Home | History | Annotate | Download | only in Lex
      1 //===--- HeaderSearch.h - Resolve Header File Locations ---------*- 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 HeaderSearch interface.
     11 //
     12 //===----------------------------------------------------------------------===//
     13 
     14 #ifndef LLVM_CLANG_LEX_HEADERSEARCH_H
     15 #define LLVM_CLANG_LEX_HEADERSEARCH_H
     16 
     17 #include "clang/Lex/DirectoryLookup.h"
     18 #include "clang/Lex/ModuleMap.h"
     19 #include "llvm/ADT/StringMap.h"
     20 #include "llvm/ADT/StringSet.h"
     21 #include "llvm/Support/Allocator.h"
     22 #include "llvm/ADT/OwningPtr.h"
     23 #include <vector>
     24 
     25 namespace clang {
     26 
     27 class DiagnosticsEngine;
     28 class ExternalIdentifierLookup;
     29 class FileEntry;
     30 class FileManager;
     31 class IdentifierInfo;
     32 
     33 /// HeaderFileInfo - The preprocessor keeps track of this information for each
     34 /// file that is #included.
     35 struct HeaderFileInfo {
     36   /// isImport - True if this is a #import'd or #pragma once file.
     37   unsigned isImport : 1;
     38 
     39   /// isPragmaOnce - True if this is  #pragma once file.
     40   unsigned isPragmaOnce : 1;
     41 
     42   /// DirInfo - Keep track of whether this is a system header, and if so,
     43   /// whether it is C++ clean or not.  This can be set by the include paths or
     44   /// by #pragma gcc system_header.  This is an instance of
     45   /// SrcMgr::CharacteristicKind.
     46   unsigned DirInfo : 2;
     47 
     48   /// \brief Whether this header file info was supplied by an external source.
     49   unsigned External : 1;
     50 
     51   /// \brief Whether this structure is considered to already have been
     52   /// "resolved", meaning that it was loaded from the external source.
     53   unsigned Resolved : 1;
     54 
     55   /// \brief Whether this is a header inside a framework that is currently
     56   /// being built.
     57   ///
     58   /// When a framework is being built, the headers have not yet been placed
     59   /// into the appropriate framework subdirectories, and therefore are
     60   /// provided via a header map. This bit indicates when this is one of
     61   /// those framework headers.
     62   unsigned IndexHeaderMapHeader : 1;
     63 
     64   /// NumIncludes - This is the number of times the file has been included
     65   /// already.
     66   unsigned short NumIncludes;
     67 
     68   /// \brief The ID number of the controlling macro.
     69   ///
     70   /// This ID number will be non-zero when there is a controlling
     71   /// macro whose IdentifierInfo may not yet have been loaded from
     72   /// external storage.
     73   unsigned ControllingMacroID;
     74 
     75   /// ControllingMacro - If this file has a #ifndef XXX (or equivalent) guard
     76   /// that protects the entire contents of the file, this is the identifier
     77   /// for the macro that controls whether or not it has any effect.
     78   ///
     79   /// Note: Most clients should use getControllingMacro() to access
     80   /// the controlling macro of this header, since
     81   /// getControllingMacro() is able to load a controlling macro from
     82   /// external storage.
     83   const IdentifierInfo *ControllingMacro;
     84 
     85   /// \brief If this header came from a framework include, this is the name
     86   /// of the framework.
     87   StringRef Framework;
     88 
     89   HeaderFileInfo()
     90     : isImport(false), isPragmaOnce(false), DirInfo(SrcMgr::C_User),
     91       External(false), Resolved(false), IndexHeaderMapHeader(false),
     92       NumIncludes(0), ControllingMacroID(0), ControllingMacro(0)  {}
     93 
     94   /// \brief Retrieve the controlling macro for this header file, if
     95   /// any.
     96   const IdentifierInfo *getControllingMacro(ExternalIdentifierLookup *External);
     97 
     98   /// \brief Determine whether this is a non-default header file info, e.g.,
     99   /// it corresponds to an actual header we've included or tried to include.
