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_BASIC_SOURCEMANAGER_H
36#define LLVM_CLANG_BASIC_SOURCEMANAGER_H
37
38#include "clang/Basic/FileManager.h"
39#include "clang/Basic/SourceLocation.h"
40#include "llvm/ADT/ArrayRef.h"
41#include "llvm/ADT/BitVector.h"
42#include "llvm/ADT/DenseMap.h"
43#include "llvm/ADT/DenseSet.h"
44#include "llvm/ADT/IntrusiveRefCntPtr.h"
45#include "llvm/ADT/PointerIntPair.h"
46#include "llvm/ADT/SmallVector.h"
47#include "llvm/ADT/StringRef.h"
48#include "llvm/Support/Allocator.h"
49#include "llvm/Support/Compiler.h"
50#include "llvm/Support/MemoryBuffer.h"
51#include <cassert>
52#include <cstddef>
53#include <map>
54#include <memory>
55#include <string>
56#include <utility>
57#include <vector>
58
59namespace clang {
60
61class ASTReader;
62class ASTWriter;
63class DiagnosticsEngine;
64class LineTableInfo;
65class SourceManager;
66
67/// \brief Public enums and private classes that are part of the
68/// SourceManager implementation.
69namespace SrcMgr {
70
71 /// \brief Indicates whether a file or directory holds normal user code,
72 /// system code, or system code which is implicitly 'extern "C"' in C++ mode.
73 ///
74 /// Entire directories can be tagged with this (this is maintained by
75 /// DirectoryLookup and friends) as can specific FileInfos when a \#pragma
76 /// system_header is seen or in various other cases.
77 ///
78 enum CharacteristicKind {
79 C_User, C_System, C_ExternCSystem, C_User_ModuleMap, C_System_ModuleMap
80 };
81
82 /// Determine whether a file / directory characteristic is for system code.
83 inline bool isSystem(CharacteristicKind CK) {
84 return CK != C_User && CK != C_User_ModuleMap;
85 }
86
87 /// Determine whether a file characteristic is for a module map.
88 inline bool isModuleMap(CharacteristicKind CK) {
89 return CK == C_User_ModuleMap || CK == C_System_ModuleMap;
90 }
91
92 /// \brief One instance of this struct is kept for every file loaded or used.
93 ///
94 /// This object owns the MemoryBuffer object.
95 class LLVM_ALIGNAS(8) ContentCache {
96 enum CCFlags {
97 /// \brief Whether the buffer is invalid.
98 InvalidFlag = 0x01,
99
100 /// \brief Whether the buffer should not be freed on destruction.
101 DoNotFreeFlag = 0x02
102 };
103
104 /// \brief The actual buffer containing the characters from the input
105 /// file.
106 ///
107 /// This is owned by the ContentCache object. The bits indicate
108 /// whether the buffer is invalid.
109 mutable llvm::PointerIntPair<llvm::MemoryBuffer *, 2> Buffer;
110
111 public:
112 /// \brief Reference to the file entry representing this ContentCache.
113 ///
114 /// This reference does not own the FileEntry object.
115 ///
116 /// It is possible for this to be NULL if the ContentCache encapsulates
117 /// an imaginary text buffer.
118 const FileEntry *OrigEntry;
119
120 /// \brief References the file which the contents were actually loaded from.
121 ///
122 /// Can be different from 'Entry' if we overridden the contents of one file
123 /// with the contents of another file.
124 const FileEntry *ContentsEntry;
125
126 /// \brief A bump pointer allocated array of offsets for each source line.
127 ///
128 /// This is lazily computed. This is owned by the SourceManager
129 /// BumpPointerAllocator object.
130 unsigned *SourceLineCache = nullptr;
131
132 /// \brief The number of lines in this ContentCache.
133 ///
134 /// This is only valid if SourceLineCache is non-null.
135 unsigned NumLines = 0;
136
137 /// \brief Indicates whether the buffer itself was provided to override
138 /// the actual file contents.
