor is responsible for producing output or if
735   /// it is producing tokens to be consumed by Parse and Sema.
setPreprocessedOutput(bool IsPreprocessedOutput)736   void setPreprocessedOutput(bool IsPreprocessedOutput) {
737     PreprocessedOutput = IsPreprocessedOutput;
738   }
739 
740   /// Returns true if the preprocessor is responsible for generating output,
741   /// false if it is producing tokens to be consumed by Parse and Sema.
isPreprocessedOutput()742   bool isPreprocessedOutput() const { return PreprocessedOutput; }
743 
744   /// \brief Return true if we are lexing directly from the specified lexer.
isCurrentLexer(const PreprocessorLexer * L)745   bool isCurrentLexer(const PreprocessorLexer *L) const {
746     return CurPPLexer == L;
747   }
748 
749   /// \brief Return the current lexer being lexed from.
750   ///
751   /// Note that this ignores any potentially active macro expansions and _Pragma
752   /// expansions going on at the time.
getCurrentLexer()753   PreprocessorLexer *getCurrentLexer() const { return CurPPLexer; }
754 
755   /// \brief Return the current file lexer being lexed from.
756   ///
757   /// Note that this ignores any potentially active macro expansions and _Pragma
758   /// expansions going on at the time.
759   PreprocessorLexer *getCurrentFileLexer() const;
760 
761   /// \brief Return the submodule owning the file being lexed.
getCurrentSubmodule()762   Module *getCurrentSubmodule() const { return CurSubmodule; }
763 
764   /// \brief Returns the FileID for the preprocessor predefines.
getPredefinesFileID()765   FileID getPredefinesFileID() const { return PredefinesFileID; }
766 
767   /// \{
768   /// \brief Accessors for preprocessor callbacks.
769   ///
770   /// Note that this class takes ownership of any PPCallbacks object given to
771   /// it.
getPPCallbacks()772   PPCallbacks *getPPCallbacks() const { return Callbacks.get(); }
addPPCallbacks(std::unique_ptr<PPCallbacks> C)773   void addPPCallbacks(std::unique_ptr<PPCallbacks> C) {
774     if (Callbacks)
775       C = llvm::make_unique<PPChainedCallbacks>(std::move(C),
776                                                 std::move(Callbacks));
777     Callbacks = std::move(C);
778   }
779   /// \}
780 
isMacroDefined(StringRef Id)781   bool isMacroDefined(StringRef Id) {
782     return isMacroDefined(&Identifiers.get(Id));
783   }
isMacroDefined(const IdentifierInfo * II)784   bool isMacroDefined(const IdentifierInfo *II) {
785     return II->hasMacroDefinition() &&
786            (!getLangOpts().Modules || (bool)getMacroDefinition(II));
787   }
788 
789   /// \brief Determine whether II is defined as a macro within the module M,
790   /// if that is a module that we've already preprocessed. Does not check for
791   /// macros imported into M.
isMacroDefinedInLocalModule(const IdentifierInfo * II,Module * M)792   bool isMacroDefinedInLocalModule(const IdentifierInfo *II, Module *M) {
793     if (!II->hasMacroDefinition())
794       return false;
795     auto I = Submodules.find(M);
796     if (I == Submodules.end())
797       return false;
798     auto J = I->second.Macros.find(II);
799     if (J == I->second.Macros.end())
800       return false;
801     auto *MD = J->second.getLatest();
802     return MD && MD->isDefined();
803   }
804 
getMacroDefinition(const IdentifierInfo * II)805   MacroDefinition getMacroDefinition(const IdentifierInfo *II) {
806     if (!II->hasMacroDefinition())
807       return MacroDefinition();
808 
809     MacroState &S = CurSubmoduleState->Macros[II];
810     auto *MD = S.getLatest();
811     while (MD && isa<VisibilityMacroDirective>(MD))
812       MD = MD->getPrevious();
813     return MacroDefinition(dyn_cast_or_null<DefMacroDirective>(MD),
814                            S.getActiveModuleMacros(*this, II),
815                            S.isAmbiguous(*this, II));
816   }
817 
getMacroDefinitionAtLoc(const IdentifierInfo * II,SourceLocation Loc)818   MacroDefinition getMacroDefinitionAtLoc(const IdentifierInfo *II,
819                                           SourceLocation Loc) {
820     if (!II->hadMacroDefinition())
821       return MacroDefinition();
822 
823     MacroState &S = CurSubmoduleState->Macros[II];
824     MacroDirective::DefInfo DI;
825     if (auto *MD = S.getLatest())
826       DI = MD->findDirectiveAtLoc(Loc, getSourceManager());
827     // FIXME: Compute the set of active module macros at the specified location.
828     return MacroDefinition(DI.getDirective(),
829                            S.getActiveModuleMacros(*this, II),
830                            S.isAmbiguous(*this, II));
831   }
832 
833   /// \brief Given an identifier, return its latest non-imported MacroDirective
834   /// if it is \#define'd and not \#undef'd, or null if it isn't \#define'd.
getLocalMacroDirective(const IdentifierInfo * II)835   MacroDirective *getLocalMacroDirective(const IdentifierInfo *II) const {
836     if (!II->hasMacroDefinition())
837       return nullptr;
838 
839     auto *MD = getLocalMacroDirectiveHistory(II);
840     if (!MD || MD->getDefinition().isUndefined())
841       return nullptr;
842 
843     return MD;
844   }
845 
getMacroInfo(const IdentifierInfo * II)846   const MacroInfo *getMacroInfo(const IdentifierInfo *II) const {
847     return const_cast<Preprocessor*>(this)->getMacroInfo(II);
848   }
849 
getMacroInfo(const IdentifierInfo * II)850   MacroInfo *getMacroInfo(const IdentifierInfo *II) {
851     if (!II->hasMacroDefinition())
852       return nullptr;
853     if (auto MD = getMacroDefinition(II))
854       return MD.getMacroInfo();
855     return nullptr;
856   }
857 
858   /// \brief Given an identifier, return the latest non-imported macro
859   /// directive for that identifier.
860   ///
861   /// One can iterate over all previous macro directives from the most recent
862   /// one.
863   MacroDirective *getLocalMacroDirectiveHistory(const IdentifierInfo *II) const;
864 
865   /// \brief Add a directive to the macro directive history for this identifier.
866   void appendMacroDirective(IdentifierInfo *II, MacroDirective *MD);
appendDefMacroDirective(IdentifierInfo * II,MacroInfo * MI,SourceLocation Loc)867   DefMacroDirective *appendDefMacroDirective(IdentifierInfo *II, MacroInfo *MI,
868                                              SourceLocation Loc) {
869     DefMacroDirective *MD = AllocateDefMacroDirective(MI, Loc);
870     appendMacroDirective(II, MD);
871     return MD;
872   }
appendDefMacroDirective(IdentifierInfo * II,MacroInfo * MI)873   DefMacroDirective *appendDefMacroDirective(IdentifierInfo *II,
874                                              MacroInfo *MI) {
875     return appendDefMacroDirective(II, MI, MI->getDefinitionLoc());
876   }
877   /// \brief Set a MacroDirective that was loaded from a PCH file.
878   void setLoadedMacroDirective(IdentifierInfo *II, MacroDirective *MD);
879 
880   /// \brief Register an exported macro for a module and identifier.