    100   bool isNonDefault() const {
    101     return isImport || isPragmaOnce || NumIncludes || ControllingMacro ||
    102       ControllingMacroID;
    103   }
    104 };
    105 
    106 /// \brief An external source of header file information, which may supply
    107 /// information about header files already included.
    108 class ExternalHeaderFileInfoSource {
    109 public:
    110   virtual ~ExternalHeaderFileInfoSource();
    111 
    112   /// \brief Retrieve the header file information for the given file entry.
    113   ///
    114   /// \returns Header file information for the given file entry, with the
    115   /// \c External bit set. If the file entry is not known, return a
    116   /// default-constructed \c HeaderFileInfo.
    117   virtual HeaderFileInfo GetHeaderFileInfo(const FileEntry *FE) = 0;
    118 };
    119 
    120 /// HeaderSearch - This class encapsulates the information needed to find the
    121 /// file referenced by a #include or #include_next, (sub-)framework lookup, etc.
    122 class HeaderSearch {
    123   /// This structure is used to record entries in our framework cache.
    124   struct FrameworkCacheEntry {
    125     /// The directory entry which should be used for the cached framework.
    126     const DirectoryEntry *Directory;
    127 
    128     /// Whether this framework has been "user-specified" to be treated as if it
    129     /// were a system framework (even if it was found outside a system framework
    130     /// directory).
    131     bool IsUserSpecifiedSystemFramework;
    132   };
    133 
    134   FileManager &FileMgr;
    135   DiagnosticsEngine &Diags;
    136   /// #include search path information.  Requests for #include "x" search the
    137   /// directory of the #including file first, then each directory in SearchDirs
    138   /// consecutively. Requests for <x> search the current dir first, then each
    139   /// directory in SearchDirs, starting at AngledDirIdx, consecutively.  If
    140   /// NoCurDirSearch is true, then the check for the file in the current
    141   /// directory is suppressed.
    142   std::vector<DirectoryLookup> SearchDirs;
    143   unsigned AngledDirIdx;
    144   unsigned SystemDirIdx;
    145   bool NoCurDirSearch;
    146 
    147   /// \brief The path to the module cache.
    148   std::string ModuleCachePath;
    149 
    150   /// FileInfo - This contains all of the preprocessor-specific data about files
    151   /// that are included.  The vector is indexed by the FileEntry's UID.
    152   ///
    153   std::vector<HeaderFileInfo> FileInfo;
    154 
    155   /// LookupFileCache - This is keeps track of each lookup performed by
    156   /// LookupFile.  The first part of the value is the starting index in
    157   /// SearchDirs that the cached search was performed from.  If there is a hit
    158   /// and this value doesn't match the current query, the cache has to be
    159   /// ignored.  The second value is the entry in SearchDirs that satisfied the
    160   /// query.
    161   llvm::StringMap<std::pair<unsigned, unsigned>, llvm::BumpPtrAllocator>
    162     LookupFileCache;
    163 
    164   /// FrameworkMap - This is a collection mapping a framework or subframework
    165   /// name like "Carbon" to the Carbon.framework directory.
    166   llvm::StringMap<FrameworkCacheEntry, llvm::BumpPtrAllocator> FrameworkMap;
    167 
    168   /// IncludeAliases - maps include file names (including the quotes or
    169   /// angle brackets) to other include file names.  This is used to support the
    170   /// include_alias pragma for Microsoft compatibility.
    171   typedef llvm::StringMap<std::string, llvm::BumpPtrAllocator>
    172     IncludeAliasMap;
    173   OwningPtr<IncludeAliasMap> IncludeAliases;
    174 
    175   /// HeaderMaps - This is a mapping from FileEntry -> HeaderMap, uniquing
    176   /// headermaps.  This vector owns the headermap.
    177   std::vector<std::pair<const FileEntry*, const HeaderMap*> > HeaderMaps;
    178 
    179   /// \brief The mapping between modules and headers.
    180   ModuleMap ModMap;
    181 
    182   /// \brief Describes whether a given directory has a module map in it.
    183   llvm::DenseMap<const DirectoryEntry *, bool> DirectoryHasModuleMap;
    184 
    185   /// \brief Uniqued set of framework names, which is used to track which
    186   /// headers were included as framework headers.