139 ///
140 /// When true, the original entry may be a virtual file that does not
141 /// exist.
142 unsigned BufferOverridden : 1;
143
144 /// \brief True if this content cache was initially created for a source
145 /// file considered as a system one.
146 unsigned IsSystemFile : 1;
147
148 /// \brief True if this file may be transient, that is, if it might not
149 /// exist at some later point in time when this content entry is used,
150 /// after serialization and deserialization.
151 unsigned IsTransient : 1;
152
153 ContentCache(const FileEntry *Ent = nullptr) : ContentCache(Ent, Ent) {}
154
155 ContentCache(const FileEntry *Ent, const FileEntry *contentEnt)
156 : Buffer(nullptr, false), OrigEntry(Ent), ContentsEntry(contentEnt),
157 BufferOverridden(false), IsSystemFile(false), IsTransient(false) {}
158
159 /// The copy ctor does not allow copies where source object has either
160 /// a non-NULL Buffer or SourceLineCache. Ownership of allocated memory
161 /// is not transferred, so this is a logical error.
162 ContentCache(const ContentCache &RHS)
163 : Buffer(nullptr, false), BufferOverridden(false), IsSystemFile(false),
164 IsTransient(false) {
165 OrigEntry = RHS.OrigEntry;
166 ContentsEntry = RHS.ContentsEntry;
167
168 assert(RHS.Buffer.getPointer() == nullptr &&
169 RHS.SourceLineCache == nullptr &&
170 "Passed ContentCache object cannot own a buffer.");
171
172 NumLines = RHS.NumLines;
173 }
174
175 ContentCache &operator=(const ContentCache& RHS) = delete;
176
177 ~ContentCache();
178
179 /// \brief Returns the memory buffer for the associated content.
180 ///
181 /// \param Diag Object through which diagnostics will be emitted if the
182 /// buffer cannot be retrieved.
183 ///
184 /// \param Loc If specified, is the location that invalid file diagnostics
185 /// will be emitted at.
186 ///
187 /// \param Invalid If non-NULL, will be set \c true if an error occurred.
188 llvm::MemoryBuffer *getBuffer(DiagnosticsEngine &Diag,
189 const SourceManager &SM,
190 SourceLocation Loc = SourceLocation(),
191 bool *Invalid = nullptr) const;
192
193 /// \brief Returns the size of the content encapsulated by this
194 /// ContentCache.
195 ///
196 /// This can be the size of the source file or the size of an
197 /// arbitrary scratch buffer. If the ContentCache encapsulates a source
198 /// file this size is retrieved from the file's FileEntry.
199 unsigned getSize() const;
200
201 /// \brief Returns the number of bytes actually mapped for this
202 /// ContentCache.
203 ///
204 /// This can be 0 if the MemBuffer was not actually expanded.
205 unsigned getSizeBytesMapped() const;
206
207 /// Returns the kind of memory used to back the memory buffer for
208 /// this content cache. This is used for performance analysis.
209 llvm::MemoryBuffer::BufferKind getMemoryBufferKind() const;
210
211 /// \brief Get the underlying buffer, returning NULL if the buffer is not
212 /// yet available.
213 llvm::MemoryBuffer *getRawBuffer() const { return Buffer.getPointer(); }
214
215 /// \brief Replace the existing buffer (which will be deleted)
216 /// with the given buffer.
217 void replaceBuffer(llvm::MemoryBuffer *B, bool DoNotFree = false);
218
219 /// \brief Determine whether the buffer itself is invalid.
220 bool isBufferInvalid() const {
221 return Buffer.getInt() & InvalidFlag;
222 }
223
224 /// \brief Determine whether the buffer should be freed.
225 bool shouldFreeBuffer() const {
226 return (Buffer.getInt() & DoNotFreeFlag) == 0;
227 }
228 };
229
230 // Assert that the \c ContentCache objects will always be 8-byte aligned so
231 // that we can pack 3 bits of integer into pointers to such objects.