881   ModuleMacro *addModuleMacro(Module *Mod, IdentifierInfo *II, MacroInfo *Macro,
882                               ArrayRef<ModuleMacro *> Overrides, bool &IsNew);
883   ModuleMacro *getModuleMacro(Module *Mod, IdentifierInfo *II);
884 
885   /// \brief Get the list of leaf (non-overridden) module macros for a name.
getLeafModuleMacros(const IdentifierInfo * II)886   ArrayRef<ModuleMacro*> getLeafModuleMacros(const IdentifierInfo *II) const {
887     auto I = LeafModuleMacros.find(II);
888     if (I != LeafModuleMacros.end())
889       return I->second;
890     return None;
891   }
892 
893   /// \{
894   /// Iterators for the macro history table. Currently defined macros have
895   /// IdentifierInfo::hasMacroDefinition() set and an empty
896   /// MacroInfo::getUndefLoc() at the head of the list.
897   typedef MacroMap::const_iterator macro_iterator;
898   macro_iterator macro_begin(bool IncludeExternalMacros = true) const;
899   macro_iterator macro_end(bool IncludeExternalMacros = true) const;
900   llvm::iterator_range<macro_iterator>
901   macros(bool IncludeExternalMacros = true) const {
902     return llvm::make_range(macro_begin(IncludeExternalMacros),
903                             macro_end(IncludeExternalMacros));
904   }
905   /// \}
906 
907   /// \brief Return the name of the macro defined before \p Loc that has
908   /// spelling \p Tokens.  If there are multiple macros with same spelling,
909   /// return the last one defined.
910   StringRef getLastMacroWithSpelling(SourceLocation Loc,
911                                      ArrayRef<TokenValue> Tokens) const;
912 
getPredefines()913   const std::string &getPredefines() const { return Predefines; }
914   /// \brief Set the predefines for this Preprocessor.
915   ///
916   /// These predefines are automatically injected when parsing the main file.
setPredefines(const char * P)917   void setPredefines(const char *P) { Predefines = P; }
setPredefines(StringRef P)918   void setPredefines(StringRef P) { Predefines = P; }
919 
920   /// Return information about the specified preprocessor
921   /// identifier token.
getIdentifierInfo(StringRef Name)922   IdentifierInfo *getIdentifierInfo(StringRef Name) const {
923     return &Identifiers.get(Name);
924   }
925 
926   /// \brief Add the specified pragma handler to this preprocessor.
927   ///
928   /// If \p Namespace is non-null, then it is a token required to exist on the
929   /// pragma line before the pragma string starts, e.g. "STDC" or "GCC".
930   void AddPragmaHandler(StringRef Namespace, PragmaHandler *Handler);
AddPragmaHandler(PragmaHandler * Handler)931   void AddPragmaHandler(PragmaHandler *Handler) {
932     AddPragmaHandler(StringRef(), Handler);
933   }
934 
935   /// \brief Remove the specific pragma handler from this preprocessor.
936   ///
937   /// If \p Namespace is non-null, then it should be the namespace that
938   /// \p Handler was added to. It is an error to remove a handler that
939   /// has not been registered.
940   void RemovePragmaHandler(StringRef Namespace, PragmaHandler *Handler);
RemovePragmaHandler(PragmaHandler * Handler)941   void RemovePragmaHandler(PragmaHandler *Handler) {
942     RemovePragmaHandler(StringRef(), Handler);
943   }
944 
945   /// Install empty handlers for all pragmas (making them ignored).
946   void IgnorePragmas();
947 
948   /// \brief Add the specified comment handler to the preprocessor.
949   void addCommentHandler(CommentHandler *Handler);
950 
951   /// \brief Remove the specified comment handler.
952   ///
953   /// It is an error to remove a handler that has not been registered.
954   void removeCommentHandler(CommentHandler *Handler);
955 
956   /// \brief Set the code completion handler to the given object.
setCodeCompletionHandler(CodeCompletionHandler & Handler)957   void setCodeCompletionHandler(CodeCompletionHandler &Handler) {
958     CodeComplete = &Handler;
959   }
960 
961   /// \brief Retrieve the current code-completion handler.
getCodeCompletionHandler()962   CodeCompletionHandler *getCodeCompletionHandler() const {
963     return CodeComplete;
964   }
965 
966   /// \brief Clear out the code completion handler.
clearCodeCompletionHandler()967   void clearCodeCompletionHandler() {
968     CodeComplete = nullptr;
969   }
970 
971   /// \brief Hook used by the lexer to invoke the "natural language" code
972   /// completion point.
973   void CodeCompleteNaturalLanguage();
974 
975   /// \brief Retrieve the preprocessing record, or NULL if there is no
976   /// preprocessing record.
getPreprocessingRecord()977   PreprocessingRecord *getPreprocessingRecord() const { return Record; }
978 
979   /// \brief Create a new preprocessing record, which will keep track of
980   /// all macro expansions, macro definitions, etc.
981   void createPreprocessingRecord();
982 
983   /// \brief Enter the specified FileID as the main source file,
984   /// which implicitly adds the builtin defines etc.
985   void EnterMainSourceFile();
986 
987   /// \brief Inform the preprocessor callbacks that processing is complete.
988   void EndSourceFile();
989 
990   /// \brief Add a source file to the top of the include stack and
991   /// start lexing tokens from it instead of the current buffer.
992   ///
993   /// Emits a diagnostic, doesn't enter the file, and returns true on error.
994   bool EnterSourceFile(FileID CurFileID, const DirectoryLookup *Dir,
995                        SourceLocation Loc);
996 
997   /// \brief Add a Macro to the top of the include stack and start lexing
998   /// tokens from it instead of the current buffer.
999   ///
1000   /// \param Args specifies the tokens input to a function-like macro.
1001   /// \param ILEnd specifies the location of the ')' for a function-like macro
1002   /// or the identifier for an object-like macro.
1003   void EnterMacro(Token &Identifier, SourceLocation ILEnd, MacroInfo *Macro,
1004                   MacroArgs *Args);
1005 
1006   /// \brief Add a "macro" context to the top of the include stack,
1007   /// which will cause the lexer to start returning the specified tokens.
1008   ///
1009   /// If \p DisableMacroExpansion is true, tokens lexed from the token stream
1010   /// will not be subject to further macro expansion. Otherwise, these tokens
1011   /// will be re-macro-expanded when/if expansion is enabled.
1012   ///
1013   /// If \p OwnsTokens is false, this method assumes that the specified stream
1014   /// of tokens has a permanent owner somewhere, so they do not need to be
1015   /// copied. If it is true, it assumes the array of tokens is allocated with
1016   /// \c new[] and must be freed.
1017   void EnterTokenStream(const Token *Toks, unsigned NumToks,
1018                         bool DisableMacroExpansion, bool OwnsTokens);
1019 
1020   /// \brief Pop the current lexer/macro exp off the top of the lexer stack.
1021   ///
1022   /// This should only be used in situations where the current state of the
1023   /// top-of-stack lexer is known.
1024   void RemoveTopOfLexerStack();
1025 
1026   /// From the point that this method is called, and until
1027   /// CommitBacktrackedTokens() or Backtrack() is called, the Preprocessor
1028   /// keeps track of the lexed tokens so that a subsequent Backtrack() call will
1029   /// make the Preprocessor re-lex the same tokens.
