• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 //===--- SourceManager.h - Track and cache source files ---------*- 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 SourceManager interface.
12 ///
13 /// There are three different types of locations in a %file: a spelling
14 /// location, an expansion location, and a presumed location.
15 ///
16 /// Given an example of:
17 /// \code
18 /// #define min(x, y) x < y ? x : y
19 /// \endcode
20 ///
21 /// and then later on a use of min:
22 /// \code
23 /// #line 17
24 /// return min(a, b);
25 /// \endcode
26 ///
27 /// The expansion location is the line in the source code where the macro
28 /// was expanded (the return statement), the spelling location is the
29 /// location in the source where the macro was originally defined,
30 /// and the presumed location is where the line directive states that
31 /// the line is 17, or any other line.
32 ///
33 //===----------------------------------------------------------------------===//
34 
35 #ifndef LLVM_CLANG_SOURCEMANAGER_H
36 #define LLVM_CLANG_SOURCEMANAGER_H
37 
38 #include "clang/Basic/FileManager.h"
39 #include "clang/Basic/LLVM.h"
40 #include "clang/Basic/SourceLocation.h"
41 #include "llvm/ADT/ArrayRef.h"
42 #include "llvm/ADT/DenseMap.h"
43 #include "llvm/ADT/DenseSet.h"
44 #include "llvm/ADT/IntrusiveRefCntPtr.h"
45 #include "llvm/ADT/OwningPtr.h"
46 #include "llvm/ADT/PointerIntPair.h"
47 #include "llvm/ADT/PointerUnion.h"
48 #include "llvm/Support/Allocator.h"
49 #include "llvm/Support/DataTypes.h"
50 #include "llvm/Support/MemoryBuffer.h"
51 #include <cassert>
52 #include <map>
53 #include <vector>
54 
55 namespace clang {
56 
57 class DiagnosticsEngine;
58 class SourceManager;
59 class FileManager;
60 class FileEntry;
61 class LineTableInfo;
62 class LangOptions;
63 class ASTWriter;
64 class ASTReader;
65 
66 /// \brief Public enums and private classes that are part of the
67 /// SourceManager implementation.
68 ///
69 namespace SrcMgr {
70   /// \brief Indicates whether a file or directory holds normal user code,
71   /// system code, or system code which is implicitly 'extern "C"' in C++ mode.
72   ///
73   /// Entire directories can be tagged with this (this is maintained by
74   /// DirectoryLookup and friends) as can specific FileInfos when a \#pragma
75   /// system_header is seen or in various other cases.
76   ///
77   enum CharacteristicKind {
78     C_User, C_System, C_ExternCSystem
79   };
80 
81   /// \brief One instance of this struct is kept for every file loaded or used.
82   ///
83   /// This object owns the MemoryBuffer object.
84   class ContentCache {
85     enum CCFlags {
86       /// \brief Whether the buffer is invalid.
87       InvalidFlag = 0x01,
88       /// \brief Whether the buffer should not be freed on destruction.
89       DoNotFreeFlag = 0x02
90     };
91 
92     /// \brief The actual buffer containing the characters from the input
93     /// file.
94     ///
95     /// This is owned by the ContentCache object.  The bits indicate
96     /// whether the buffer is invalid.
97     mutable llvm::PointerIntPair<const llvm::MemoryBuffer *, 2> Buffer;
98 
99   public:
100     /// \brief Reference to the file entry representing this ContentCache.
101     ///
102     /// This reference does not own the FileEntry object.
103     ///
104     /// It is possible for this to be NULL if the ContentCache encapsulates
105     /// an imaginary text buffer.
106     const FileEntry *OrigEntry;
107 
108     /// \brief References the file which the contents were actually loaded from.
109     ///
110     /// Can be different from 'Entry' if we overridden the contents of one file
111     /// with the contents of another file.
112     const FileEntry *ContentsEntry;
113 
114     /// \brief A bump pointer allocated array of offsets for each source line.
115     ///
116     /// This is lazily computed.  This is owned by the SourceManager
117     /// BumpPointerAllocator object.
118     unsigned *SourceLineCache;
119 
120     /// \brief The number of lines in this ContentCache.
121     ///
122     /// This is only valid if SourceLineCache is non-null.
123     unsigned NumLines : 31;
124 
125     /// \brief Indicates whether the buffer itself was provided to override
126     /// the actual file contents.
127     ///
128     /// When true, the original entry may be a virtual file that does not
129     /// exist.
130     unsigned BufferOverridden : 1;
131 
132     /// \brief True if this content cache was initially created for a source
133     /// file considered as a system one.
134     unsigned IsSystemFile : 1;
135 
136     ContentCache(const FileEntry *Ent = 0)
137       : Buffer(0, false), OrigEntry(Ent), ContentsEntry(Ent),
138         SourceLineCache(0), NumLines(0), BufferOverridden(false),
139         IsSystemFile(false) {}
140 
ContentCache(const FileEntry * Ent,const FileEntry * contentEnt)141     ContentCache(const FileEntry *Ent, const FileEntry *contentEnt)
142       : Buffer(0, false), OrigEntry(Ent), ContentsEntry(contentEnt),
143         SourceLineCache(0), NumLines(0), BufferOverridden(false),
144         IsSystemFile(false) {}
145 
146     ~ContentCache();
147 
148     /// The copy ctor does not allow copies where source object has either
149     /// a non-NULL Buffer or SourceLineCache.  Ownership of allocated memory
150     /// is not transferred, so this is a logical error.
ContentCache(const ContentCache & RHS)151     ContentCache(const ContentCache &RHS)
152       : Buffer(0, false), SourceLineCache(0), BufferOverridden(false),
153         IsSystemFile(false)
154     {
155       OrigEntry = RHS.OrigEntry;
156       ContentsEntry = RHS.ContentsEntry;
157 
158       assert (RHS.Buffer.getPointer() == 0 && RHS.SourceLineCache == 0 &&
159               "Passed ContentCache object cannot own a buffer.");
160 
161       NumLines = RHS.NumLines;
162     }
163 
164     /// \brief Returns the memory buffer for the associated content.
165     ///
166     /// \param Diag Object through which diagnostics will be emitted if the
167     ///   buffer cannot be retrieved.
168     ///
169     /// \param Loc If specified, is the location that invalid file diagnostics
170     ///   will be emitted at.
171     ///
172     /// \param Invalid If non-NULL, will be set \c true if an error occurred.
173     const llvm::MemoryBuffer *getBuffer(DiagnosticsEngine &Diag,
174                                         const SourceManager &SM,
175                                         SourceLocation Loc = SourceLocation(),
176                                         bool *Invalid = 0) const;
177 
178     /// \brief Returns the size of the content encapsulated by this
179     /// ContentCache.
180     ///
181     /// This can be the size of the source file or the size of an
182     /// arbitrary scratch buffer.  If the ContentCache encapsulates a source
183     /// file this size is retrieved from the file's FileEntry.
184     unsigned getSize() const;
185 
186     /// \brief Returns the number of bytes actually mapped for this
187     /// ContentCache.
188     ///
189     /// This can be 0 if the MemBuffer was not actually expanded.
190     unsigned getSizeBytesMapped() const;
191 
192     /// Returns the kind of memory used to back the memory buffer for
193     /// this content cache.  This is used for performance analysis.
194     llvm::MemoryBuffer::BufferKind getMemoryBufferKind() const;
195 
setBuffer(const llvm::MemoryBuffer * B)196     void setBuffer(const llvm::MemoryBuffer *B) {
197       assert(!Buffer.getPointer() && "MemoryBuffer already set.");
198       Buffer.setPointer(B);
199       Buffer.setInt(false);
200     }
201 
202     /// \brief Get the underlying buffer, returning NULL if the buffer is not
203     /// yet available.
getRawBuffer()204     const llvm::MemoryBuffer *getRawBuffer() const {
205       return Buffer.getPointer();
206     }
207 
208     /// \brief Replace the existing buffer (which will be deleted)
209     /// with the given buffer.
210     void replaceBuffer(const llvm::MemoryBuffer *B, bool DoNotFree = false);
211 
212     /// \brief Determine whether the buffer itself is invalid.
isBufferInvalid()213     bool isBufferInvalid() const {
214       return Buffer.getInt() & InvalidFlag;
215     }
216 
217     /// \brief Determine whether the buffer should be freed.
shouldFreeBuffer()218     bool shouldFreeBuffer() const {
219       return (Buffer.getInt() & DoNotFreeFlag) == 0;
220     }
221 
222   private:
223     // Disable assignments.
224     ContentCache &operator=(const ContentCache& RHS) LLVM_DELETED_FUNCTION;
225   };
226 
227   /// \brief Information about a FileID, basically just the logical file
228   /// that it represents and include stack information.
229   ///
230   /// Each FileInfo has include stack information, indicating where it came
231   /// from. This information encodes the \#include chain that a token was
232   /// expanded from. The main include file has an invalid IncludeLoc.
233   ///
234   /// FileInfos contain a "ContentCache *", with the contents of the file.
235   ///
236   class FileInfo {
237     /// \brief The location of the \#include that brought in this file.
238     ///
239     /// This is an invalid SLOC for the main file (top of the \#include chain).