    187   llvm::StringSet<llvm::BumpPtrAllocator> FrameworkNames;
    188 
    189   /// \brief Entity used to resolve the identifier IDs of controlling
    190   /// macros into IdentifierInfo pointers, as needed.
    191   ExternalIdentifierLookup *ExternalLookup;
    192 
    193   /// \brief Entity used to look up stored header file information.
    194   ExternalHeaderFileInfoSource *ExternalSource;
    195 
    196   // Various statistics we track for performance analysis.
    197   unsigned NumIncluded;
    198   unsigned NumMultiIncludeFileOptzn;
    199   unsigned NumFrameworkLookups, NumSubFrameworkLookups;
    200 
    201   // HeaderSearch doesn't support default or copy construction.
    202   explicit HeaderSearch();
    203   explicit HeaderSearch(const HeaderSearch&);
    204   void operator=(const HeaderSearch&);
    205 
    206   friend class DirectoryLookup;
    207 
    208 public:
    209   HeaderSearch(FileManager &FM, DiagnosticsEngine &Diags,
    210                const LangOptions &LangOpts, const TargetInfo *Target);
    211   ~HeaderSearch();
    212 
    213   FileManager &getFileMgr() const { return FileMgr; }
    214 
    215   /// SetSearchPaths - Interface for setting the file search paths.
    216   ///
    217   void SetSearchPaths(const std::vector<DirectoryLookup> &dirs,
    218                       unsigned angledDirIdx, unsigned systemDirIdx,
    219                       bool noCurDirSearch) {
    220     assert(angledDirIdx <= systemDirIdx && systemDirIdx <= dirs.size() &&
    221         "Directory indicies are unordered");
    222     SearchDirs = dirs;
    223     AngledDirIdx = angledDirIdx;
    224     SystemDirIdx = systemDirIdx;
    225     NoCurDirSearch = noCurDirSearch;
    226     //LookupFileCache.clear();
    227   }
    228 
    229   /// AddSearchPath - Add an additional search path.
    230   void AddSearchPath(const DirectoryLookup &dir, bool isAngled) {
    231     unsigned idx = isAngled ? SystemDirIdx : AngledDirIdx;
    232     SearchDirs.insert(SearchDirs.begin() + idx, dir);
    233     if (!isAngled)
    234       AngledDirIdx++;
    235     SystemDirIdx++;
    236   }
    237 
    238   /// HasIncludeAliasMap - Checks whether the map exists or not
    239   bool HasIncludeAliasMap() const {
    240     return IncludeAliases;
    241   }
    242 
    243   /// AddIncludeAlias - Map the source include name to the dest include name.
    244   /// The Source should include the angle brackets or quotes, the dest
    245   /// should not.  This allows for distinction between <> and "" headers.
    246   void AddIncludeAlias(StringRef Source, StringRef Dest) {
    247     if (!IncludeAliases)
    248       IncludeAliases.reset(new IncludeAliasMap);
    249     (*IncludeAliases)[Source] = Dest;
    250   }
    251 
    252   /// MapHeaderToIncludeAlias - Maps one header file name to a different header
    253   /// file name, for use with the include_alias pragma.  Note that the source
    254   /// file name should include the angle brackets or quotes.  Returns StringRef
    255   /// as null if the header cannot be mapped.
    256   StringRef MapHeaderToIncludeAlias(StringRef Source) {
    257     assert(IncludeAliases && "Trying to map headers when there's no map");
    258 
    259     // Do any filename replacements before anything else
    260     IncludeAliasMap::const_iterator Iter = IncludeAliases->find(Source);
    261     if (Iter != IncludeAliases->end())
    262       return Iter->second;
    263     return StringRef();
    264   }
    265 
    266   /// \brief Set the path to the module cache.
    267   void setModuleCachePath(StringRef CachePath) {
    268     ModuleCachePath = CachePath;
    269   }
    270 
    271   /// \brief Retrieve the path to the module cache.
    272   StringRef getModuleCachePath() const { return ModuleCachePath; }
    273 
    274   /// ClearFileInfo - Forget everything we know about headers so far.