1030   ///
1031   /// Nested backtracks are allowed, meaning that EnableBacktrackAtThisPos can
1032   /// be called multiple times and CommitBacktrackedTokens/Backtrack calls will
1033   /// be combined with the EnableBacktrackAtThisPos calls in reverse order.
1034   ///
1035   /// NOTE: *DO NOT* forget to call either CommitBacktrackedTokens or Backtrack
1036   /// at some point after EnableBacktrackAtThisPos. If you don't, caching of
1037   /// tokens will continue indefinitely.
1038   ///
1039   void EnableBacktrackAtThisPos();
1040 
1041   /// \brief Disable the last EnableBacktrackAtThisPos call.
1042   void CommitBacktrackedTokens();
1043 
1044   /// \brief Make Preprocessor re-lex the tokens that were lexed since
1045   /// EnableBacktrackAtThisPos() was previously called.
1046   void Backtrack();
1047 
1048   /// \brief True if EnableBacktrackAtThisPos() was called and
1049   /// caching of tokens is on.
isBacktrackEnabled()1050   bool isBacktrackEnabled() const { return !BacktrackPositions.empty(); }
1051 
1052   /// \brief Lex the next token for this preprocessor.
1053   void Lex(Token &Result);
1054 
1055   void LexAfterModuleImport(Token &Result);
1056 
1057   void makeModuleVisible(Module *M, SourceLocation Loc);
1058 
getModuleImportLoc(Module * M)1059   SourceLocation getModuleImportLoc(Module *M) const {
1060     return CurSubmoduleState->VisibleModules.getImportLoc(M);
1061   }
1062 
1063   /// \brief Lex a string literal, which may be the concatenation of multiple
1064   /// string literals and may even come from macro expansion.
1065   /// \returns true on success, false if a error diagnostic has been generated.
LexStringLiteral(Token & Result,std::string & String,const char * DiagnosticTag,bool AllowMacroExpansion)1066   bool LexStringLiteral(Token &Result, std::string &String,
1067                         const char *DiagnosticTag, bool AllowMacroExpansion) {
1068     if (AllowMacroExpansion)
1069       Lex(Result);
1070     else
1071       LexUnexpandedToken(Result);
1072     return FinishLexStringLiteral(Result, String, DiagnosticTag,
1073                                   AllowMacroExpansion);
1074   }
1075 
1076   /// \brief Complete the lexing of a string literal where the first token has
1077   /// already been lexed (see LexStringLiteral).
1078   bool FinishLexStringLiteral(Token &Result, std::string &String,
1079                               const char *DiagnosticTag,
1080                               bool AllowMacroExpansion);
1081 
1082   /// \brief Lex a token.  If it's a comment, keep lexing until we get
1083   /// something not a comment.
1084   ///
1085   /// This is useful in -E -C mode where comments would foul up preprocessor
1086   /// directive handling.
LexNonComment(Token & Result)1087   void LexNonComment(Token &Result) {
1088     do
1089       Lex(Result);
1090     while (Result.getKind() == tok::comment);
1091   }
1092 
1093   /// \brief Just like Lex, but disables macro expansion of identifier tokens.
LexUnexpandedToken(Token & Result)1094   void LexUnexpandedToken(Token &Result) {
1095     // Disable macro expansion.
1096     bool OldVal = DisableMacroExpansion;
1097     DisableMacroExpansion = true;
1098     // Lex the token.
1099     Lex(Result);
1100 
1101     // Reenable it.
1102     DisableMacroExpansion = OldVal;
1103   }
1104 
1105   /// \brief Like LexNonComment, but this disables macro expansion of
1106   /// identifier tokens.
LexUnexpandedNonComment(Token & Result)1107   void LexUnexpandedNonComment(Token &Result) {
1108     do
1109       LexUnexpandedToken(Result);
1110     while (Result.getKind() == tok::comment);
1111   }
1112 
1113   /// \brief Parses a simple integer literal to get its numeric value.  Floating
1114   /// point literals and user defined literals are rejected.  Used primarily to
1115   /// handle pragmas that accept integer arguments.
1116   bool parseSimpleIntegerLiteral(Token &Tok, uint64_t &Value);
1117 
1118   /// Disables macro expansion everywhere except for preprocessor directives.
SetMacroExpansionOnlyInDirectives()1119   void SetMacroExpansionOnlyInDirectives() {
1120     DisableMacroExpansion = true;
1121     MacroExpansionInDirectivesOverride = true;
1122   }
1123 
1124   /// \brief Peeks ahead N tokens and returns that token without consuming any
1125   /// tokens.
1126   ///
1127   /// LookAhead(0) returns the next token that would be returned by Lex(),
1128   /// LookAhead(1) returns the token after it, etc.  This returns normal
1129   /// tokens after phase 5.  As such, it is equivalent to using
1130   /// 'Lex', not 'LexUnexpandedToken'.
LookAhead(unsigned N)1131   const Token &LookAhead(unsigned N) {
1132     if (CachedLexPos + N < CachedTokens.size())
1133       return CachedTokens[CachedLexPos+N];
1134     else
1135       return PeekAhead(N+1);
1136   }
1137 
1138   /// \brief When backtracking is enabled and tokens are cached,
1139   /// this allows to revert a specific number of tokens.
1140   ///
1141   /// Note that the number of tokens being reverted should be up to the last
1142   /// backtrack position, not more.
RevertCachedTokens(unsigned N)1143   void RevertCachedTokens(unsigned N) {
1144     assert(isBacktrackEnabled() &&
1145            "Should only be called when tokens are cached for backtracking");
1146     assert(signed(CachedLexPos) - signed(N) >= signed(BacktrackPositions.back())
1147          && "Should revert tokens up to the last backtrack position, not more");
1148     assert(signed(CachedLexPos) - signed(N) >= 0 &&
1149            "Corrupted backtrack positions ?");
1150     CachedLexPos -= N;
1151   }
1152 
1153   /// \brief Enters a token in the token stream to be lexed next.
1154   ///
1155   /// If BackTrack() is called afterwards, the token will remain at the
1156   /// insertion point.
EnterToken(const Token & Tok)1157   void EnterToken(const Token &Tok) {
1158     EnterCachingLexMode();
1159     CachedTokens.insert(CachedTokens.begin()+CachedLexPos, Tok);
1160   }
1161 
1162   /// We notify the Preprocessor that if it is caching tokens (because
1163   /// backtrack is enabled) it should replace the most recent cached tokens
1164   /// with the given annotation token. This function has no effect if
1165   /// backtracking is not enabled.
1166   ///
1167   /// Note that the use of this function is just for optimization, so that the
1168   /// cached tokens doesn't get re-parsed and re-resolved after a backtrack is
1169   /// invoked.
AnnotateCachedTokens(const Token & Tok)1170   void AnnotateCachedTokens(const Token &Tok) {
1171     assert(Tok.isAnnotation() && "Expected annotation token");
1172     if (CachedLexPos != 0 && isBacktrackEnabled())
1173       AnnotatePreviousCachedTokens(Tok);
1174   }
1175 
1176   /// Get the location of the last cached token, suitable for setting the end
1177   /// location of an annotation token.
getLastCachedTokenLocation()1178   SourceLocation getLastCachedTokenLocation() const {
1179     assert(CachedLexPos != 0);
1180     return CachedTokens[CachedLexPos-1].getLastLoc();
1181   }
1182 
1183   /// \brief Replace the last token with an annotation token.