240     unsigned IncludeLoc;  // Really a SourceLocation
241 
242     /// \brief Number of FileIDs (files and macros) that were created during
243     /// preprocessing of this \#include, including this SLocEntry.
244     ///
245     /// Zero means the preprocessor didn't provide such info for this SLocEntry.
246     unsigned NumCreatedFIDs;
247 
248     /// \brief Contains the ContentCache* and the bits indicating the
249     /// characteristic of the file and whether it has \#line info, all
250     /// bitmangled together.
251     uintptr_t Data;
252 
253     friend class clang::SourceManager;
254     friend class clang::ASTWriter;
255     friend class clang::ASTReader;
256   public:
257     /// \brief Return a FileInfo object.
get(SourceLocation IL,const ContentCache * Con,CharacteristicKind FileCharacter)258     static FileInfo get(SourceLocation IL, const ContentCache *Con,
259                         CharacteristicKind FileCharacter) {
260       FileInfo X;
261       X.IncludeLoc = IL.getRawEncoding();
262       X.NumCreatedFIDs = 0;
263       X.Data = (uintptr_t)Con;
264       assert((X.Data & 7) == 0 &&"ContentCache pointer insufficiently aligned");
265       assert((unsigned)FileCharacter < 4 && "invalid file character");
266       X.Data |= (unsigned)FileCharacter;
267       return X;
268     }
269 
getIncludeLoc()270     SourceLocation getIncludeLoc() const {
271       return SourceLocation::getFromRawEncoding(IncludeLoc);
272     }
getContentCache()273     const ContentCache* getContentCache() const {
274       return reinterpret_cast<const ContentCache*>(Data & ~uintptr_t(7));
275     }
276 
277     /// \brief Return whether this is a system header or not.
getFileCharacteristic()278     CharacteristicKind getFileCharacteristic() const {
279       return (CharacteristicKind)(Data & 3);
280     }
281 
282     /// \brief Return true if this FileID has \#line directives in it.
hasLineDirectives()283     bool hasLineDirectives() const { return (Data & 4) != 0; }
284 
285     /// \brief Set the flag that indicates that this FileID has
286     /// line table entries associated with it.
setHasLineDirectives()287     void setHasLineDirectives() {
288       Data |= 4;
289     }
290   };
291 
292   /// \brief Each ExpansionInfo encodes the expansion location - where
293   /// the token was ultimately expanded, and the SpellingLoc - where the actual
294   /// character data for the token came from.
295   class ExpansionInfo {
296     // Really these are all SourceLocations.
297 
298     /// \brief Where the spelling for the token can be found.
299     unsigned SpellingLoc;
300 
301     /// In a macro expansion, ExpansionLocStart and ExpansionLocEnd
302     /// indicate the start and end of the expansion. In object-like macros,
303     /// they will be the same. In a function-like macro expansion, the start
304     /// will be the identifier and the end will be the ')'. Finally, in
305     /// macro-argument instantiations, the end will be 'SourceLocation()', an
306     /// invalid location.
307     unsigned ExpansionLocStart, ExpansionLocEnd;
308 
309   public:
getSpellingLoc()310     SourceLocation getSpellingLoc() const {
311       return SourceLocation::getFromRawEncoding(SpellingLoc);
312     }
getExpansionLocStart()313     SourceLocation getExpansionLocStart() const {
314       return SourceLocation::getFromRawEncoding(ExpansionLocStart);
315     }
getExpansionLocEnd()316     SourceLocation getExpansionLocEnd() const {
317       SourceLocation EndLoc =
318         SourceLocation::getFromRawEncoding(ExpansionLocEnd);
319       return EndLoc.isInvalid() ? getExpansionLocStart() : EndLoc;
320     }
321 
getExpansionLocRange()322     std::pair<SourceLocation,SourceLocation> getExpansionLocRange() const {
323       return std::make_pair(getExpansionLocStart(), getExpansionLocEnd());
324     }
325 
isMacroArgExpansion()326     bool isMacroArgExpansion() const {
327       // Note that this needs to return false for default constructed objects.
328       return getExpansionLocStart().isValid() &&
329         SourceLocation::getFromRawEncoding(ExpansionLocEnd).isInvalid();
330     }
331 
isMacroBodyExpansion()332     bool isMacroBodyExpansion() const {
333       return getExpansionLocStart().isValid() &&
334         SourceLocation::getFromRawEncoding(ExpansionLocEnd).isValid();
335     }
336 
isFunctionMacroExpansion()337     bool isFunctionMacroExpansion() const {
338       return getExpansionLocStart().isValid() &&
339           getExpansionLocStart() != getExpansionLocEnd();
340     }
341 
342     /// \brief Return a ExpansionInfo for an expansion.
343     ///
344     /// Start and End specify the expansion range (where the macro is
345     /// expanded), and SpellingLoc specifies the spelling location (where
346     /// the characters from the token come from). All three can refer to
347     /// normal File SLocs or expansion locations.
create(SourceLocation SpellingLoc,SourceLocation Start,SourceLocation End)348     static ExpansionInfo create(SourceLocation SpellingLoc,
349                                 SourceLocation Start, SourceLocation End) {
350       ExpansionInfo X;
351       X.SpellingLoc = SpellingLoc.getRawEncoding();
352       X.ExpansionLocStart = Start.getRawEncoding();
353       X.ExpansionLocEnd = End.getRawEncoding();
354       return X;
355     }
356 
357     /// \brief Return a special ExpansionInfo for the expansion of
358     /// a macro argument into a function-like macro's body.
359     ///
360     /// ExpansionLoc specifies the expansion location (where the macro is
361     /// expanded). This doesn't need to be a range because a macro is always
362     /// expanded at a macro parameter reference, and macro parameters are
363     /// always exactly one token. SpellingLoc specifies the spelling location
364     /// (where the characters from the token come from). ExpansionLoc and
365     /// SpellingLoc can both refer to normal File SLocs or expansion locations.
366     ///
367     /// Given the code:
368     /// \code
369     ///   #define F(x) f(x)
370     ///   F(42);
371     /// \endcode
372     ///
373     /// When expanding '\c F(42)', the '\c x' would call this with an
374     /// SpellingLoc pointing at '\c 42' and an ExpansionLoc pointing at its
375     /// location in the definition of '\c F'.
createForMacroArg(SourceLocation SpellingLoc,SourceLocation ExpansionLoc)376     static ExpansionInfo createForMacroArg(SourceLocation SpellingLoc,
377                                            SourceLocation ExpansionLoc) {
378       // We store an intentionally invalid source location for the end of the
379       // expansion range to mark that this is a macro argument ion rather than
380       // a normal one.
381       return create(SpellingLoc, ExpansionLoc, SourceLocation());
382     }
383   };
384 
385   /// \brief This is a discriminated union of FileInfo and ExpansionInfo.
386   ///
387   /// SourceManager keeps an array of these objects, and they are uniquely
388   /// identified by the FileID datatype.
389   class SLocEntry {
390     unsigned Offset;   // low bit is set for expansion info.
391     union {
392       FileInfo File;
393       ExpansionInfo Expansion;
394     };
395   public:
getOffset()396     unsigned getOffset() const { return Offset >> 1; }
397 
isExpansion()398     bool isExpansion() const { return Offset & 1; }
isFile()399     bool isFile() const { return !isExpansion(); }
400 
getFile()401     const FileInfo &getFile() const {
402       assert(isFile() && "Not a file SLocEntry!");
403       return File;
404     }
405 
getExpansion()406     const ExpansionInfo &getExpansion() const {
407       assert(isExpansion() && "Not a macro expansion SLocEntry!");
408       return Expansion;
409     }
410 
get(unsigned Offset,const FileInfo & FI)411     static SLocEntry get(unsigned Offset, const FileInfo &FI) {
412       SLocEntry E;
413       E.Offset = Offset << 1;
414       E.File = FI;
415       return E;
416     }
417 
get(unsigned Offset,const ExpansionInfo & Expansion)418     static SLocEntry get(unsigned Offset, const ExpansionInfo &Expansion) {
419       SLocEntry E;
420       E.Offset = (Offset << 1) | 1;
421       E.Expansion = Expansion;
422       return E;
423     }
424   };
425 }  // end SrcMgr namespace.
426 
427 /// \brief External source of source location entries.
428 class ExternalSLocEntrySource {
429 public:
430   virtual ~ExternalSLocEntrySource();
431 
432   /// \brief Read the source location entry with index ID, which will always be
433   /// less than -1.
434   ///
435   /// \returns true if an error occurred that prevented the source-location
436   /// entry from being loaded.
437   virtual bool ReadSLocEntry(int ID) = 0;
438 
439   /// \brief Retrieve the module import location and name for the given ID, if
440   /// in fact it was loaded from a module (rather than, say, a precompiled
441   /// header).
442   virtual std::pair<SourceLocation, StringRef> getModuleImportLoc(int ID) = 0;
443 };
444 
445 
446 /// \brief Holds the cache used by isBeforeInTranslationUnit.
447 ///
448 /// The cache structure is complex enough to be worth breaking out of
449 /// SourceManager.
450 class InBeforeInTUCacheEntry {
451   /// \brief The FileID's of the cached query.
452   ///
453   /// If these match up with a subsequent query, the result can be reused.