    275   void ClearFileInfo() {
    276     FileInfo.clear();
    277   }
    278 
    279   void SetExternalLookup(ExternalIdentifierLookup *EIL) {
    280     ExternalLookup = EIL;
    281   }
    282 
    283   ExternalIdentifierLookup *getExternalLookup() const {
    284     return ExternalLookup;
    285   }
    286 
    287   /// \brief Set the external source of header information.
    288   void SetExternalSource(ExternalHeaderFileInfoSource *ES) {
    289     ExternalSource = ES;
    290   }
    291 
    292   /// \brief Set the target information for the header search, if not
    293   /// already known.
    294   void setTarget(const TargetInfo &Target);
    295 
    296   /// LookupFile - Given a "foo" or <foo> reference, look up the indicated file,
    297   /// return null on failure.
    298   ///
    299   /// \returns If successful, this returns 'UsedDir', the DirectoryLookup member
    300   /// the file was found in, or null if not applicable.
    301   ///
    302   /// \param isAngled indicates whether the file reference is a <> reference.
    303   ///
    304   /// \param CurDir If non-null, the file was found in the specified directory
    305   /// search location.  This is used to implement #include_next.
    306   ///
    307   /// \param CurFileEnt If non-null, indicates where the #including file is, in
    308   /// case a relative search is needed.
    309   ///
    310   /// \param SearchPath If non-null, will be set to the search path relative
    311   /// to which the file was found. If the include path is absolute, SearchPath
    312   /// will be set to an empty string.
    313   ///
    314   /// \param RelativePath If non-null, will be set to the path relative to
    315   /// SearchPath at which the file was found. This only differs from the
    316   /// Filename for framework includes.
    317   ///
    318   /// \param SuggestedModule If non-null, and the file found is semantically
    319   /// part of a known module, this will be set to the module that should
    320   /// be imported instead of preprocessing/parsing the file found.
    321   const FileEntry *LookupFile(StringRef Filename, bool isAngled,
    322                               const DirectoryLookup *FromDir,
    323                               const DirectoryLookup *&CurDir,
    324                               const FileEntry *CurFileEnt,
    325                               SmallVectorImpl<char> *SearchPath,
    326                               SmallVectorImpl<char> *RelativePath,
    327                               Module **SuggestedModule,
    328                               bool SkipCache = false);
    329 
    330   /// LookupSubframeworkHeader - Look up a subframework for the specified
    331   /// #include file.  For example, if #include'ing <HIToolbox/HIToolbox.h> from
    332   /// within ".../Carbon.framework/Headers/Carbon.h", check to see if HIToolbox
    333   /// is a subframework within Carbon.framework.  If so, return the FileEntry
    334   /// for the designated file, otherwise return null.
    335   const FileEntry *LookupSubframeworkHeader(
    336       StringRef Filename,
    337       const FileEntry *RelativeFileEnt,
    338       SmallVectorImpl<char> *SearchPath,
    339       SmallVectorImpl<char> *RelativePath);
    340 
    341   /// LookupFrameworkCache - Look up the specified framework name in our
    342   /// framework cache, returning the DirectoryEntry it is in if we know,
    343   /// otherwise, return null.
    344   FrameworkCacheEntry &LookupFrameworkCache(StringRef FWName) {
    345     return FrameworkMap.GetOrCreateValue(FWName).getValue();
    346   }
    347 
    348   /// ShouldEnterIncludeFile - Mark the specified file as a target of of a
    349   /// #include, #include_next, or #import directive.  Return false if #including
    350   /// the file will have no effect or true if we should include it.
    351   bool ShouldEnterIncludeFile(const FileEntry *File, bool isImport);
    352 
    353 
    354   /// getFileDirFlavor - Return whether the specified file is a normal header,
    355   /// a system header, or a C++ friendly system header.
    356   SrcMgr::CharacteristicKind getFileDirFlavor(const FileEntry *File) {
    357     return (SrcMgr::CharacteristicKind)getFileInfo(File).DirInfo;
    358   }
    359 
    360   /// MarkFileIncludeOnce - Mark the specified file as a "once only" file, e.g.
    361   /// due to #pragma once.