1184   ///
1185   /// Like AnnotateCachedTokens(), this routine replaces an
1186   /// already-parsed (and resolved) token with an annotation
1187   /// token. However, this routine only replaces the last token with
1188   /// the annotation token; it does not affect any other cached
1189   /// tokens. This function has no effect if backtracking is not
1190   /// enabled.
ReplaceLastTokenWithAnnotation(const Token & Tok)1191   void ReplaceLastTokenWithAnnotation(const Token &Tok) {
1192     assert(Tok.isAnnotation() && "Expected annotation token");
1193     if (CachedLexPos != 0 && isBacktrackEnabled())
1194       CachedTokens[CachedLexPos-1] = Tok;
1195   }
1196 
1197   /// Update the current token to represent the provided
1198   /// identifier, in order to cache an action performed by typo correction.
TypoCorrectToken(const Token & Tok)1199   void TypoCorrectToken(const Token &Tok) {
1200     assert(Tok.getIdentifierInfo() && "Expected identifier token");
1201     if (CachedLexPos != 0 && isBacktrackEnabled())
1202       CachedTokens[CachedLexPos-1] = Tok;
1203   }
1204 
1205   /// \brief Recompute the current lexer kind based on the CurLexer/CurPTHLexer/
1206   /// CurTokenLexer pointers.
1207   void recomputeCurLexerKind();
1208 
1209   /// \brief Returns true if incremental processing is enabled
isIncrementalProcessingEnabled()1210   bool isIncrementalProcessingEnabled() const { return IncrementalProcessing; }
1211 
1212   /// \brief Enables the incremental processing
1213   void enableIncrementalProcessing(bool value = true) {
1214     IncrementalProcessing = value;
1215   }
1216 
1217   /// \brief Specify the point at which code-completion will be performed.
1218   ///
1219   /// \param File the file in which code completion should occur. If
1220   /// this file is included multiple times, code-completion will
1221   /// perform completion the first time it is included. If NULL, this
1222   /// function clears out the code-completion point.
1223   ///
1224   /// \param Line the line at which code completion should occur
1225   /// (1-based).
1226   ///
1227   /// \param Column the column at which code completion should occur
1228   /// (1-based).
1229   ///
1230   /// \returns true if an error occurred, false otherwise.
1231   bool SetCodeCompletionPoint(const FileEntry *File,
1232                               unsigned Line, unsigned Column);
1233 
1234   /// \brief Determine if we are performing code completion.
isCodeCompletionEnabled()1235   bool isCodeCompletionEnabled() const { return CodeCompletionFile != nullptr; }
1236 
1237   /// \brief Returns the location of the code-completion point.
1238   ///
1239   /// Returns an invalid location if code-completion is not enabled or the file
1240   /// containing the code-completion point has not been lexed yet.
getCodeCompletionLoc()1241   SourceLocation getCodeCompletionLoc() const { return CodeCompletionLoc; }
1242 
1243   /// \brief Returns the start location of the file of code-completion point.
1244   ///
1245   /// Returns an invalid location if code-completion is not enabled or the file
1246   /// containing the code-completion point has not been lexed yet.
getCodeCompletionFileLoc()1247   SourceLocation getCodeCompletionFileLoc() const {
1248     return CodeCompletionFileLoc;
1249   }
1250 
1251   /// \brief Returns true if code-completion is enabled and we have hit the
1252   /// code-completion point.
isCodeCompletionReached()1253   bool isCodeCompletionReached() const { return CodeCompletionReached; }
1254 
1255   /// \brief Note that we hit the code-completion point.
setCodeCompletionReached()1256   void setCodeCompletionReached() {
1257     assert(isCodeCompletionEnabled() && "Code-completion not enabled!");
1258     CodeCompletionReached = true;
1259     // Silence any diagnostics that occur after we hit the code-completion.
1260     getDiagnostics().setSuppressAllDiagnostics(true);
1261   }
1262 
1263   /// \brief The location of the currently-active \#pragma clang
1264   /// arc_cf_code_audited begin.
1265   ///
1266   /// Returns an invalid location if there is no such pragma active.
getPragmaARCCFCodeAuditedLoc()1267   SourceLocation getPragmaARCCFCodeAuditedLoc() const {
1268     return PragmaARCCFCodeAuditedLoc;
1269   }
1270 
1271   /// \brief Set the location of the currently-active \#pragma clang
1272   /// arc_cf_code_audited begin.  An invalid location ends the pragma.
setPragmaARCCFCodeAuditedLoc(SourceLocation Loc)1273   void setPragmaARCCFCodeAuditedLoc(SourceLocation Loc) {
1274     PragmaARCCFCodeAuditedLoc = Loc;
1275   }
1276 
1277   /// \brief The location of the currently-active \#pragma clang
1278   /// assume_nonnull begin.
1279   ///
1280   /// Returns an invalid location if there is no such pragma active.
getPragmaAssumeNonNullLoc()1281   SourceLocation getPragmaAssumeNonNullLoc() const {
1282     return PragmaAssumeNonNullLoc;
1283   }
1284 
1285   /// \brief Set the location of the currently-active \#pragma clang
1286   /// assume_nonnull begin.  An invalid location ends the pragma.
setPragmaAssumeNonNullLoc(SourceLocation Loc)1287   void setPragmaAssumeNonNullLoc(SourceLocation Loc) {
1288     PragmaAssumeNonNullLoc = Loc;
1289   }
1290 
1291   /// \brief Set the directory in which the main file should be considered
1292   /// to have been found, if it is not a real file.
setMainFileDir(const DirectoryEntry * Dir)1293   void setMainFileDir(const DirectoryEntry *Dir) {
1294     MainFileDir = Dir;
1295   }
1296 
1297   /// \brief Instruct the preprocessor to skip part of the main source file.
1298   ///
1299   /// \param Bytes The number of bytes in the preamble to skip.
1300   ///
1301   /// \param StartOfLine Whether skipping these bytes puts the lexer at the
1302   /// start of a line.
setSkipMainFilePreamble(unsigned Bytes,bool StartOfLine)1303   void setSkipMainFilePreamble(unsigned Bytes, bool StartOfLine) {
1304     SkipMainFilePreamble.first = Bytes;
1305     SkipMainFilePreamble.second = StartOfLine;
1306   }
1307 
1308   /// Forwarding function for diagnostics.  This emits a diagnostic at
1309   /// the specified Token's location, translating the token's start
1310   /// position in the current buffer into a SourcePosition object for rendering.
Diag(SourceLocation Loc,unsigned DiagID)1311   DiagnosticBuilder Diag(SourceLocation Loc, unsigned DiagID) const {
1312     return Diags->Report(Loc, DiagID);
1313   }
1314 
Diag(const Token & Tok,unsigned DiagID)1315   DiagnosticBuilder Diag(const Token &Tok, unsigned DiagID) const {
1316     return Diags->Report(Tok.getLocation(), DiagID);
1317   }
1318 
1319   /// Return the 'spelling' of the token at the given
1320   /// location; does not go up to the spelling location or down to the
1321   /// expansion location.