454   FileID LQueryFID, RQueryFID;
455 
456   /// \brief True if LQueryFID was created before RQueryFID.
457   ///
458   /// This is used to compare macro expansion locations.
459   bool IsLQFIDBeforeRQFID;
460 
461   /// \brief The file found in common between the two \#include traces, i.e.,
462   /// the nearest common ancestor of the \#include tree.
463   FileID CommonFID;
464 
465   /// \brief The offset of the previous query in CommonFID.
466   ///
467   /// Usually, this represents the location of the \#include for QueryFID, but
468   /// if LQueryFID is a parent of RQueryFID (or vice versa) then these can be a
469   /// random token in the parent.
470   unsigned LCommonOffset, RCommonOffset;
471 public:
472   /// \brief Return true if the currently cached values match up with
473   /// the specified LHS/RHS query.
474   ///
475   /// If not, we can't use the cache.
isCacheValid(FileID LHS,FileID RHS)476   bool isCacheValid(FileID LHS, FileID RHS) const {
477     return LQueryFID == LHS && RQueryFID == RHS;
478   }
479 
480   /// \brief If the cache is valid, compute the result given the
481   /// specified offsets in the LHS/RHS FileID's.
getCachedResult(unsigned LOffset,unsigned ROffset)482   bool getCachedResult(unsigned LOffset, unsigned ROffset) const {
483     // If one of the query files is the common file, use the offset.  Otherwise,
484     // use the #include loc in the common file.
485     if (LQueryFID != CommonFID) LOffset = LCommonOffset;
486     if (RQueryFID != CommonFID) ROffset = RCommonOffset;
487 
488     // It is common for multiple macro expansions to be "included" from the same
489     // location (expansion location), in which case use the order of the FileIDs
490     // to determine which came first. This will also take care the case where
491     // one of the locations points at the inclusion/expansion point of the other
492     // in which case its FileID will come before the other.
493     if (LOffset == ROffset)
494       return IsLQFIDBeforeRQFID;
495 
496     return LOffset < ROffset;
497   }
498 
499   /// \brief Set up a new query.
setQueryFIDs(FileID LHS,FileID RHS,bool isLFIDBeforeRFID)500   void setQueryFIDs(FileID LHS, FileID RHS, bool isLFIDBeforeRFID) {
501     assert(LHS != RHS);
502     LQueryFID = LHS;
503     RQueryFID = RHS;
504     IsLQFIDBeforeRQFID = isLFIDBeforeRFID;
505   }
506 
clear()507   void clear() {
508     LQueryFID = RQueryFID = FileID();
509     IsLQFIDBeforeRQFID = false;
510   }
511 
setCommonLoc(FileID commonFID,unsigned lCommonOffset,unsigned rCommonOffset)512   void setCommonLoc(FileID commonFID, unsigned lCommonOffset,
513                     unsigned rCommonOffset) {
514     CommonFID = commonFID;
515     LCommonOffset = lCommonOffset;
516     RCommonOffset = rCommonOffset;
517   }
518 
519 };
520 
521 /// \brief The stack used when building modules on demand, which is used
522 /// to provide a link between the source managers of the different compiler
523 /// instances.
524 typedef ArrayRef<std::pair<std::string, FullSourceLoc> > ModuleBuildStack;
525 
526 /// \brief This class handles loading and caching of source files into memory.
527 ///
528 /// This object owns the MemoryBuffer objects for all of the loaded
529 /// files and assigns unique FileID's for each unique \#include chain.
530 ///
531 /// The SourceManager can be queried for information about SourceLocation
532 /// objects, turning them into either spelling or expansion locations. Spelling
533 /// locations represent where the bytes corresponding to a token came from and
534 /// expansion locations represent where the location is in the user's view. In
535 /// the case of a macro expansion, for example, the spelling location indicates
536 /// where the expanded token came from and the expansion location specifies
537 /// where it was expanded.
538 class SourceManager : public RefCountedBase<SourceManager> {
539   /// \brief DiagnosticsEngine object.
540   DiagnosticsEngine &Diag;
541 
542   FileManager &FileMgr;
543 
544   mutable llvm::BumpPtrAllocator ContentCacheAlloc;
545 
546   /// \brief Memoized information about all of the files tracked by this
547   /// SourceManager.
548   ///
549   /// This map allows us to merge ContentCache entries based
550   /// on their FileEntry*.  All ContentCache objects will thus have unique,
551   /// non-null, FileEntry pointers.
552   llvm::DenseMap<const FileEntry*, SrcMgr::ContentCache*> FileInfos;
553 
554   /// \brief True if the ContentCache for files that are overriden by other
555   /// files, should report the original file name. Defaults to true.
556   bool OverridenFilesKeepOriginalName;
557 
558   /// \brief True if non-system source files should be treated as volatile
559   /// (likely to change while trying to use them). Defaults to false.
560   bool UserFilesAreVolatile;
561 
562   struct OverriddenFilesInfoTy {
563     /// \brief Files that have been overriden with the contents from another
564     /// file.
565     llvm::DenseMap<const FileEntry *, const FileEntry *> OverriddenFiles;
566     /// \brief Files that were overridden with a memory buffer.
567     llvm::DenseSet<const FileEntry *> OverriddenFilesWithBuffer;
568   };
569 
570   /// \brief Lazily create the object keeping overridden files info, since
571   /// it is uncommonly used.
572   OwningPtr<OverriddenFilesInfoTy> OverriddenFilesInfo;
573 
getOverriddenFilesInfo()574   OverriddenFilesInfoTy &getOverriddenFilesInfo() {
575     if (!OverriddenFilesInfo)
576       OverriddenFilesInfo.reset(new OverriddenFilesInfoTy);
577     return *OverriddenFilesInfo;
578   }
579 
580   /// \brief Information about various memory buffers that we have read in.
581   ///
582   /// All FileEntry* within the stored ContentCache objects are NULL,
583   /// as they do not refer to a file.
584   std::vector<SrcMgr::ContentCache*> MemBufferInfos;
585 
586   /// \brief The table of SLocEntries that are local to this module.
587   ///
588   /// Positive FileIDs are indexes into this table. Entry 0 indicates an invalid
589   /// expansion.
590   SmallVector<SrcMgr::SLocEntry, 0> LocalSLocEntryTable;
591 
592   /// \brief The table of SLocEntries that are loaded from other modules.
593   ///
594   /// Negative FileIDs are indexes into this table. To get from ID to an index,
595   /// use (-ID - 2).
596   mutable SmallVector<SrcMgr::SLocEntry, 0> LoadedSLocEntryTable;
597 
598   /// \brief The starting offset of the next local SLocEntry.
599   ///
600   /// This is LocalSLocEntryTable.back().Offset + the size of that entry.
601   unsigned NextLocalOffset;
602 
603   /// \brief The starting offset of the latest batch of loaded SLocEntries.
604   ///
605   /// This is LoadedSLocEntryTable.back().Offset, except that that entry might
606   /// not have been loaded, so that value would be unknown.
607   unsigned CurrentLoadedOffset;
608 
609   /// \brief The highest possible offset is 2^31-1, so CurrentLoadedOffset
610   /// starts at 2^31.
611   static const unsigned MaxLoadedOffset = 1U << 31U;
612 
613   /// \brief A bitmap that indicates whether the entries of LoadedSLocEntryTable
614   /// have already been loaded from the external source.
615   ///
616   /// Same indexing as LoadedSLocEntryTable.
617   std::vector<bool> SLocEntryLoaded;
618 
619   /// \brief An external source for source location entries.
620   ExternalSLocEntrySource *ExternalSLocEntries;
621 
622   /// \brief A one-entry cache to speed up getFileID.
623   ///
624   /// LastFileIDLookup records the last FileID looked up or created, because it
625   /// is very common to look up many tokens from the same file.
626   mutable FileID LastFileIDLookup;
627 
628   /// \brief Holds information for \#line directives.
629   ///
630   /// This is referenced by indices from SLocEntryTable.
631   LineTableInfo *LineTable;
632 
633   /// \brief These ivars serve as a cache used in the getLineNumber
634   /// method which is used to speedup getLineNumber calls to nearby locations.
635   mutable FileID LastLineNoFileIDQuery;
636   mutable SrcMgr::ContentCache *LastLineNoContentCache;
637   mutable unsigned LastLineNoFilePos;
638   mutable unsigned LastLineNoResult;
639 
640   /// \brief The file ID for the main source file of the translation unit.
641   FileID MainFileID;
642 
643   /// \brief The file ID for the precompiled preamble there is one.
644   FileID PreambleFileID;
645 
646   // Statistics for -print-stats.
647   mutable unsigned NumLinearScans, NumBinaryProbes;
648 
649   /// The key value into the IsBeforeInTUCache table.
650   typedef std::pair<FileID, FileID> IsBeforeInTUCacheKey;
651 
652   /// The IsBeforeInTranslationUnitCache is a mapping from FileID pairs
653   /// to cache results.
654   typedef llvm::DenseMap<IsBeforeInTUCacheKey, InBeforeInTUCacheEntry>
655           InBeforeInTUCache;
656 
657   /// Cache results for the isBeforeInTranslationUnit method.
658   mutable InBeforeInTUCache IBTUCache;
659   mutable InBeforeInTUCacheEntry IBTUCacheOverflow;
660 
661   /// Return the cache entry for comparing the given file IDs
662   /// for isBeforeInTranslationUnit.