232 static_assert(alignof(ContentCache) >= 8,
233 "ContentCache must be 8-byte aligned.");
234
235 /// \brief Information about a FileID, basically just the logical file
236 /// that it represents and include stack information.
237 ///
238 /// Each FileInfo has include stack information, indicating where it came
239 /// from. This information encodes the \#include chain that a token was
240 /// expanded from. The main include file has an invalid IncludeLoc.
241 ///
242 /// FileInfos contain a "ContentCache *", with the contents of the file.
243 ///
244 class FileInfo {
245 friend class clang::SourceManager;
246 friend class clang::ASTWriter;
247 friend class clang::ASTReader;
248
249 /// \brief The location of the \#include that brought in this file.
250 ///
251 /// This is an invalid SLOC for the main file (top of the \#include chain).
252 unsigned IncludeLoc; // Really a SourceLocation
253
254 /// \brief Number of FileIDs (files and macros) that were created during
255 /// preprocessing of this \#include, including this SLocEntry.
256 ///
257 /// Zero means the preprocessor didn't provide such info for this SLocEntry.
258 unsigned NumCreatedFIDs : 31;
259
260 /// \brief Whether this FileInfo has any \#line directives.
261 unsigned HasLineDirectives : 1;
262
263 /// \brief The content cache and the characteristic of the file.
264 llvm::PointerIntPair<const ContentCache*, 3, CharacteristicKind>
265 ContentAndKind;
266
267 public:
268 /// \brief Return a FileInfo object.
269 static FileInfo get(SourceLocation IL, const ContentCache *Con,
270 CharacteristicKind FileCharacter) {
271 FileInfo X;
272 X.IncludeLoc = IL.getRawEncoding();
273 X.NumCreatedFIDs = 0;
274 X.HasLineDirectives = false;
275 X.ContentAndKind.setPointer(Con);
276 X.ContentAndKind.setInt(FileCharacter);
277 return X;
278 }
279
280 SourceLocation getIncludeLoc() const {
281 return SourceLocation::getFromRawEncoding(IncludeLoc);
282 }
283
284 const ContentCache *getContentCache() const {
285 return ContentAndKind.getPointer();
286 }
287
288 /// \brief Return whether this is a system header or not.
289 CharacteristicKind getFileCharacteristic() const {
290 return ContentAndKind.getInt();
291 }
292
293 /// \brief Return true if this FileID has \#line directives in it.
294 bool hasLineDirectives() const { return HasLineDirectives; }
295
296 /// \brief Set the flag that indicates that this FileID has
297 /// line table entries associated with it.
298 void setHasLineDirectives() {
299 HasLineDirectives = true;
300 }
301 };
302
303 /// \brief Each ExpansionInfo encodes the expansion location - where
304 /// the token was ultimately expanded, and the SpellingLoc - where the actual
305 /// character data for the token came from.
306 class ExpansionInfo {
307 // Really these are all SourceLocations.
308
309 /// \brief Where the spelling for the token can be found.
310 unsigned SpellingLoc;
311
312 /// In a macro expansion, ExpansionLocStart and ExpansionLocEnd
313 /// indicate the start and end of the expansion. In object-like macros,
314 /// they will be the same. In a function-like macro expansion, the start
315 /// will be the identifier and the end will be the ')'. Finally, in
316 /// macro-argument instantiations, the end will be 'SourceLocation()', an
317 /// invalid location.
318 unsigned ExpansionLocStart, ExpansionLocEnd;
319
320 public:
321 SourceLocation getSpellingLoc() const {
322 return SourceLocation::getFromRawEncoding(SpellingLoc);
323 }
324
325 SourceLocation getExpansionLocStart() const {
326 return SourceLocation::getFromRawEncoding(ExpansionLocStart);
327 }
328
329 SourceLocation getExpansionLocEnd() const {
330 SourceLocation EndLoc =
331 SourceLocation::getFromRawEncoding(ExpansionLocEnd);
332 return EndLoc.isInvalid() ? getExpansionLocStart() : EndLoc;
333 }
334
335 std::pair<SourceLocation,SourceLocation> getExpansionLocRange() const {
336 return std::make_pair(getExpansionLocStart(), getExpansionLocEnd());
337 }
338
339 bool isMacroArgExpansion() const {
340 // Note that this needs to return false for default constructed objects.