1322   ///
1323   /// \param buffer A buffer which will be used only if the token requires
1324   ///   "cleaning", e.g. if it contains trigraphs or escaped newlines
1325   /// \param invalid If non-null, will be set \c true if an error occurs.
1326   StringRef getSpelling(SourceLocation loc,
1327                         SmallVectorImpl<char> &buffer,
1328                         bool *invalid = nullptr) const {
1329     return Lexer::getSpelling(loc, buffer, SourceMgr, LangOpts, invalid);
1330   }
1331 
1332   /// \brief Return the 'spelling' of the Tok token.
1333   ///
1334   /// The spelling of a token is the characters used to represent the token in
1335   /// the source file after trigraph expansion and escaped-newline folding.  In
1336   /// particular, this wants to get the true, uncanonicalized, spelling of
1337   /// things like digraphs, UCNs, etc.
1338   ///
1339   /// \param Invalid If non-null, will be set \c true if an error occurs.
1340   std::string getSpelling(const Token &Tok, bool *Invalid = nullptr) const {
1341     return Lexer::getSpelling(Tok, SourceMgr, LangOpts, Invalid);
1342   }
1343 
1344   /// \brief Get the spelling of a token into a preallocated buffer, instead
1345   /// of as an std::string.
1346   ///
1347   /// The caller is required to allocate enough space for the token, which is
1348   /// guaranteed to be at least Tok.getLength() bytes long. The length of the
1349   /// actual result is returned.
1350   ///
1351   /// Note that this method may do two possible things: it may either fill in
1352   /// the buffer specified with characters, or it may *change the input pointer*
1353   /// to point to a constant buffer with the data already in it (avoiding a
1354   /// copy).  The caller is not allowed to modify the returned buffer pointer
1355   /// if an internal buffer is returned.
1356   unsigned getSpelling(const Token &Tok, const char *&Buffer,
1357                        bool *Invalid = nullptr) const {
1358     return Lexer::getSpelling(Tok, Buffer, SourceMgr, LangOpts, Invalid);
1359   }
1360 
1361   /// \brief Get the spelling of a token into a SmallVector.
1362   ///
1363   /// Note that the returned StringRef may not point to the
1364   /// supplied buffer if a copy can be avoided.
1365   StringRef getSpelling(const Token &Tok,
1366                         SmallVectorImpl<char> &Buffer,
1367                         bool *Invalid = nullptr) const;
1368 
1369   /// \brief Relex the token at the specified location.
1370   /// \returns true if there was a failure, false on success.
1371   bool getRawToken(SourceLocation Loc, Token &Result,
1372                    bool IgnoreWhiteSpace = false) {
1373     return Lexer::getRawToken(Loc, Result, SourceMgr, LangOpts, IgnoreWhiteSpace);
1374   }
1375 
1376   /// \brief Given a Token \p Tok that is a numeric constant with length 1,
1377   /// return the character.
1378   char
1379   getSpellingOfSingleCharacterNumericConstant(const Token &Tok,
1380                                               bool *Invalid = nullptr) const {
1381     assert(Tok.is(tok::numeric_constant) &&
1382            Tok.getLength() == 1 && "Called on unsupported token");
1383     assert(!Tok.needsCleaning() && "Token can't need cleaning with length 1");
1384 
1385     // If the token is carrying a literal data pointer, just use it.
1386     if (const char *D = Tok.getLiteralData())
1387       return *D;
1388 
1389     // Otherwise, fall back on getCharacterData, which is slower, but always
1390     // works.
1391     return *SourceMgr.getCharacterData(Tok.getLocation(), Invalid);
1392   }
1393 
1394   /// \brief Retrieve the name of the immediate macro expansion.
1395   ///
1396   /// This routine starts from a source location, and finds the name of the
1397   /// macro responsible for its immediate expansion. It looks through any
1398   /// intervening macro argument expansions to compute this. It returns a
1399   /// StringRef that refers to the SourceManager-owned buffer of the source
1400   /// where that macro name is spelled. Thus, the result shouldn't out-live
1401   /// the SourceManager.
getImmediateMacroName(SourceLocation Loc)1402   StringRef getImmediateMacroName(SourceLocation Loc) {
1403     return Lexer::getImmediateMacroName(Loc, SourceMgr, getLangOpts());
1404   }
1405 
1406   /// \brief Plop the specified string into a scratch buffer and set the
1407   /// specified token's location and length to it.
1408   ///
1409   /// If specified, the source location provides a location of the expansion
1410   /// point of the token.
1411   void CreateString(StringRef Str, Token &Tok,
1412                     SourceLocation ExpansionLocStart = SourceLocation(),
1413                     SourceLocation ExpansionLocEnd = SourceLocation());
1414 
1415   /// \brief Computes the source location just past the end of the
1416   /// token at this source location.
1417   ///
1418   /// This routine can be used to produce a source location that
1419   /// points just past the end of the token referenced by \p Loc, and
1420   /// is generally used when a diagnostic needs to point just after a
1421   /// token where it expected something different that it received. If
1422   /// the returned source location would not be meaningful (e.g., if
1423   /// it points into a macro), this routine returns an invalid
1424   /// source location.
1425   ///
1426   /// \param Offset an offset from the end of the token, where the source
1427   /// location should refer to. The default offset (0) produces a source
1428   /// location pointing just past the end of the token; an offset of 1 produces
1429   /// a source location pointing to the last character in the token, etc.
1430   SourceLocation getLocForEndOfToken(SourceLocation Loc, unsigned Offset = 0) {
1431     return Lexer::getLocForEndOfToken(Loc, Offset, SourceMgr, LangOpts);
1432   }
1433 
1434   /// \brief Returns true if the given MacroID location points at the first
1435   /// token of the macro expansion.
1436   ///
1437   /// \param MacroBegin If non-null and function returns true, it is set to
1438   /// begin location of the macro.
1439   bool isAtStartOfMacroExpansion(SourceLocation loc,
1440                                  SourceLocation *MacroBegin = nullptr) const {
1441     return Lexer::isAtStartOfMacroExpansion(loc, SourceMgr, LangOpts,
1442                                             MacroBegin);
1443   }
1444 
1445   /// \brief Returns true if the given MacroID location points at the last
1446   /// token of the macro expansion.
1447   ///
1448   /// \param MacroEnd If non-null and function returns true, it is set to
1449   /// end location of the macro.
1450   bool isAtEndOfMacroExpansion(SourceLocation loc,
1451                                SourceLocation *MacroEnd = nullptr) const {
1452     return Lexer::isAtEndOfMacroExpansion(loc, SourceMgr, LangOpts, MacroEnd);
1453   }
1454 
1455   /// \brief Print the token to stderr, used for debugging.
1456   void DumpToken(const Token &Tok, bool DumpFlags = false) const;
1457   void DumpLocation(SourceLocation Loc) const;
1458   void DumpMacro(const MacroInfo &MI) const;
1459   void dumpMacroInfo(const IdentifierInfo *II);
1460 
1461   /// \brief Given a location that specifies the start of a
1462   /// token, return a new location that specifies a character within the token.
AdvanceToTokenCharacter(SourceLocation TokStart,unsigned Char)1463   SourceLocation AdvanceToTokenCharacter(SourceLocation TokStart,
1464                                          unsigned Char) const {
1465     return Lexer::AdvanceToTokenCharacter(TokStart, Char, SourceMgr, LangOpts);
1466   }
1467 
1468   /// \brief Increment the counters for the number of token paste operations
1469   /// performed.