663   InBeforeInTUCacheEntry &getInBeforeInTUCache(FileID LFID, FileID RFID) const;
664 
665   // Cache for the "fake" buffer used for error-recovery purposes.
666   mutable llvm::MemoryBuffer *FakeBufferForRecovery;
667 
668   mutable SrcMgr::ContentCache *FakeContentCacheForRecovery;
669 
670   /// \brief Lazily computed map of macro argument chunks to their expanded
671   /// source location.
672   typedef std::map<unsigned, SourceLocation> MacroArgsMap;
673 
674   mutable llvm::DenseMap<FileID, MacroArgsMap *> MacroArgsCacheMap;
675 
676   /// \brief The stack of modules being built, which is used to detect
677   /// cycles in the module dependency graph as modules are being built, as
678   /// well as to describe why we're rebuilding a particular module.
679   ///
680   /// There is no way to set this value from the command line. If we ever need
681   /// to do so (e.g., if on-demand module construction moves out-of-process),
682   /// we can add a cc1-level option to do so.
683   SmallVector<std::pair<std::string, FullSourceLoc>, 2> StoredModuleBuildStack;
684 
685   // SourceManager doesn't support copy construction.
686   explicit SourceManager(const SourceManager&) LLVM_DELETED_FUNCTION;
687   void operator=(const SourceManager&) LLVM_DELETED_FUNCTION;
688 public:
689   SourceManager(DiagnosticsEngine &Diag, FileManager &FileMgr,
690                 bool UserFilesAreVolatile = false);
691   ~SourceManager();
692 
693   void clearIDTables();
694 
getDiagnostics()695   DiagnosticsEngine &getDiagnostics() const { return Diag; }
696 
getFileManager()697   FileManager &getFileManager() const { return FileMgr; }
698 
699   /// \brief Set true if the SourceManager should report the original file name
700   /// for contents of files that were overriden by other files.Defaults to true.
setOverridenFilesKeepOriginalName(bool value)701   void setOverridenFilesKeepOriginalName(bool value) {
702     OverridenFilesKeepOriginalName = value;
703   }
704 
705   /// \brief True if non-system source files should be treated as volatile
706   /// (likely to change while trying to use them).
userFilesAreVolatile()707   bool userFilesAreVolatile() const { return UserFilesAreVolatile; }
708 
709   /// \brief Retrieve the module build stack.
getModuleBuildStack()710   ModuleBuildStack getModuleBuildStack() const {
711     return StoredModuleBuildStack;
712   }
713 
714   /// \brief Set the module build stack.
setModuleBuildStack(ModuleBuildStack stack)715   void setModuleBuildStack(ModuleBuildStack stack) {
716     StoredModuleBuildStack.clear();
717     StoredModuleBuildStack.append(stack.begin(), stack.end());
718   }
719 
720   /// \brief Push an entry to the module build stack.
pushModuleBuildStack(StringRef moduleName,FullSourceLoc importLoc)721   void pushModuleBuildStack(StringRef moduleName, FullSourceLoc importLoc) {
722     StoredModuleBuildStack.push_back(std::make_pair(moduleName.str(),importLoc));
723   }
724 
725   /// \brief Create the FileID for a memory buffer that will represent the
726   /// FileID for the main source.
727   ///
728   /// One example of when this would be used is when the main source is read
729   /// from STDIN.
730   FileID createMainFileIDForMemBuffer(const llvm::MemoryBuffer *Buffer,
731                              SrcMgr::CharacteristicKind Kind = SrcMgr::C_User) {
732     assert(MainFileID.isInvalid() && "MainFileID already set!");
733     MainFileID = createFileIDForMemBuffer(Buffer, Kind);
734     return MainFileID;
735   }
736 
737   //===--------------------------------------------------------------------===//
738   // MainFileID creation and querying methods.
739   //===--------------------------------------------------------------------===//
740 
741   /// \brief Returns the FileID of the main source file.
getMainFileID()742   FileID getMainFileID() const { return MainFileID; }
743 
744   /// \brief Create the FileID for the main source file.
745   FileID createMainFileID(const FileEntry *SourceFile,
746                           SrcMgr::CharacteristicKind Kind = SrcMgr::C_User) {
747     assert(MainFileID.isInvalid() && "MainFileID already set!");
748     MainFileID = createFileID(SourceFile, SourceLocation(), Kind);
749     return MainFileID;
750   }
751 
752   /// \brief Set the file ID for the main source file.
setMainFileID(FileID FID)753   void setMainFileID(FileID FID) {
754     assert(MainFileID.isInvalid() && "MainFileID already set!");
755     MainFileID = FID;
756   }
757 
758   /// \brief Set the file ID for the precompiled preamble.
setPreambleFileID(FileID Preamble)759   void setPreambleFileID(FileID Preamble) {
760     assert(PreambleFileID.isInvalid() && "PreambleFileID already set!");
761     PreambleFileID = Preamble;
762   }
763 
764   /// \brief Get the file ID for the precompiled preamble if there is one.
getPreambleFileID()765   FileID getPreambleFileID() const { return PreambleFileID; }
766 
767   //===--------------------------------------------------------------------===//
768   // Methods to create new FileID's and macro expansions.
769   //===--------------------------------------------------------------------===//
770 
771   /// \brief Create a new FileID that represents the specified file
772   /// being \#included from the specified IncludePosition.
773   ///
774   /// This translates NULL into standard input.
775   FileID createFileID(const FileEntry *SourceFile, SourceLocation IncludePos,
776                       SrcMgr::CharacteristicKind FileCharacter,
777                       int LoadedID = 0, unsigned LoadedOffset = 0) {
778     const SrcMgr::ContentCache *
779       IR = getOrCreateContentCache(SourceFile,
780                               /*isSystemFile=*/FileCharacter != SrcMgr::C_User);
781     assert(IR && "getOrCreateContentCache() cannot return NULL");
782     return createFileID(IR, IncludePos, FileCharacter, LoadedID, LoadedOffset);
783   }
784 
785   /// \brief Create a new FileID that represents the specified memory buffer.
786   ///
787   /// This does no caching of the buffer and takes ownership of the
788   /// MemoryBuffer, so only pass a MemoryBuffer to this once.
789   FileID createFileIDForMemBuffer(const llvm::MemoryBuffer *Buffer,
790                       SrcMgr::CharacteristicKind FileCharacter = SrcMgr::C_User,
791                                   int LoadedID = 0, unsigned LoadedOffset = 0,
792                                  SourceLocation IncludeLoc = SourceLocation()) {
793     return createFileID(createMemBufferContentCache(Buffer), IncludeLoc,
794                         FileCharacter, LoadedID, LoadedOffset);
795   }
796 
797   /// \brief Return a new SourceLocation that encodes the
798   /// fact that a token from SpellingLoc should actually be referenced from
799   /// ExpansionLoc, and that it represents the expansion of a macro argument
800   /// into the function-like macro body.
801   SourceLocation createMacroArgExpansionLoc(SourceLocation Loc,
802                                             SourceLocation ExpansionLoc,
803                                             unsigned TokLength);
804 
805   /// \brief Return a new SourceLocation that encodes the fact
806   /// that a token from SpellingLoc should actually be referenced from
807   /// ExpansionLoc.
808   SourceLocation createExpansionLoc(SourceLocation Loc,
809                                     SourceLocation ExpansionLocStart,
810                                     SourceLocation ExpansionLocEnd,
811                                     unsigned TokLength,
812                                     int LoadedID = 0,
813                                     unsigned LoadedOffset = 0);
814 
815   /// \brief Retrieve the memory buffer associated with the given file.
816   ///
817   /// \param Invalid If non-NULL, will be set \c true if an error
818   /// occurs while retrieving the memory buffer.
819   const llvm::MemoryBuffer *getMemoryBufferForFile(const FileEntry *File,
820                                                    bool *Invalid = 0);
821 
822   /// \brief Override the contents of the given source file by providing an
823   /// already-allocated buffer.
824   ///
825   /// \param SourceFile the source file whose contents will be overriden.
826   ///
827   /// \param Buffer the memory buffer whose contents will be used as the
828   /// data in the given source file.
829   ///
830   /// \param DoNotFree If true, then the buffer will not be freed when the
831   /// source manager is destroyed.
832   void overrideFileContents(const FileEntry *SourceFile,
833                             const llvm::MemoryBuffer *Buffer,
834                             bool DoNotFree = false);
835 
836   /// \brief Override the given source file with another one.
837   ///
838   /// \param SourceFile the source file which will be overriden.
839   ///
840   /// \param NewFile the file whose contents will be used as the
841   /// data instead of the contents of the given source file.
842   void overrideFileContents(const FileEntry *SourceFile,
843                             const FileEntry *NewFile);
844 
845   /// \brief Returns true if the file contents have been overridden.
isFileOverridden(const FileEntry * File)846   bool isFileOverridden(const FileEntry *File) {
847     if (OverriddenFilesInfo) {
848       if (OverriddenFilesInfo->OverriddenFilesWithBuffer.count(File))
849         return true;
850       if (OverriddenFilesInfo->OverriddenFiles.find(File) !=
851           OverriddenFilesInfo->OverriddenFiles.end())
852         return true;
853     }
854     return false;
855   }
856 
857   /// \brief Disable overridding the contents of a file, previously enabled
858   /// with #overrideFileContents.