    362   void MarkFileIncludeOnce(const FileEntry *File) {
    363     HeaderFileInfo &FI = getFileInfo(File);
    364     FI.isImport = true;
    365     FI.isPragmaOnce = true;
    366   }
    367 
    368   /// MarkFileSystemHeader - Mark the specified file as a system header, e.g.
    369   /// due to #pragma GCC system_header.
    370   void MarkFileSystemHeader(const FileEntry *File) {
    371     getFileInfo(File).DirInfo = SrcMgr::C_System;
    372   }
    373 
    374   /// IncrementIncludeCount - Increment the count for the number of times the
    375   /// specified FileEntry has been entered.
    376   void IncrementIncludeCount(const FileEntry *File) {
    377     ++getFileInfo(File).NumIncludes;
    378   }
    379 
    380   /// SetFileControllingMacro - Mark the specified file as having a controlling
    381   /// macro.  This is used by the multiple-include optimization to eliminate
    382   /// no-op #includes.
    383   void SetFileControllingMacro(const FileEntry *File,
    384                                const IdentifierInfo *ControllingMacro) {
    385     getFileInfo(File).ControllingMacro = ControllingMacro;
    386   }
    387 
    388   /// \brief Determine whether this file is intended to be safe from
    389   /// multiple inclusions, e.g., it has #pragma once or a controlling
    390   /// macro.
    391   ///
    392   /// This routine does not consider the effect of #import
    393   bool isFileMultipleIncludeGuarded(const FileEntry *File);
    394 
    395   /// CreateHeaderMap - This method returns a HeaderMap for the specified
    396   /// FileEntry, uniquing them through the the 'HeaderMaps' datastructure.
    397   const HeaderMap *CreateHeaderMap(const FileEntry *FE);
    398 
    399   /// \brief Retrieve the name of the module file that should be used to
    400   /// load the given module.
    401   ///
    402   /// \param Module The module whose module file name will be returned.
    403   ///
    404   /// \returns The name of the module file that corresponds to this module,
    405   /// or an empty string if this module does not correspond to any module file.
    406   std::string getModuleFileName(Module *Module);
    407 
    408   /// \brief Retrieve the name of the module file that should be used to
    409   /// load a module with the given name.
    410   ///
    411   /// \param Module The module whose module file name will be returned.
    412   ///
    413   /// \returns The name of the module file that corresponds to this module,
    414   /// or an empty string if this module does not correspond to any module file.
    415   std::string getModuleFileName(StringRef ModuleName);
    416 
    417   /// \brief Lookup a module Search for a module with the given name.
    418   ///
    419   /// \param ModuleName The name of the module we're looking for.
    420   ///
    421   /// \param AllowSearch Whether we are allowed to search in the various
    422   /// search directories to produce a module definition. If not, this lookup
    423   /// will only return an already-known module.
    424   ///
    425   /// \returns The module with the given name.
    426   Module *lookupModule(StringRef ModuleName, bool AllowSearch = true);
    427 
    428   void IncrementFrameworkLookupCount() { ++NumFrameworkLookups; }
    429 
    430   /// \brief Determine whether there is a module map that may map the header
    431   /// with the given file name to a (sub)module.
    432   ///
    433   /// \param Filename The name of the file.
    434   ///
    435   /// \param Root The "root" directory, at which we should stop looking for
    436   /// module maps.
    437   bool hasModuleMap(StringRef Filename, const DirectoryEntry *Root);
    438 
    439   /// \brief Retrieve the module that corresponds to the given file, if any.
    440   ///
    441   /// \param File The header that we wish to map to a module.
    442   Module *findModuleForHeader(const FileEntry *File);
    443 
    444   /// \brief Read the contents of the given module map file.
    445   ///
    446   /// \param File The module map file.
    447   ///
    448   /// \param OnlyModule If non-NULL, this will receive the
    449   ///
    450   /// \returns true if an error occurred, false otherwise.
    451   bool loadModuleMapFile(const FileEntry *File);
    452 
    453   /// \brief Collect the set of all known, top-level modules.
    454   ///
    455   /// \param Modules Will be filled with the set of known, top-level modules.
    456   void collectAllModules(llvm::SmallVectorImpl<Module *> &Modules);
    457 
    458 private:
    459   /// \brief Retrieve a module with the given name, which may be part of the
    460   /// given framework.