341 return getExpansionLocStart().isValid() &&
342 SourceLocation::getFromRawEncoding(ExpansionLocEnd).isInvalid();
343 }
344
345 bool isMacroBodyExpansion() const {
346 return getExpansionLocStart().isValid() &&
347 SourceLocation::getFromRawEncoding(ExpansionLocEnd).isValid();
348 }
349
350 bool isFunctionMacroExpansion() const {
351 return getExpansionLocStart().isValid() &&
352 getExpansionLocStart() != getExpansionLocEnd();
353 }
354
355 /// \brief Return a ExpansionInfo for an expansion.
356 ///
357 /// Start and End specify the expansion range (where the macro is
358 /// expanded), and SpellingLoc specifies the spelling location (where
359 /// the characters from the token come from). All three can refer to
360 /// normal File SLocs or expansion locations.
361 static ExpansionInfo create(SourceLocation SpellingLoc,
362 SourceLocation Start, SourceLocation End) {
363 ExpansionInfo X;
364 X.SpellingLoc = SpellingLoc.getRawEncoding();
365 X.ExpansionLocStart = Start.getRawEncoding();
366 X.ExpansionLocEnd = End.getRawEncoding();
367 return X;
368 }
369
370 /// \brief Return a special ExpansionInfo for the expansion of
371 /// a macro argument into a function-like macro's body.
372 ///
373 /// ExpansionLoc specifies the expansion location (where the macro is
374 /// expanded). This doesn't need to be a range because a macro is always
375 /// expanded at a macro parameter reference, and macro parameters are
376 /// always exactly one token. SpellingLoc specifies the spelling location
377 /// (where the characters from the token come from). ExpansionLoc and
378 /// SpellingLoc can both refer to normal File SLocs or expansion locations.
379 ///
380 /// Given the code:
381 /// \code
382 /// #define F(x) f(x)
383 /// F(42);
384 /// \endcode
385 ///
386 /// When expanding '\c F(42)', the '\c x' would call this with an
387 /// SpellingLoc pointing at '\c 42' and an ExpansionLoc pointing at its
388 /// location in the definition of '\c F'.
389 static ExpansionInfo createForMacroArg(SourceLocation SpellingLoc,
390 SourceLocation ExpansionLoc) {
391 // We store an intentionally invalid source location for the end of the
392 // expansion range to mark that this is a macro argument ion rather than
393 // a normal one.
394 return create(SpellingLoc, ExpansionLoc, SourceLocation());
395 }
396 };
397
398 /// \brief This is a discriminated union of FileInfo and ExpansionInfo.
399 ///
400 /// SourceManager keeps an array of these objects, and they are uniquely
401 /// identified by the FileID datatype.
402 class SLocEntry {
403 unsigned Offset : 31;
404 unsigned IsExpansion : 1;
405 union {
406 FileInfo File;
407 ExpansionInfo Expansion;
408 };
409
410 public:
411 SLocEntry() : Offset(), IsExpansion(), File() {}
412
413 unsigned getOffset() const { return Offset; }
414
415 bool isExpansion() const { return IsExpansion; }
416 bool isFile() const { return !isExpansion(); }
417
418 const FileInfo &getFile() const {
419 assert(isFile() && "Not a file SLocEntry!");
420 return File;
421 }
422
423 const ExpansionInfo &getExpansion() const {
424 assert(isExpansion() && "Not a macro expansion SLocEntry!");
425 return Expansion;
426 }
427
428 static SLocEntry get(unsigned Offset, const FileInfo &FI) {
429 assert(!(Offset & (1 << 31)) && "Offset is too large");
430 SLocEntry E;
431 E.Offset = Offset;
432 E.IsExpansion = false;
433 E.File = FI;
434 return E;
435 }
436
437 static SLocEntry get(unsigned Offset, const ExpansionInfo &Expansion) {
438 assert(!(Offset & (1 << 31)) && "Offset is too large");
439 SLocEntry E;
440 E.Offset = Offset;
441 E.IsExpansion = true;
442 E.Expansion = Expansion;
443 return E;
444 }
445 };
446
447} // namespace SrcMgr
448
449/// \brief External source of source location entries.