859   ///
860   /// This should be called before parsing has begun.
861   void disableFileContentsOverride(const FileEntry *File);
862 
863   //===--------------------------------------------------------------------===//
864   // FileID manipulation methods.
865   //===--------------------------------------------------------------------===//
866 
867   /// \brief Return the buffer for the specified FileID.
868   ///
869   /// If there is an error opening this buffer the first time, this
870   /// manufactures a temporary buffer and returns a non-empty error string.
871   const llvm::MemoryBuffer *getBuffer(FileID FID, SourceLocation Loc,
872                                       bool *Invalid = 0) const {
873     bool MyInvalid = false;
874     const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &MyInvalid);
875     if (MyInvalid || !Entry.isFile()) {
876       if (Invalid)
877         *Invalid = true;
878 
879       return getFakeBufferForRecovery();
880     }
881 
882     return Entry.getFile().getContentCache()->getBuffer(Diag, *this, Loc,
883                                                         Invalid);
884   }
885 
886   const llvm::MemoryBuffer *getBuffer(FileID FID, bool *Invalid = 0) const {
887     bool MyInvalid = false;
888     const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &MyInvalid);
889     if (MyInvalid || !Entry.isFile()) {
890       if (Invalid)
891         *Invalid = true;
892 
893       return getFakeBufferForRecovery();
894     }
895 
896     return Entry.getFile().getContentCache()->getBuffer(Diag, *this,
897                                                         SourceLocation(),
898                                                         Invalid);
899   }
900 
901   /// \brief Returns the FileEntry record for the provided FileID.
getFileEntryForID(FileID FID)902   const FileEntry *getFileEntryForID(FileID FID) const {
903     bool MyInvalid = false;
904     const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &MyInvalid);
905     if (MyInvalid || !Entry.isFile())
906       return 0;
907 
908     const SrcMgr::ContentCache *Content = Entry.getFile().getContentCache();
909     if (!Content)
910       return 0;
911     return Content->OrigEntry;
912   }
913 
914   /// \brief Returns the FileEntry record for the provided SLocEntry.
getFileEntryForSLocEntry(const SrcMgr::SLocEntry & sloc)915   const FileEntry *getFileEntryForSLocEntry(const SrcMgr::SLocEntry &sloc) const
916   {
917     const SrcMgr::ContentCache *Content = sloc.getFile().getContentCache();
918     if (!Content)
919       return 0;
920     return Content->OrigEntry;
921   }
922 
923   /// \brief Return a StringRef to the source buffer data for the
924   /// specified FileID.
925   ///
926   /// \param FID The file ID whose contents will be returned.
927   /// \param Invalid If non-NULL, will be set true if an error occurred.
928   StringRef getBufferData(FileID FID, bool *Invalid = 0) const;
929 
930   /// \brief Get the number of FileIDs (files and macros) that were created
931   /// during preprocessing of \p FID, including it.
getNumCreatedFIDsForFileID(FileID FID)932   unsigned getNumCreatedFIDsForFileID(FileID FID) const {
933     bool Invalid = false;
934     const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
935     if (Invalid || !Entry.isFile())
936       return 0;
937 
938     return Entry.getFile().NumCreatedFIDs;
939   }
940 
941   /// \brief Set the number of FileIDs (files and macros) that were created
942   /// during preprocessing of \p FID, including it.
setNumCreatedFIDsForFileID(FileID FID,unsigned NumFIDs)943   void setNumCreatedFIDsForFileID(FileID FID, unsigned NumFIDs) const {
944     bool Invalid = false;
945     const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
946     if (Invalid || !Entry.isFile())
947       return;
948 
949     assert(Entry.getFile().NumCreatedFIDs == 0 && "Already set!");
950     const_cast<SrcMgr::FileInfo &>(Entry.getFile()).NumCreatedFIDs = NumFIDs;
951   }
952 
953   //===--------------------------------------------------------------------===//
954   // SourceLocation manipulation methods.
955   //===--------------------------------------------------------------------===//
956 
957   /// \brief Return the FileID for a SourceLocation.
958   ///
959   /// This is a very hot method that is used for all SourceManager queries
960   /// that start with a SourceLocation object.  It is responsible for finding
961   /// the entry in SLocEntryTable which contains the specified location.
962   ///
getFileID(SourceLocation SpellingLoc)963   FileID getFileID(SourceLocation SpellingLoc) const {
964     unsigned SLocOffset = SpellingLoc.getOffset();
965 
966     // If our one-entry cache covers this offset, just return it.
967     if (isOffsetInFileID(LastFileIDLookup, SLocOffset))
968       return LastFileIDLookup;
969 
970     return getFileIDSlow(SLocOffset);
971   }
972 
973   /// \brief Return the filename of the file containing a SourceLocation.
getFilename(SourceLocation SpellingLoc)974   StringRef getFilename(SourceLocation SpellingLoc) const {
975     if (const FileEntry *F = getFileEntryForID(getFileID(SpellingLoc)))
976       return F->getName();
977     return StringRef();
978   }
979 
980   /// \brief Return the source location corresponding to the first byte of
981   /// the specified file.
getLocForStartOfFile(FileID FID)982   SourceLocation getLocForStartOfFile(FileID FID) const {
983     bool Invalid = false;
984     const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
985     if (Invalid || !Entry.isFile())
986       return SourceLocation();
987 
988     unsigned FileOffset = Entry.getOffset();
989     return SourceLocation::getFileLoc(FileOffset);
990   }
991 
992   /// \brief Return the source location corresponding to the last byte of the
993   /// specified file.
getLocForEndOfFile(FileID FID)994   SourceLocation getLocForEndOfFile(FileID FID) const {
995     bool Invalid = false;
996     const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
997     if (Invalid || !Entry.isFile())
998       return SourceLocation();
999 
1000     unsigned FileOffset = Entry.getOffset();
1001     return SourceLocation::getFileLoc(FileOffset + getFileIDSize(FID) - 1);
1002   }
1003 
1004   /// \brief Returns the include location if \p FID is a \#include'd file
1005   /// otherwise it returns an invalid location.
getIncludeLoc(FileID FID)1006   SourceLocation getIncludeLoc(FileID FID) const {
1007     bool Invalid = false;
1008     const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
1009     if (Invalid || !Entry.isFile())
1010       return SourceLocation();
1011 
1012     return Entry.getFile().getIncludeLoc();
1013   }
1014 
1015   // \brief Returns the import location if the given source location is
1016   // located within a module, or an invalid location if the source location
1017   // is within the current translation unit.
1018   std::pair<SourceLocation, StringRef>
getModuleImportLoc(SourceLocation Loc)1019   getModuleImportLoc(SourceLocation Loc) const {
1020     FileID FID = getFileID(Loc);
1021 
1022     // Positive file IDs are in the current translation unit, and -1 is a
1023     // placeholder.
1024     if (FID.ID >= -1)
1025       return std::make_pair(SourceLocation(), "");
1026 
1027     return ExternalSLocEntries->getModuleImportLoc(FID.ID);
1028   }
1029 
1030   /// \brief Given a SourceLocation object \p Loc, return the expansion
1031   /// location referenced by the ID.
getExpansionLoc(SourceLocation Loc)1032   SourceLocation getExpansionLoc(SourceLocation Loc) const {
1033     // Handle the non-mapped case inline, defer to out of line code to handle
1034     // expansions.
1035     if (Loc.isFileID()) return Loc;
1036     return getExpansionLocSlowCase(Loc);
1037   }
1038 
1039   /// \brief Given \p Loc, if it is a macro location return the expansion
1040   /// location or the spelling location, depending on if it comes from a
1041   /// macro argument or not.
getFileLoc(SourceLocation Loc)1042   SourceLocation getFileLoc(SourceLocation Loc) const {
1043     if (Loc.isFileID()) return Loc;
1044     return getFileLocSlowCase(Loc);
1045   }
1046 
1047   /// \brief Return the start/end of the expansion information for an
1048   /// expansion location.
1049   ///
1050   /// \pre \p Loc is required to be an expansion location.
1051   std::pair<SourceLocation,SourceLocation>
1052   getImmediateExpansionRange(SourceLocation Loc) const;
1053 
1054   /// \brief Given a SourceLocation object, return the range of
1055   /// tokens covered by the expansion the ultimate file.
1056   std::pair<SourceLocation,SourceLocation>
1057   getExpansionRange(SourceLocation Loc) const;
1058 
1059 
1060   /// \brief Given a SourceLocation object, return the spelling
1061   /// location referenced by the ID.
1062   ///
1063   /// This is the place where the characters that make up the lexed token
1064   /// can be found.
getSpellingLoc(SourceLocation Loc)1065   SourceLocation getSpellingLoc(SourceLocation Loc) const {
1066     // Handle the non-mapped case inline, defer to out of line code to handle
1067     // expansions.
1068     if (Loc.isFileID()) return Loc;
1069     return getSpellingLocSlowCase(Loc);
1070   }
1071 
1072   /// \brief Given a SourceLocation object, return the spelling location
1073   /// referenced by the ID.