1470   ///
1471   /// If fast was specified, this is a 'fast paste' case we handled.
IncrementPasteCounter(bool isFast)1472   void IncrementPasteCounter(bool isFast) {
1473     if (isFast)
1474       ++NumFastTokenPaste;
1475     else
1476       ++NumTokenPaste;
1477   }
1478 
1479   void PrintStats();
1480 
1481   size_t getTotalMemory() const;
1482 
1483   /// When the macro expander pastes together a comment (/##/) in Microsoft
1484   /// mode, this method handles updating the current state, returning the
1485   /// token on the next source line.
1486   void HandleMicrosoftCommentPaste(Token &Tok);
1487 
1488   //===--------------------------------------------------------------------===//
1489   // Preprocessor callback methods.  These are invoked by a lexer as various
1490   // directives and events are found.
1491 
1492   /// Given a tok::raw_identifier token, look up the
1493   /// identifier information for the token and install it into the token,
1494   /// updating the token kind accordingly.
1495   IdentifierInfo *LookUpIdentifierInfo(Token &Identifier) const;
1496 
1497 private:
1498   llvm::DenseMap<IdentifierInfo*,unsigned> PoisonReasons;
1499 
1500 public:
1501 
1502   /// \brief Specifies the reason for poisoning an identifier.
1503   ///
1504   /// If that identifier is accessed while poisoned, then this reason will be
1505   /// used instead of the default "poisoned" diagnostic.
1506   void SetPoisonReason(IdentifierInfo *II, unsigned DiagID);
1507 
1508   /// \brief Display reason for poisoned identifier.
1509   void HandlePoisonedIdentifier(Token & Tok);
1510 
MaybeHandlePoisonedIdentifier(Token & Identifier)1511   void MaybeHandlePoisonedIdentifier(Token & Identifier) {
1512     if(IdentifierInfo * II = Identifier.getIdentifierInfo()) {
1513       if(II->isPoisoned()) {
1514         HandlePoisonedIdentifier(Identifier);
1515       }
1516     }
1517   }
1518 
1519 private:
1520   /// Identifiers used for SEH handling in Borland. These are only
1521   /// allowed in particular circumstances
1522   // __except block
1523   IdentifierInfo *Ident__exception_code,
1524                  *Ident___exception_code,
1525                  *Ident_GetExceptionCode;
1526   // __except filter expression
1527   IdentifierInfo *Ident__exception_info,
1528                  *Ident___exception_info,
1529                  *Ident_GetExceptionInfo;
1530   // __finally
1531   IdentifierInfo *Ident__abnormal_termination,
1532                  *Ident___abnormal_termination,
1533                  *Ident_AbnormalTermination;
1534 
1535   const char *getCurLexerEndPos();
1536 
1537 public:
1538   void PoisonSEHIdentifiers(bool Poison = true); // Borland
1539 
1540   /// \brief Callback invoked when the lexer reads an identifier and has
1541   /// filled in the tokens IdentifierInfo member.
1542   ///
1543   /// This callback potentially macro expands it or turns it into a named
1544   /// token (like 'for').
1545   ///
1546   /// \returns true if we actually computed a token, false if we need to
1547   /// lex again.
1548   bool HandleIdentifier(Token &Identifier);
1549 
1550 
1551   /// \brief Callback invoked when the lexer hits the end of the current file.
1552   ///
1553   /// This either returns the EOF token and returns true, or
1554   /// pops a level off the include stack and returns false, at which point the
1555   /// client should call lex again.
1556   bool HandleEndOfFile(Token &Result, bool isEndOfMacro = false);
1557 
1558   /// \brief Callback invoked when the current TokenLexer hits the end of its
1559   /// token stream.
1560   bool HandleEndOfTokenLexer(Token &Result);
1561 
1562   /// \brief Callback invoked when the lexer sees a # token at the start of a
1563   /// line.
1564   ///
1565   /// This consumes the directive, modifies the lexer/preprocessor state, and
1566   /// advances the lexer(s) so that the next token read is the correct one.
1567   void HandleDirective(Token &Result);
1568 
1569   /// \brief Ensure that the next token is a tok::eod token.
1570   ///
1571   /// If not, emit a diagnostic and consume up until the eod.
1572   /// If \p EnableMacros is true, then we consider macros that expand to zero
1573   /// tokens as being ok.
1574   void CheckEndOfDirective(const char *Directive, bool EnableMacros = false);
1575 
1576   /// \brief Read and discard all tokens remaining on the current line until
1577   /// the tok::eod token is found.
1578   void DiscardUntilEndOfDirective();
1579 
1580   /// \brief Returns true if the preprocessor has seen a use of
1581   /// __DATE__ or __TIME__ in the file so far.
SawDateOrTime()1582   bool SawDateOrTime() const {
1583     return DATELoc != SourceLocation() || TIMELoc != SourceLocation();
1584   }
getCounterValue()1585   unsigned getCounterValue() const { return CounterValue; }
setCounterValue(unsigned V)1586   void setCounterValue(unsigned V) { CounterValue = V; }
1587 
1588   /// \brief Retrieves the module that we're currently building, if any.
1589   Module *getCurrentModule();
1590 
1591   /// \brief Allocate a new MacroInfo object with the provided SourceLocation.
1592   MacroInfo *AllocateMacroInfo(SourceLocation L);
1593 
1594   /// \brief Allocate a new MacroInfo object loaded from an AST file.
1595   MacroInfo *AllocateDeserializedMacroInfo(SourceLocation L,
1596                                            unsigned SubModuleID);
1597 
1598   /// \brief Turn the specified lexer token into a fully checked and spelled
1599   /// filename, e.g. as an operand of \#include.
1600   ///
1601   /// The caller is expected to provide a buffer that is large enough to hold
1602   /// the spelling of the filename, but is also expected to handle the case
1603   /// when this method decides to use a different buffer.
1604   ///
1605   /// \returns true if the input filename was in <>'s or false if it was
1606   /// in ""'s.
1607   bool GetIncludeFilenameSpelling(SourceLocation Loc,StringRef &Filename);
1608 
1609   /// \brief Given a "foo" or \<foo> reference, look up the indicated file.
1610   ///
1611   /// Returns null on failure.  \p isAngled indicates whether the file
1612   /// reference is for system \#include's or not (i.e. using <> instead of "").
1613   const FileEntry *LookupFile(SourceLocation FilenameLoc, StringRef Filename,
1614                               bool isAngled, const DirectoryLookup *FromDir,
1615                               const FileEntry *FromFile,
1616                               const DirectoryLookup *&CurDir,
1617                               SmallVectorImpl<char> *SearchPath,
1618                               SmallVectorImpl<char> *RelativePath,
1619                               ModuleMap::KnownHeader *SuggestedModule,
1620                               bool SkipCache = false);
1621 
1622   /// \brief Get the DirectoryLookup structure used to find the current
1623   /// FileEntry, if CurLexer is non-null and if applicable.
1624   ///
1625   /// This allows us to implement \#include_next and find directory-specific
1626   /// properties.