450class ExternalSLocEntrySource {
451public:
452 virtual ~ExternalSLocEntrySource();
453
454 /// \brief Read the source location entry with index ID, which will always be
455 /// less than -1.
456 ///
457 /// \returns true if an error occurred that prevented the source-location
458 /// entry from being loaded.
459 virtual bool ReadSLocEntry(int ID) = 0;
460
461 /// \brief Retrieve the module import location and name for the given ID, if
462 /// in fact it was loaded from a module (rather than, say, a precompiled
463 /// header).
464 virtual std::pair<SourceLocation, StringRef> getModuleImportLoc(int ID) = 0;
465};
466
467/// \brief Holds the cache used by isBeforeInTranslationUnit.
468///
469/// The cache structure is complex enough to be worth breaking out of
470/// SourceManager.
471class InBeforeInTUCacheEntry {
472 /// \brief The FileID's of the cached query.
473 ///
474 /// If these match up with a subsequent query, the result can be reused.
475 FileID LQueryFID, RQueryFID;
476
477 /// \brief True if LQueryFID was created before RQueryFID.
478 ///
479 /// This is used to compare macro expansion locations.
480 bool IsLQFIDBeforeRQFID;
481
482 /// \brief The file found in common between the two \#include traces, i.e.,
483 /// the nearest common ancestor of the \#include tree.
484 FileID CommonFID;
485
486 /// \brief The offset of the previous query in CommonFID.
487 ///
488 /// Usually, this represents the location of the \#include for QueryFID, but
489 /// if LQueryFID is a parent of RQueryFID (or vice versa) then these can be a
490 /// random token in the parent.
491 unsigned LCommonOffset, RCommonOffset;
492
493public:
494 /// \brief Return true if the currently cached values match up with
495 /// the specified LHS/RHS query.
496 ///
497 /// If not, we can't use the cache.
498 bool isCacheValid(FileID LHS, FileID RHS) const {
499 return LQueryFID == LHS && RQueryFID == RHS;
500 }
501
502 /// \brief If the cache is valid, compute the result given the
503 /// specified offsets in the LHS/RHS FileID's.
504 bool getCachedResult(unsigned LOffset, unsigned ROffset) const {
505 // If one of the query files is the common file, use the offset. Otherwise,
506 // use the #include loc in the common file.
507 if (LQueryFID != CommonFID) LOffset = LCommonOffset;
508 if (RQueryFID != CommonFID) ROffset = RCommonOffset;
509
510 // It is common for multiple macro expansions to be "included" from the same
511 // location (expansion location), in which case use the order of the FileIDs
512 // to determine which came first. This will also take care the case where
513 // one of the locations points at the inclusion/expansion point of the other
514 // in which case its FileID will come before the other.
515 if (LOffset == ROffset)
516 return IsLQFIDBeforeRQFID;
517
518 return LOffset < ROffset;
519 }
520
521 /// \brief Set up a new query.
522 void setQueryFIDs(FileID LHS, FileID RHS, bool isLFIDBeforeRFID) {
523 assert(LHS != RHS);
524 LQueryFID = LHS;
525 RQueryFID = RHS;
526 IsLQFIDBeforeRQFID = isLFIDBeforeRFID;
527 }
528
529 void clear() {
530 LQueryFID = RQueryFID = FileID();
531 IsLQFIDBeforeRQFID = false;
532 }
533
534 void setCommonLoc(FileID commonFID, unsigned lCommonOffset,
535 unsigned rCommonOffset) {
536 CommonFID = commonFID;
537 LCommonOffset = lCommonOffset;
538 RCommonOffset = rCommonOffset;
539 }
540};
541
542/// \brief The stack used when building modules on demand, which is used
543/// to provide a link between the source managers of the different compiler
544/// instances.