1074   ///
1075   /// This is the first level down towards the place where the characters
1076   /// that make up the lexed token can be found.  This should not generally
1077   /// be used by clients.
1078   SourceLocation getImmediateSpellingLoc(SourceLocation Loc) const;
1079 
1080   /// \brief Decompose the specified location into a raw FileID + Offset pair.
1081   ///
1082   /// The first element is the FileID, the second is the offset from the
1083   /// start of the buffer of the location.
getDecomposedLoc(SourceLocation Loc)1084   std::pair<FileID, unsigned> getDecomposedLoc(SourceLocation Loc) const {
1085     FileID FID = getFileID(Loc);
1086     bool Invalid = false;
1087     const SrcMgr::SLocEntry &E = getSLocEntry(FID, &Invalid);
1088     if (Invalid)
1089       return std::make_pair(FileID(), 0);
1090     return std::make_pair(FID, Loc.getOffset()-E.getOffset());
1091   }
1092 
1093   /// \brief Decompose the specified location into a raw FileID + Offset pair.
1094   ///
1095   /// If the location is an expansion record, walk through it until we find
1096   /// the final location expanded.
1097   std::pair<FileID, unsigned>
getDecomposedExpansionLoc(SourceLocation Loc)1098   getDecomposedExpansionLoc(SourceLocation Loc) const {
1099     FileID FID = getFileID(Loc);
1100     bool Invalid = false;
1101     const SrcMgr::SLocEntry *E = &getSLocEntry(FID, &Invalid);
1102     if (Invalid)
1103       return std::make_pair(FileID(), 0);
1104 
1105     unsigned Offset = Loc.getOffset()-E->getOffset();
1106     if (Loc.isFileID())
1107       return std::make_pair(FID, Offset);
1108 
1109     return getDecomposedExpansionLocSlowCase(E);
1110   }
1111 
1112   /// \brief Decompose the specified location into a raw FileID + Offset pair.
1113   ///
1114   /// If the location is an expansion record, walk through it until we find
1115   /// its spelling record.
1116   std::pair<FileID, unsigned>
getDecomposedSpellingLoc(SourceLocation Loc)1117   getDecomposedSpellingLoc(SourceLocation Loc) const {
1118     FileID FID = getFileID(Loc);
1119     bool Invalid = false;
1120     const SrcMgr::SLocEntry *E = &getSLocEntry(FID, &Invalid);
1121     if (Invalid)
1122       return std::make_pair(FileID(), 0);
1123 
1124     unsigned Offset = Loc.getOffset()-E->getOffset();
1125     if (Loc.isFileID())
1126       return std::make_pair(FID, Offset);
1127     return getDecomposedSpellingLocSlowCase(E, Offset);
1128   }
1129 
1130   /// \brief Returns the offset from the start of the file that the
1131   /// specified SourceLocation represents.
1132   ///
1133   /// This is not very meaningful for a macro ID.
getFileOffset(SourceLocation SpellingLoc)1134   unsigned getFileOffset(SourceLocation SpellingLoc) const {
1135     return getDecomposedLoc(SpellingLoc).second;
1136   }
1137 
1138   /// \brief Tests whether the given source location represents a macro
1139   /// argument's expansion into the function-like macro definition.
1140   ///
1141   /// Such source locations only appear inside of the expansion
1142   /// locations representing where a particular function-like macro was
1143   /// expanded.
1144   bool isMacroArgExpansion(SourceLocation Loc) const;
1145 
1146   /// \brief Tests whether the given source location represents the expansion of
1147   /// a macro body.
1148   ///
1149   /// This is equivalent to testing whether the location is part of a macro
1150   /// expansion but not the expansion of an argument to a function-like macro.
1151   bool isMacroBodyExpansion(SourceLocation Loc) const;
1152 
1153   /// \brief Returns true if \p Loc is inside the [\p Start, +\p Length)
1154   /// chunk of the source location address space.
1155   ///
1156   /// If it's true and \p RelativeOffset is non-null, it will be set to the
1157   /// relative offset of \p Loc inside the chunk.
1158   bool isInSLocAddrSpace(SourceLocation Loc,
1159                          SourceLocation Start, unsigned Length,
1160                          unsigned *RelativeOffset = 0) const {
1161     assert(((Start.getOffset() < NextLocalOffset &&
1162                Start.getOffset()+Length <= NextLocalOffset) ||
1163             (Start.getOffset() >= CurrentLoadedOffset &&
1164                 Start.getOffset()+Length < MaxLoadedOffset)) &&
1165            "Chunk is not valid SLoc address space");
1166     unsigned LocOffs = Loc.getOffset();
1167     unsigned BeginOffs = Start.getOffset();
1168     unsigned EndOffs = BeginOffs + Length;
1169     if (LocOffs >= BeginOffs && LocOffs < EndOffs) {
1170       if (RelativeOffset)
1171         *RelativeOffset = LocOffs - BeginOffs;
1172       return true;
1173     }
1174 
1175     return false;
1176   }
1177 
1178   /// \brief Return true if both \p LHS and \p RHS are in the local source
1179   /// location address space or the loaded one.
1180   ///
1181   /// If it's true and \p RelativeOffset is non-null, it will be set to the
1182   /// offset of \p RHS relative to \p LHS.
isInSameSLocAddrSpace(SourceLocation LHS,SourceLocation RHS,int * RelativeOffset)1183   bool isInSameSLocAddrSpace(SourceLocation LHS, SourceLocation RHS,
1184                              int *RelativeOffset) const {
1185     unsigned LHSOffs = LHS.getOffset(), RHSOffs = RHS.getOffset();
1186     bool LHSLoaded = LHSOffs >= CurrentLoadedOffset;
1187     bool RHSLoaded = RHSOffs >= CurrentLoadedOffset;
1188 
1189     if (LHSLoaded == RHSLoaded) {
1190       if (RelativeOffset)
1191         *RelativeOffset = RHSOffs - LHSOffs;
1192       return true;
1193     }
1194 
1195     return false;
1196   }
1197 
1198   //===--------------------------------------------------------------------===//
1199   // Queries about the code at a SourceLocation.
1200   //===--------------------------------------------------------------------===//
1201 
1202   /// \brief Return a pointer to the start of the specified location
1203   /// in the appropriate spelling MemoryBuffer.
1204   ///
1205   /// \param Invalid If non-NULL, will be set \c true if an error occurs.
1206   const char *getCharacterData(SourceLocation SL, bool *Invalid = 0) const;
1207 
1208   /// \brief Return the column # for the specified file position.
1209   ///
1210   /// This is significantly cheaper to compute than the line number.  This
1211   /// returns zero if the column number isn't known.  This may only be called
1212   /// on a file sloc, so you must choose a spelling or expansion location
1213   /// before calling this method.
1214   unsigned getColumnNumber(FileID FID, unsigned FilePos,
1215                            bool *Invalid = 0) const;
1216   unsigned getSpellingColumnNumber(SourceLocation Loc, bool *Invalid = 0) const;
1217   unsigned getExpansionColumnNumber(SourceLocation Loc,
1218                                     bool *Invalid = 0) const;
1219   unsigned getPresumedColumnNumber(SourceLocation Loc, bool *Invalid = 0) const;
1220 
1221 
1222   /// \brief Given a SourceLocation, return the spelling line number
1223   /// for the position indicated.
1224   ///
1225   /// This requires building and caching a table of line offsets for the
1226   /// MemoryBuffer, so this is not cheap: use only when about to emit a
1227   /// diagnostic.
1228   unsigned getLineNumber(FileID FID, unsigned FilePos, bool *Invalid = 0) const;
1229   unsigned getSpellingLineNumber(SourceLocation Loc, bool *Invalid = 0) const;
1230   unsigned getExpansionLineNumber(SourceLocation Loc, bool *Invalid = 0) const;
1231   unsigned getPresumedLineNumber(SourceLocation Loc, bool *Invalid = 0) const;
1232 
1233   /// \brief Return the filename or buffer identifier of the buffer the
1234   /// location is in.
1235   ///
1236   /// Note that this name does not respect \#line directives.  Use
1237   /// getPresumedLoc for normal clients.
1238   const char *getBufferName(SourceLocation Loc, bool *Invalid = 0) const;
1239 
1240   /// \brief Return the file characteristic of the specified source
1241   /// location, indicating whether this is a normal file, a system
1242   /// header, or an "implicit extern C" system header.
1243   ///
1244   /// This state can be modified with flags on GNU linemarker directives like:
1245   /// \code
1246   ///   # 4 "foo.h" 3
1247   /// \endcode
1248   /// which changes all source locations in the current file after that to be
1249   /// considered to be from a system header.
1250   SrcMgr::CharacteristicKind getFileCharacteristic(SourceLocation Loc) const;
1251 
1252   /// \brief Returns the "presumed" location of a SourceLocation specifies.
1253   ///
1254   /// A "presumed location" can be modified by \#line or GNU line marker
1255   /// directives.  This provides a view on the data that a user should see
1256   /// in diagnostics, for example.
1257   ///
1258   /// Note that a presumed location is always given as the expansion point of
1259   /// an expansion location, not at the spelling location.
1260   ///
1261   /// \returns The presumed location of the specified SourceLocation. If the
1262   /// presumed location cannot be calculate (e.g., because \p Loc is invalid
1263   /// or the file containing \p Loc has changed on disk), returns an invalid
1264   /// presumed location.