GetCurDirLookup()1627   const DirectoryLookup *GetCurDirLookup() { return CurDirLookup; }
1628 
1629   /// \brief Return true if we're in the top-level file, not in a \#include.
1630   bool isInPrimaryFile() const;
1631 
1632   /// \brief Handle cases where the \#include name is expanded
1633   /// from a macro as multiple tokens, which need to be glued together.
1634   ///
1635   /// This occurs for code like:
1636   /// \code
1637   ///    \#define FOO <x/y.h>
1638   ///    \#include FOO
1639   /// \endcode
1640   /// because in this case, "<x/y.h>" is returned as 7 tokens, not one.
1641   ///
1642   /// This code concatenates and consumes tokens up to the '>' token.  It
1643   /// returns false if the > was found, otherwise it returns true if it finds
1644   /// and consumes the EOD marker.
1645   bool ConcatenateIncludeName(SmallString<128> &FilenameBuffer,
1646                               SourceLocation &End);
1647 
1648   /// \brief Lex an on-off-switch (C99 6.10.6p2) and verify that it is
1649   /// followed by EOD.  Return true if the token is not a valid on-off-switch.
1650   bool LexOnOffSwitch(tok::OnOffSwitch &OOS);
1651 
1652   bool CheckMacroName(Token &MacroNameTok, MacroUse isDefineUndef,
1653                       bool *ShadowFlag = nullptr);
1654 
1655 private:
1656 
PushIncludeMacroStack()1657   void PushIncludeMacroStack() {
1658     assert(CurLexerKind != CLK_CachingLexer && "cannot push a caching lexer");
1659     IncludeMacroStack.emplace_back(
1660         CurLexerKind, CurSubmodule, std::move(CurLexer), std::move(CurPTHLexer),
1661         CurPPLexer, std::move(CurTokenLexer), CurDirLookup);
1662     CurPPLexer = nullptr;
1663   }
1664 
PopIncludeMacroStack()1665   void PopIncludeMacroStack() {
1666     CurLexer = std::move(IncludeMacroStack.back().TheLexer);
1667     CurPTHLexer = std::move(IncludeMacroStack.back().ThePTHLexer);
1668     CurPPLexer = IncludeMacroStack.back().ThePPLexer;
1669     CurTokenLexer = std::move(IncludeMacroStack.back().TheTokenLexer);
1670     CurDirLookup  = IncludeMacroStack.back().TheDirLookup;
1671     CurSubmodule = IncludeMacroStack.back().TheSubmodule;
1672     CurLexerKind = IncludeMacroStack.back().CurLexerKind;
1673     IncludeMacroStack.pop_back();
1674   }
1675 
1676   void PropagateLineStartLeadingSpaceInfo(Token &Result);
1677 
1678   void EnterSubmodule(Module *M, SourceLocation ImportLoc);
1679   void LeaveSubmodule();
1680 
1681   /// Update the set of active module macros and ambiguity flag for a module
1682   /// macro name.
1683   void updateModuleMacroInfo(const IdentifierInfo *II, ModuleMacroInfo &Info);
1684 
1685   /// \brief Allocate a new MacroInfo object.
1686   MacroInfo *AllocateMacroInfo();
1687 
1688   DefMacroDirective *AllocateDefMacroDirective(MacroInfo *MI,
1689                                                SourceLocation Loc);
1690   UndefMacroDirective *AllocateUndefMacroDirective(SourceLocation UndefLoc);
1691   VisibilityMacroDirective *AllocateVisibilityMacroDirective(SourceLocation Loc,
1692                                                              bool isPublic);
1693 
1694   /// \brief Lex and validate a macro name, which occurs after a
1695   /// \#define or \#undef.
1696   ///
1697   /// \param MacroNameTok Token that represents the name defined or undefined.
1698   /// \param IsDefineUndef Kind if preprocessor directive.
1699   /// \param ShadowFlag Points to flag that is set if macro name shadows
1700   ///                   a keyword.
1701   ///
1702   /// This emits a diagnostic, sets the token kind to eod,
1703   /// and discards the rest of the macro line if the macro name is invalid.
1704   void ReadMacroName(Token &MacroNameTok, MacroUse IsDefineUndef = MU_Other,
1705                      bool *ShadowFlag = nullptr);
1706 
1707   /// The ( starting an argument list of a macro definition has just been read.
1708   /// Lex the rest of the arguments and the closing ), updating \p MI with
1709   /// what we learn and saving in \p LastTok the last token read.
1710   /// Return true if an error occurs parsing the arg list.
1711   bool ReadMacroDefinitionArgList(MacroInfo *MI, Token& LastTok);
1712 
1713   /// We just read a \#if or related directive and decided that the
1714   /// subsequent tokens are in the \#if'd out portion of the
1715   /// file.  Lex the rest of the file, until we see an \#endif.  If \p
1716   /// FoundNonSkipPortion is true, then we have already emitted code for part of
1717   /// this \#if directive, so \#else/\#elif blocks should never be entered. If
1718   /// \p FoundElse is false, then \#else directives are ok, if not, then we have
1719   /// already seen one so a \#else directive is a duplicate.  When this returns,
1720   /// the caller can lex the first valid token.
1721   void SkipExcludedConditionalBlock(SourceLocation IfTokenLoc,
1722                                     bool FoundNonSkipPortion, bool FoundElse,
1723                                     SourceLocation ElseLoc = SourceLocation());
1724 
1725   /// \brief A fast PTH version of SkipExcludedConditionalBlock.
1726   void PTHSkipExcludedConditionalBlock();
1727 
1728   /// \brief Evaluate an integer constant expression that may occur after a
1729   /// \#if or \#elif directive and return it as a bool.
1730   ///
1731   /// If the expression is equivalent to "!defined(X)" return X in IfNDefMacro.
1732   bool EvaluateDirectiveExpression(IdentifierInfo *&IfNDefMacro);
1733 
1734   /// \brief Install the standard preprocessor pragmas:
1735   /// \#pragma GCC poison/system_header/dependency and \#pragma once.
1736   void RegisterBuiltinPragmas();
1737 
1738   /// \brief Register builtin macros such as __LINE__ with the identifier table.
1739   void RegisterBuiltinMacros();
1740 
1741   /// If an identifier token is read that is to be expanded as a macro, handle
1742   /// it and return the next token as 'Tok'.  If we lexed a token, return true;
1743   /// otherwise the caller should lex again.
1744   bool HandleMacroExpandedIdentifier(Token &Tok, const MacroDefinition &MD);
1745 
1746   /// \brief Cache macro expanded tokens for TokenLexers.
1747   //
1748   /// Works like a stack; a TokenLexer adds the macro expanded tokens that is
1749   /// going to lex in the cache and when it finishes the tokens are removed
1750   /// from the end of the cache.
1751   Token *cacheMacroExpandedTokens(TokenLexer *tokLexer,
1752                                   ArrayRef<Token> tokens);
1753   void removeCachedMacroExpandedTokensOfLastLexer();
1754   friend void TokenLexer::ExpandFunctionArguments();
1755 
1756   /// Determine whether the next preprocessor token to be
1757   /// lexed is a '('.  If so, consume the token and return true, if not, this
1758   /// method should have no observable side-effect on the lexed tokens.
1759   bool isNextPPTokenLParen();
1760 
1761   /// After reading "MACRO(", this method is invoked to read all of the formal
1762   /// arguments specified for the macro invocation.  Returns null on error.