    461   ///
    462   /// \param Name The name of the module to retrieve.
    463   ///
    464   /// \param Dir The framework directory (e.g., ModuleName.framework).
    465   ///
    466   /// \param IsSystem Whether the framework directory is part of the system
    467   /// frameworks.
    468   ///
    469   /// \returns The module, if found; otherwise, null.
    470   Module *loadFrameworkModule(StringRef Name,
    471                               const DirectoryEntry *Dir,
    472                               bool IsSystem);
    473 
    474 public:
    475   /// \brief Retrieve the module map.
    476   ModuleMap &getModuleMap() { return ModMap; }
    477 
    478   unsigned header_file_size() const { return FileInfo.size(); }
    479 
    480   // Used by ASTReader.
    481   void setHeaderFileInfoForUID(HeaderFileInfo HFI, unsigned UID);
    482 
    483   /// getFileInfo - Return the HeaderFileInfo structure for the specified
    484   /// FileEntry.
    485   const HeaderFileInfo &getFileInfo(const FileEntry *FE) const {
    486     return const_cast<HeaderSearch*>(this)->getFileInfo(FE);
    487   }
    488 
    489   // Used by external tools
    490   typedef std::vector<DirectoryLookup>::const_iterator search_dir_iterator;
    491   search_dir_iterator search_dir_begin() const { return SearchDirs.begin(); }
    492   search_dir_iterator search_dir_end() const { return SearchDirs.end(); }
    493   unsigned search_dir_size() const { return SearchDirs.size(); }
    494 
    495   search_dir_iterator quoted_dir_begin() const {
    496     return SearchDirs.begin();
    497   }
    498   search_dir_iterator quoted_dir_end() const {
    499     return SearchDirs.begin() + AngledDirIdx;
    500   }
    501 
    502   search_dir_iterator angled_dir_begin() const {
    503     return SearchDirs.begin() + AngledDirIdx;
    504   }
    505   search_dir_iterator angled_dir_end() const {
    506     return SearchDirs.begin() + SystemDirIdx;
    507   }
    508 
    509   search_dir_iterator system_dir_begin() const {
    510     return SearchDirs.begin() + SystemDirIdx;
    511   }
    512   search_dir_iterator system_dir_end() const { return SearchDirs.end(); }
    513 
    514   /// \brief Retrieve a uniqued framework name.
    515   StringRef getUniqueFrameworkName(StringRef Framework);
    516 
    517   void PrintStats();
    518 
    519   size_t getTotalMemory() const;
    520 
    521   static std::string NormalizeDashIncludePath(StringRef File,
    522                                               FileManager &FileMgr);
    523 
    524 private:
    525   /// \brief Describes what happened when we tried to load a module map file.
    526   enum LoadModuleMapResult {
    527     /// \brief The module map file had already been loaded.
    528     LMM_AlreadyLoaded,
    529     /// \brief The module map file was loaded by this invocation.
    530     LMM_NewlyLoaded,
    531     /// \brief There is was directory with the given name.
    532     LMM_NoDirectory,
    533     /// \brief There was either no module map file or the module map file was
    534     /// invalid.
    535     LMM_InvalidModuleMap
    536   };
    537 
    538   /// \brief Try to load the module map file in the given directory.
    539   ///
    540   /// \param DirName The name of the directory where we will look for a module
    541   /// map file.
    542   ///
    543   /// \returns The result of attempting to load the module map file from the
    544   /// named directory.
    545   LoadModuleMapResult loadModuleMapFile(StringRef DirName);
    546 
    547   /// \brief Try to load the module map file in the given directory.
    548   ///
    549   /// \param Dir The directory where we will look for a module map file.
    550   ///
    551   /// \returns The result of attempting to load the module map file from the
    552   /// named directory.
    553   LoadModuleMapResult loadModuleMapFile(const DirectoryEntry *Dir);
    554 
    555   /// getFileInfo - Return the HeaderFileInfo structure for the specified
    556   /// FileEntry.
    557   HeaderFileInfo &getFileInfo(const FileEntry *FE);
    558 };
    559 
    560 }  // end namespace clang
    561 
    562 #endif
    563