545using ModuleBuildStack = ArrayRef<std::pair<std::string, FullSourceLoc>>;
546
547/// \brief This class handles loading and caching of source files into memory.
548///
549/// This object owns the MemoryBuffer objects for all of the loaded
550/// files and assigns unique FileID's for each unique \#include chain.
551///
552/// The SourceManager can be queried for information about SourceLocation
553/// objects, turning them into either spelling or expansion locations. Spelling
554/// locations represent where the bytes corresponding to a token came from and
555/// expansion locations represent where the location is in the user's view. In
556/// the case of a macro expansion, for example, the spelling location indicates
557/// where the expanded token came from and the expansion location specifies
558/// where it was expanded.
559class SourceManager : public RefCountedBase<SourceManager> {
560 /// \brief DiagnosticsEngine object.
561 DiagnosticsEngine &Diag;
562
563 FileManager &FileMgr;
564
565 mutable llvm::BumpPtrAllocator ContentCacheAlloc;
566
567 /// \brief Memoized information about all of the files tracked by this
568 /// SourceManager.
569 ///
570 /// This map allows us to merge ContentCache entries based
571 /// on their FileEntry*. All ContentCache objects will thus have unique,
572 /// non-null, FileEntry pointers.
573 llvm::DenseMap<const FileEntry*, SrcMgr::ContentCache*> FileInfos;
574
575 /// \brief True if the ContentCache for files that are overridden by other
576 /// files, should report the original file name. Defaults to true.
577 bool OverridenFilesKeepOriginalName = true;
578
579 /// \brief True if non-system source files should be treated as volatile
580 /// (likely to change while trying to use them). Defaults to false.
581 bool UserFilesAreVolatile;
582
583 /// \brief True if all files read during this compilation should be treated
584 /// as transient (may not be present in later compilations using a module
585 /// file created from this compilation). Defaults to false.
586 bool FilesAreTransient = false;
587
588 struct OverriddenFilesInfoTy {
589 /// \brief Files that have been overridden with the contents from another
590 /// file.
591 llvm::DenseMap<const FileEntry *, const FileEntry *> OverriddenFiles;
592
593 /// \brief Files that were overridden with a memory buffer.
594 llvm::DenseSet<const FileEntry *> OverriddenFilesWithBuffer;
595 };
596
597 /// \brief Lazily create the object keeping overridden files info, since
598 /// it is uncommonly used.
599 std::unique_ptr<OverriddenFilesInfoTy> OverriddenFilesInfo;
600
601 OverriddenFilesInfoTy &getOverriddenFilesInfo() {
602 if (!OverriddenFilesInfo)
603 OverriddenFilesInfo.reset(new OverriddenFilesInfoTy);
604 return *OverriddenFilesInfo;
605 }
606
607 /// \brief Information about various memory buffers that we have read in.
608 ///
609 /// All FileEntry* within the stored ContentCache objects are NULL,
610 /// as they do not refer to a file.
611 std::vector<SrcMgr::ContentCache*> MemBufferInfos;
612
613 /// \brief The table of SLocEntries that are local to this module.
614 ///
615 /// Positive FileIDs are indexes into this table. Entry 0 indicates an invalid
616 /// expansion.
617 SmallVector<SrcMgr::SLocEntry, 0> LocalSLocEntryTable;
618
619 /// \brief The table of SLocEntries that are loaded from other modules.
620 ///
621 /// Negative FileIDs are indexes into this table. To get from ID to an index,
622 /// use (-ID - 2).