1763   MacroArgs *ReadFunctionLikeMacroArgs(Token &MacroName, MacroInfo *MI,
1764                                        SourceLocation &ExpansionEnd);
1765 
1766   /// \brief If an identifier token is read that is to be expanded
1767   /// as a builtin macro, handle it and return the next token as 'Tok'.
1768   void ExpandBuiltinMacro(Token &Tok);
1769 
1770   /// \brief Read a \c _Pragma directive, slice it up, process it, then
1771   /// return the first token after the directive.
1772   /// This assumes that the \c _Pragma token has just been read into \p Tok.
1773   void Handle_Pragma(Token &Tok);
1774 
1775   /// \brief Like Handle_Pragma except the pragma text is not enclosed within
1776   /// a string literal.
1777   void HandleMicrosoft__pragma(Token &Tok);
1778 
1779   /// \brief Add a lexer to the top of the include stack and
1780   /// start lexing tokens from it instead of the current buffer.
1781   void EnterSourceFileWithLexer(Lexer *TheLexer, const DirectoryLookup *Dir);
1782 
1783   /// \brief Add a lexer to the top of the include stack and
1784   /// start getting tokens from it using the PTH cache.
1785   void EnterSourceFileWithPTH(PTHLexer *PL, const DirectoryLookup *Dir);
1786 
1787   /// \brief Set the FileID for the preprocessor predefines.
setPredefinesFileID(FileID FID)1788   void setPredefinesFileID(FileID FID) {
1789     assert(PredefinesFileID.isInvalid() && "PredefinesFileID already set!");
1790     PredefinesFileID = FID;
1791   }
1792 
1793   /// \brief Returns true if we are lexing from a file and not a
1794   /// pragma or a macro.
IsFileLexer(const Lexer * L,const PreprocessorLexer * P)1795   static bool IsFileLexer(const Lexer* L, const PreprocessorLexer* P) {
1796     return L ? !L->isPragmaLexer() : P != nullptr;
1797   }
1798 
IsFileLexer(const IncludeStackInfo & I)1799   static bool IsFileLexer(const IncludeStackInfo& I) {
1800     return IsFileLexer(I.TheLexer.get(), I.ThePPLexer);
1801   }
1802 
IsFileLexer()1803   bool IsFileLexer() const {
1804     return IsFileLexer(CurLexer.get(), CurPPLexer);
1805   }
1806 
1807   //===--------------------------------------------------------------------===//
1808   // Caching stuff.
1809   void CachingLex(Token &Result);
InCachingLexMode()1810   bool InCachingLexMode() const {
1811     // If the Lexer pointers are 0 and IncludeMacroStack is empty, it means
1812     // that we are past EOF, not that we are in CachingLex mode.
1813     return !CurPPLexer && !CurTokenLexer && !CurPTHLexer &&
1814            !IncludeMacroStack.empty();
1815   }
1816   void EnterCachingLexMode();
ExitCachingLexMode()1817   void ExitCachingLexMode() {
1818     if (InCachingLexMode())
1819       RemoveTopOfLexerStack();
1820   }
1821   const Token &PeekAhead(unsigned N);
1822   void AnnotatePreviousCachedTokens(const Token &Tok);
1823 
1824   //===--------------------------------------------------------------------===//
1825   /// Handle*Directive - implement the various preprocessor directives.  These
1826   /// should side-effect the current preprocessor object so that the next call
1827   /// to Lex() will return the appropriate token next.
1828   void HandleLineDirective(Token &Tok);
1829   void HandleDigitDirective(Token &Tok);
1830   void HandleUserDiagnosticDirective(Token &Tok, bool isWarning);
1831   void HandleIdentSCCSDirective(Token &Tok);
1832   void HandleMacroPublicDirective(Token &Tok);
1833   void HandleMacroPrivateDirective(Token &Tok);
1834 
1835   // File inclusion.
1836   void HandleIncludeDirective(SourceLocation HashLoc,
1837                               Token &Tok,
1838                               const DirectoryLookup *LookupFrom = nullptr,
1839                               const FileEntry *LookupFromFile = nullptr,
1840                               bool isImport = false);
1841   void HandleIncludeNextDirective(SourceLocation HashLoc, Token &Tok);
1842   void HandleIncludeMacrosDirective(SourceLocation HashLoc, Token &Tok);
1843   void HandleImportDirective(SourceLocation HashLoc, Token &Tok);
1844   void HandleMicrosoftImportDirective(Token &Tok);
1845 
1846 public:
1847   // Module inclusion testing.
1848   /// \brief Find the module that owns the source or header file that
1849   /// \p Loc points to. If the location is in a file that was included
1850   /// into a module, or is outside any module, returns nullptr.
1851   Module *getModuleForLocation(SourceLocation Loc);
1852 
1853   /// \brief Find the module that contains the specified location, either
1854   /// directly or indirectly.
1855   Module *getModuleContainingLocation(SourceLocation Loc);
1856 
1857 private:
1858   // Macro handling.
1859   void HandleDefineDirective(Token &Tok, bool ImmediatelyAfterTopLevelIfndef);
1860   void HandleUndefDirective(Token &Tok);
1861 
1862   // Conditional Inclusion.
1863   void HandleIfdefDirective(Token &Tok, bool isIfndef,
1864                             bool ReadAnyTokensBeforeDirective);
1865   void HandleIfDirective(Token &Tok, bool ReadAnyTokensBeforeDirective);
1866   void HandleEndifDirective(Token &Tok);
1867   void HandleElseDirective(Token &Tok);
1868   void HandleElifDirective(Token &Tok);
1869 
1870   // Pragmas.
1871   void HandlePragmaDirective(SourceLocation IntroducerLoc,
1872                              PragmaIntroducerKind Introducer);
1873 public:
1874   void HandlePragmaOnce(Token &OnceTok);
1875   void HandlePragmaMark();
1876   void HandlePragmaPoison(Token &PoisonTok);
1877   void HandlePragmaSystemHeader(Token &SysHeaderTok);
1878   void HandlePragmaDependency(Token &DependencyTok);
1879   void HandlePragmaPushMacro(Token &Tok);
1880   void HandlePragmaPopMacro(Token &Tok);
1881   void HandlePragmaIncludeAlias(Token &Tok);
1882   IdentifierInfo *ParsePragmaPushOrPopMacro(Token &Tok);
1883 
1884   // Return true and store the first token only if any CommentHandler
1885   // has inserted some tokens and getCommentRetentionState() is false.
1886   bool HandleComment(Token &Token, SourceRange Comment);
1887 
1888   /// \brief A macro is used, update information about macros that need unused
1889   /// warnings.
1890   void markMacroAsUsed(MacroInfo *MI);
1891 };
1892 
1893 /// \brief Abstract base class that describes a handler that will receive
1894 /// source ranges for each of the comments encountered in the source file.
1895 class CommentHandler {
1896 public:
1897   virtual ~CommentHandler();
1898 
1899   // The handler shall return true if it has pushed any tokens
1900   // to be read using e.g. EnterToken or EnterTokenStream.
1901   virtual bool HandleComment(Preprocessor &PP, SourceRange Comment) = 0;
1902 };
1903 
1904 }  // end namespace clang
1905 
1906 #endif
1907