1265   PresumedLoc getPresumedLoc(SourceLocation Loc,
1266                              bool UseLineDirectives = true) const;
1267 
1268   /// \brief Returns true if both SourceLocations correspond to the same file.
isFromSameFile(SourceLocation Loc1,SourceLocation Loc2)1269   bool isFromSameFile(SourceLocation Loc1, SourceLocation Loc2) const {
1270     return getFileID(Loc1) == getFileID(Loc2);
1271   }
1272 
1273   /// \brief Returns true if the file of provided SourceLocation is the main
1274   /// file.
isFromMainFile(SourceLocation Loc)1275   bool isFromMainFile(SourceLocation Loc) const {
1276     return getFileID(Loc) == getMainFileID();
1277   }
1278 
1279   /// \brief Returns if a SourceLocation is in a system header.
isInSystemHeader(SourceLocation Loc)1280   bool isInSystemHeader(SourceLocation Loc) const {
1281     return getFileCharacteristic(Loc) != SrcMgr::C_User;
1282   }
1283 
1284   /// \brief Returns if a SourceLocation is in an "extern C" system header.
isInExternCSystemHeader(SourceLocation Loc)1285   bool isInExternCSystemHeader(SourceLocation Loc) const {
1286     return getFileCharacteristic(Loc) == SrcMgr::C_ExternCSystem;
1287   }
1288 
1289   /// \brief Returns whether \p Loc is expanded from a macro in a system header.
isInSystemMacro(SourceLocation loc)1290   bool isInSystemMacro(SourceLocation loc) {
1291     return loc.isMacroID() && isInSystemHeader(getSpellingLoc(loc));
1292   }
1293 
1294   /// \brief The size of the SLocEnty that \p FID represents.
1295   unsigned getFileIDSize(FileID FID) const;
1296 
1297   /// \brief Given a specific FileID, returns true if \p Loc is inside that
1298   /// FileID chunk and sets relative offset (offset of \p Loc from beginning
1299   /// of FileID) to \p relativeOffset.
1300   bool isInFileID(SourceLocation Loc, FileID FID,
1301                   unsigned *RelativeOffset = 0) const {
1302     unsigned Offs = Loc.getOffset();
1303     if (isOffsetInFileID(FID, Offs)) {
1304       if (RelativeOffset)
1305         *RelativeOffset = Offs - getSLocEntry(FID).getOffset();
1306       return true;
1307     }
1308 
1309     return false;
1310   }
1311 
1312   //===--------------------------------------------------------------------===//
1313   // Line Table Manipulation Routines
1314   //===--------------------------------------------------------------------===//
1315 
1316   /// \brief Return the uniqued ID for the specified filename.
1317   ///
1318   unsigned getLineTableFilenameID(StringRef Str);
1319 
1320   /// \brief Add a line note to the line table for the FileID and offset
1321   /// specified by Loc.
1322   ///
1323   /// If FilenameID is -1, it is considered to be unspecified.
1324   void AddLineNote(SourceLocation Loc, unsigned LineNo, int FilenameID);
1325   void AddLineNote(SourceLocation Loc, unsigned LineNo, int FilenameID,
1326                    bool IsFileEntry, bool IsFileExit,
1327                    bool IsSystemHeader, bool IsExternCHeader);
1328 
1329   /// \brief Determine if the source manager has a line table.
hasLineTable()1330   bool hasLineTable() const { return LineTable != 0; }
1331 
1332   /// \brief Retrieve the stored line table.
1333   LineTableInfo &getLineTable();
1334 
1335   //===--------------------------------------------------------------------===//
1336   // Queries for performance analysis.
1337   //===--------------------------------------------------------------------===//
1338 
1339   /// \brief Return the total amount of physical memory allocated by the
1340   /// ContentCache allocator.
getContentCacheSize()1341   size_t getContentCacheSize() const {
1342     return ContentCacheAlloc.getTotalMemory();
1343   }
1344 
1345   struct MemoryBufferSizes {
1346     const size_t malloc_bytes;
1347     const size_t mmap_bytes;
1348 
MemoryBufferSizesMemoryBufferSizes1349     MemoryBufferSizes(size_t malloc_bytes, size_t mmap_bytes)
1350       : malloc_bytes(malloc_bytes), mmap_bytes(mmap_bytes) {}
1351   };
1352 
1353   /// \brief Return the amount of memory used by memory buffers, breaking down
1354   /// by heap-backed versus mmap'ed memory.
1355   MemoryBufferSizes getMemoryBufferSizes() const;
1356 
1357   /// \brief Return the amount of memory used for various side tables and
1358   /// data structures in the SourceManager.
1359   size_t getDataStructureSizes() const;
1360 
1361   //===--------------------------------------------------------------------===//
1362   // Other miscellaneous methods.
1363   //===--------------------------------------------------------------------===//
1364 
1365   /// \brief Get the source location for the given file:line:col triplet.
1366   ///
1367   /// If the source file is included multiple times, the source location will
1368   /// be based upon the first inclusion.
1369   SourceLocation translateFileLineCol(const FileEntry *SourceFile,
1370                                       unsigned Line, unsigned Col) const;
1371 
1372   /// \brief Get the FileID for the given file.
1373   ///
1374   /// If the source file is included multiple times, the FileID will be the
1375   /// first inclusion.
1376   FileID translateFile(const FileEntry *SourceFile) const;
1377 
1378   /// \brief Get the source location in \p FID for the given line:col.
1379   /// Returns null location if \p FID is not a file SLocEntry.
1380   SourceLocation translateLineCol(FileID FID,
1381                                   unsigned Line, unsigned Col) const;
1382 
1383   /// \brief If \p Loc points inside a function macro argument, the returned
1384   /// location will be the macro location in which the argument was expanded.
1385   /// If a macro argument is used multiple times, the expanded location will
1386   /// be at the first expansion of the argument.
1387   /// e.g.
1388   ///   MY_MACRO(foo);
1389   ///             ^
1390   /// Passing a file location pointing at 'foo', will yield a macro location
1391   /// where 'foo' was expanded into.
1392   SourceLocation getMacroArgExpandedLocation(SourceLocation Loc) const;
1393 
1394   /// \brief Determines the order of 2 source locations in the translation unit.
1395   ///
1396   /// \returns true if LHS source location comes before RHS, false otherwise.
1397   bool isBeforeInTranslationUnit(SourceLocation LHS, SourceLocation RHS) const;
1398 
1399   /// \brief Determines the order of 2 source locations in the "source location
1400   /// address space".
isBeforeInSLocAddrSpace(SourceLocation LHS,SourceLocation RHS)1401   bool isBeforeInSLocAddrSpace(SourceLocation LHS, SourceLocation RHS) const {
1402     return isBeforeInSLocAddrSpace(LHS, RHS.getOffset());
1403   }
1404 
1405   /// \brief Determines the order of a source location and a source location
1406   /// offset in the "source location address space".
1407   ///
1408   /// Note that we always consider source locations loaded from
isBeforeInSLocAddrSpace(SourceLocation LHS,unsigned RHS)1409   bool isBeforeInSLocAddrSpace(SourceLocation LHS, unsigned RHS) const {
1410     unsigned LHSOffset = LHS.getOffset();
1411     bool LHSLoaded = LHSOffset >= CurrentLoadedOffset;
1412     bool RHSLoaded = RHS >= CurrentLoadedOffset;
1413     if (LHSLoaded == RHSLoaded)
1414       return LHSOffset < RHS;
1415 
1416     return LHSLoaded;
1417   }
1418 
1419   // Iterators over FileInfos.
1420   typedef llvm::DenseMap<const FileEntry*, SrcMgr::ContentCache*>
1421       ::const_iterator fileinfo_iterator;
fileinfo_begin()1422   fileinfo_iterator fileinfo_begin() const { return FileInfos.begin(); }
fileinfo_end()1423   fileinfo_iterator fileinfo_end() const { return FileInfos.end(); }
hasFileInfo(const FileEntry * File)1424   bool hasFileInfo(const FileEntry *File) const {
1425     return FileInfos.find(File) != FileInfos.end();
1426   }
1427 
1428   /// \brief Print statistics to stderr.
1429   ///
1430   void PrintStats() const;
1431 
1432   /// \brief Get the number of local SLocEntries we have.
local_sloc_entry_size()1433   unsigned local_sloc_entry_size() const { return LocalSLocEntryTable.size(); }
1434 
1435   /// \brief Get a local SLocEntry. This is exposed for indexing.
1436   const SrcMgr::SLocEntry &getLocalSLocEntry(unsigned Index,
1437                                              bool *Invalid = 0) const {
1438     assert(Index < LocalSLocEntryTable.size() && "Invalid index");
1439     return LocalSLocEntryTable[Index];
1440   }
1441 
1442   /// \brief Get the number of loaded SLocEntries we have.
loaded_sloc_entry_size()1443   unsigned loaded_sloc_entry_size() const { return LoadedSLocEntryTable.size();}
1444 
1445   /// \brief Get a loaded SLocEntry. This is exposed for indexing.