623 mutable SmallVector<SrcMgr::SLocEntry, 0> LoadedSLocEntryTable;
624
625 /// \brief The starting offset of the next local SLocEntry.
626 ///
627 /// This is LocalSLocEntryTable.back().Offset + the size of that entry.
628 unsigned NextLocalOffset;
629
630 /// \brief The starting offset of the latest batch of loaded SLocEntries.
631 ///
632 /// This is LoadedSLocEntryTable.back().Offset, except that that entry might
633 /// not have been loaded, so that value would be unknown.
634 unsigned CurrentLoadedOffset;
635
636 /// \brief The highest possible offset is 2^31-1, so CurrentLoadedOffset
637 /// starts at 2^31.
638 static const unsigned MaxLoadedOffset = 1U << 31U;
639
640 /// \brief A bitmap that indicates whether the entries of LoadedSLocEntryTable
641 /// have already been loaded from the external source.
642 ///
643 /// Same indexing as LoadedSLocEntryTable.
644 llvm::BitVector SLocEntryLoaded;
645
646 /// \brief An external source for source location entries.
647 ExternalSLocEntrySource *ExternalSLocEntries = nullptr;
648
649 /// \brief A one-entry cache to speed up getFileID.
650 ///
651 /// LastFileIDLookup records the last FileID looked up or created, because it
652 /// is very common to look up many tokens from the same file.
653 mutable FileID LastFileIDLookup;
654
655 /// \brief Holds information for \#line directives.
656 ///
657 /// This is referenced by indices from SLocEntryTable.
658 LineTableInfo *LineTable = nullptr;
659
660 /// \brief These ivars serve as a cache used in the getLineNumber
661 /// method which is used to speedup getLineNumber calls to nearby locations.
662 mutable FileID LastLineNoFileIDQuery;
663 mutable SrcMgr::ContentCache *LastLineNoContentCache;
664 mutable unsigned LastLineNoFilePos;
665 mutable unsigned LastLineNoResult;
666
667 /// \brief The file ID for the main source file of the translation unit.
668 FileID MainFileID;
669
670 /// \brief The file ID for the precompiled preamble there is one.
671 FileID PreambleFileID;
672
673 // Statistics for -print-stats.
674 mutable unsigned NumLinearScans = 0;
675 mutable unsigned NumBinaryProbes = 0;
676
677 /// \brief Associates a FileID with its "included/expanded in" decomposed
678 /// location.
679 ///
680 /// Used to cache results from and speed-up \c getDecomposedIncludedLoc
681 /// function.
682 mutable llvm::DenseMap<FileID, std::pair<FileID, unsigned>> IncludedLocMap;
683
684 /// The key value into the IsBeforeInTUCache table.
685 using IsBeforeInTUCacheKey = std::pair<FileID, FileID>;
686
687 /// The IsBeforeInTranslationUnitCache is a mapping from FileID pairs
688 /// to cache results.
689 using InBeforeInTUCache =
690 llvm::DenseMap<IsBeforeInTUCacheKey, InBeforeInTUCacheEntry>;
691
692 /// Cache results for the isBeforeInTranslationUnit method.
693 mutable InBeforeInTUCache IBTUCache;
694 mutable InBeforeInTUCacheEntry IBTUCacheOverflow;
695
696 /// Return the cache entry for comparing the given file IDs
697 /// for isBeforeInTranslationUnit.
698 InBeforeInTUCacheEntry &getInBeforeInTUCache(FileID LFID, FileID RFID) const;
699
700 // Cache for the "fake" buffer used for error-recovery purposes.
701 mutable std::unique_ptr<llvm::MemoryBuffer> FakeBufferForRecovery;
702
703 mutable std::unique_ptr<SrcMgr::ContentCache> FakeContentCacheForRecovery;
704
705 /// \brief Lazily computed map of macro argument chunks to their expanded
706 /// source location.