1446   const SrcMgr::SLocEntry &getLoadedSLocEntry(unsigned Index,
1447                                               bool *Invalid = 0) const {
1448     assert(Index < LoadedSLocEntryTable.size() && "Invalid index");
1449     if (SLocEntryLoaded[Index])
1450       return LoadedSLocEntryTable[Index];
1451     return loadSLocEntry(Index, Invalid);
1452   }
1453 
1454   const SrcMgr::SLocEntry &getSLocEntry(FileID FID, bool *Invalid = 0) const {
1455     if (FID.ID == 0 || FID.ID == -1) {
1456       if (Invalid) *Invalid = true;
1457       return LocalSLocEntryTable[0];
1458     }
1459     return getSLocEntryByID(FID.ID);
1460   }
1461 
getNextLocalOffset()1462   unsigned getNextLocalOffset() const { return NextLocalOffset; }
1463 
setExternalSLocEntrySource(ExternalSLocEntrySource * Source)1464   void setExternalSLocEntrySource(ExternalSLocEntrySource *Source) {
1465     assert(LoadedSLocEntryTable.empty() &&
1466            "Invalidating existing loaded entries");
1467     ExternalSLocEntries = Source;
1468   }
1469 
1470   /// \brief Allocate a number of loaded SLocEntries, which will be actually
1471   /// loaded on demand from the external source.
1472   ///
1473   /// NumSLocEntries will be allocated, which occupy a total of TotalSize space
1474   /// in the global source view. The lowest ID and the base offset of the
1475   /// entries will be returned.
1476   std::pair<int, unsigned>
1477   AllocateLoadedSLocEntries(unsigned NumSLocEntries, unsigned TotalSize);
1478 
1479   /// \brief Returns true if \p Loc came from a PCH/Module.
isLoadedSourceLocation(SourceLocation Loc)1480   bool isLoadedSourceLocation(SourceLocation Loc) const {
1481     return Loc.getOffset() >= CurrentLoadedOffset;
1482   }
1483 
1484   /// \brief Returns true if \p Loc did not come from a PCH/Module.
isLocalSourceLocation(SourceLocation Loc)1485   bool isLocalSourceLocation(SourceLocation Loc) const {
1486     return Loc.getOffset() < NextLocalOffset;
1487   }
1488 
1489   /// \brief Returns true if \p FID came from a PCH/Module.
isLoadedFileID(FileID FID)1490   bool isLoadedFileID(FileID FID) const {
1491     assert(FID.ID != -1 && "Using FileID sentinel value");
1492     return FID.ID < 0;
1493   }
1494 
1495   /// \brief Returns true if \p FID did not come from a PCH/Module.
isLocalFileID(FileID FID)1496   bool isLocalFileID(FileID FID) const {
1497     return !isLoadedFileID(FID);
1498   }
1499 
1500   /// Gets the location of the immediate macro caller, one level up the stack
1501   /// toward the initial macro typed into the source.
getImmediateMacroCallerLoc(SourceLocation Loc)1502   SourceLocation getImmediateMacroCallerLoc(SourceLocation Loc) const {
1503     if (!Loc.isMacroID()) return Loc;
1504 
1505     // When we have the location of (part of) an expanded parameter, its
1506     // spelling location points to the argument as expanded in the macro call,
1507     // and therefore is used to locate the macro caller.
1508     if (isMacroArgExpansion(Loc))
1509       return getImmediateSpellingLoc(Loc);
1510 
1511     // Otherwise, the caller of the macro is located where this macro is
1512     // expanded (while the spelling is part of the macro definition).
1513     return getImmediateExpansionRange(Loc).first;
1514   }
1515 
1516 private:
1517   const llvm::MemoryBuffer *getFakeBufferForRecovery() const;
1518   const SrcMgr::ContentCache *getFakeContentCacheForRecovery() const;
1519 
1520   const SrcMgr::SLocEntry &loadSLocEntry(unsigned Index, bool *Invalid) const;
1521 
1522   /// \brief Get the entry with the given unwrapped FileID.
getSLocEntryByID(int ID)1523   const SrcMgr::SLocEntry &getSLocEntryByID(int ID) const {
1524     assert(ID != -1 && "Using FileID sentinel value");
1525     if (ID < 0)
1526       return getLoadedSLocEntryByID(ID);
1527     return getLocalSLocEntry(static_cast<unsigned>(ID));
1528   }
1529 
1530   const SrcMgr::SLocEntry &getLoadedSLocEntryByID(int ID,
1531                                                   bool *Invalid = 0) const {
1532     return getLoadedSLocEntry(static_cast<unsigned>(-ID - 2), Invalid);
1533   }
1534 
1535   /// Implements the common elements of storing an expansion info struct into
1536   /// the SLocEntry table and producing a source location that refers to it.
1537   SourceLocation createExpansionLocImpl(const SrcMgr::ExpansionInfo &Expansion,
1538                                         unsigned TokLength,
1539                                         int LoadedID = 0,
1540                                         unsigned LoadedOffset = 0);
1541 
1542   /// \brief Return true if the specified FileID contains the
1543   /// specified SourceLocation offset.  This is a very hot method.
isOffsetInFileID(FileID FID,unsigned SLocOffset)1544   inline bool isOffsetInFileID(FileID FID, unsigned SLocOffset) const {
1545     const SrcMgr::SLocEntry &Entry = getSLocEntry(FID);
1546     // If the entry is after the offset, it can't contain it.
1547     if (SLocOffset < Entry.getOffset()) return false;
1548 
1549     // If this is the very last entry then it does.
1550     if (FID.ID == -2)
1551       return true;
1552 
1553     // If it is the last local entry, then it does if the location is local.
1554     if (FID.ID+1 == static_cast<int>(LocalSLocEntryTable.size()))
1555       return SLocOffset < NextLocalOffset;
1556 
1557     // Otherwise, the entry after it has to not include it. This works for both
1558     // local and loaded entries.
1559     return SLocOffset < getSLocEntryByID(FID.ID+1).getOffset();
1560   }
1561 
1562   /// \brief Create a new fileID for the specified ContentCache and
1563   /// include position.
1564   ///
1565   /// This works regardless of whether the ContentCache corresponds to a
1566   /// file or some other input source.
1567   FileID createFileID(const SrcMgr::ContentCache* File,
1568                       SourceLocation IncludePos,
1569                       SrcMgr::CharacteristicKind DirCharacter,
1570                       int LoadedID, unsigned LoadedOffset);
1571 
1572   const SrcMgr::ContentCache *
1573     getOrCreateContentCache(const FileEntry *SourceFile,
1574                             bool isSystemFile = false);
1575 
1576   /// \brief Create a new ContentCache for the specified  memory buffer.
1577   const SrcMgr::ContentCache*
1578   createMemBufferContentCache(const llvm::MemoryBuffer *Buf);
1579 
1580   FileID getFileIDSlow(unsigned SLocOffset) const;
1581   FileID getFileIDLocal(unsigned SLocOffset) const;
1582   FileID getFileIDLoaded(unsigned SLocOffset) const;
1583 
1584   SourceLocation getExpansionLocSlowCase(SourceLocation Loc) const;
1585   SourceLocation getSpellingLocSlowCase(SourceLocation Loc) const;
1586   SourceLocation getFileLocSlowCase(SourceLocation Loc) const;
1587 
1588   std::pair<FileID, unsigned>
1589   getDecomposedExpansionLocSlowCase(const SrcMgr::SLocEntry *E) const;
1590   std::pair<FileID, unsigned>
1591   getDecomposedSpellingLocSlowCase(const SrcMgr::SLocEntry *E,
1592                                    unsigned Offset) const;
1593   void computeMacroArgsCache(MacroArgsMap *&MacroArgsCache, FileID FID) const;
1594   void associateFileChunkWithMacroArgExp(MacroArgsMap &MacroArgsCache,
1595                                          FileID FID,
1596                                          SourceLocation SpellLoc,
1597                                          SourceLocation ExpansionLoc,
1598                                          unsigned ExpansionLength) const;
1599   friend class ASTReader;
1600   friend class ASTWriter;
1601 };
1602 
1603 /// \brief Comparison function object.
1604 template<typename T>
1605 class BeforeThanCompare;
1606 
1607 /// \brief Compare two source locations.
1608 template<>
1609 class BeforeThanCompare<SourceLocation> {
1610   SourceManager &SM;
1611 
1612 public:
BeforeThanCompare(SourceManager & SM)1613   explicit BeforeThanCompare(SourceManager &SM) : SM(SM) { }
1614 
operator()1615   bool operator()(SourceLocation LHS, SourceLocation RHS) const {
1616     return SM.isBeforeInTranslationUnit(LHS, RHS);
1617   }
1618 };
1619 
1620 /// \brief Compare two non-overlapping source ranges.
1621 template<>
1622 class BeforeThanCompare<SourceRange> {
1623   SourceManager &SM;
1624 
1625 public:
BeforeThanCompare(SourceManager & SM)1626   explicit BeforeThanCompare(SourceManager &SM) : SM(SM) { }
1627 
operator()1628   bool operator()(SourceRange LHS, SourceRange RHS) {
1629     return SM.isBeforeInTranslationUnit(LHS.getBegin(), RHS.getBegin());
1630   }
1631 };
1632 
1633 }  // end namespace clang
1634 
1635 
1636 #endif
1637