707 using MacroArgsMap = std::map<unsigned, SourceLocation>;
708
709 mutable llvm::DenseMap<FileID, std::unique_ptr<MacroArgsMap>>
710 MacroArgsCacheMap;
711
712 /// \brief The stack of modules being built, which is used to detect
713 /// cycles in the module dependency graph as modules are being built, as
714 /// well as to describe why we're rebuilding a particular module.
715 ///
716 /// There is no way to set this value from the command line. If we ever need
717 /// to do so (e.g., if on-demand module construction moves out-of-process),
718 /// we can add a cc1-level option to do so.
719 SmallVector<std::pair<std::string, FullSourceLoc>, 2> StoredModuleBuildStack;
720
721public:
722 SourceManager(DiagnosticsEngine &Diag, FileManager &FileMgr,
723 bool UserFilesAreVolatile = false);
724 explicit SourceManager(const SourceManager &) = delete;
725 SourceManager &operator=(const SourceManager &) = delete;
726 ~SourceManager();
727
728 void clearIDTables();
729
730 /// Initialize this source manager suitably to replay the compilation
731 /// described by \p Old. Requires that \p Old outlive \p *this.
732 void initializeForReplay(const SourceManager &Old);
733
734 DiagnosticsEngine &getDiagnostics() const { return Diag; }
735
736 FileManager &getFileManager() const { return FileMgr; }
737
738 /// \brief Set true if the SourceManager should report the original file name
739 /// for contents of files that were overridden by other files. Defaults to
740 /// true.
741 void setOverridenFilesKeepOriginalName(bool value) {
742 OverridenFilesKeepOriginalName = value;
743 }
744
745 /// \brief True if non-system source files should be treated as volatile
746 /// (likely to change while trying to use them).
747 bool userFilesAreVolatile() const { return UserFilesAreVolatile; }
748
749 /// \brief Retrieve the module build stack.
750 ModuleBuildStack getModuleBuildStack() const {
751 return StoredModuleBuildStack;
752 }
753
754 /// \brief Set the module build stack.
755 void setModuleBuildStack(ModuleBuildStack stack) {
756 StoredModuleBuildStack.clear();
757 StoredModuleBuildStack.append(stack.begin(), stack.end());
758 }
759
760 /// \brief Push an entry to the module build stack.
761 void pushModuleBuildStack(StringRef moduleName, FullSourceLoc importLoc) {
762 StoredModuleBuildStack.push_back(std::make_pair(moduleName.str(),importLoc));
763 }
764
765 //===--------------------------------------------------------------------===//
766 // MainFileID creation and querying methods.
767 //===--------------------------------------------------------------------===//
768
769 /// \brief Returns the FileID of the main source file.
770 FileID getMainFileID() const { return MainFileID; }
771
772 /// \brief Set the file ID for the main source file.
773 void setMainFileID(FileID FID) {
774 MainFileID = FID;
775 }
776
777 /// \brief Set the file ID for the precompiled preamble.
778 void setPreambleFileID(FileID Preamble) {
779 assert(PreambleFileID.isInvalid() && "PreambleFileID already set!");
780 PreambleFileID = Preamble;
781 }
782
783 /// \brief Get the file ID for the precompiled preamble if there is one.
784 FileID getPreambleFileID() const { return PreambleFileID; }
785
786 //===--------------------------------------------------------------------===//
787 // Methods to create new FileID's and macro expansions.
788 //===--------------------------------------------------------------------===//
789
790 /// \brief Create a new FileID that represents the specified file
791 /// being \#included from the specified IncludePosition.
792 ///
793 /// This translates NULL into standard input.
794 FileID createFileID(const FileEntry *SourceFile, SourceLocation IncludePos,
795 SrcMgr::CharacteristicKind FileCharacter,
796 int LoadedID = 0, unsigned LoadedOffset = 0) {
797 const SrcMgr::ContentCache *IR =
798 getOrCreateContentCache(SourceFile, isSystem(FileCharacter));
799 assert(IR && "getOrCreateContentCache() cannot return NULL");
800 return createFileID(IR, IncludePos, FileCharacter, LoadedID, LoadedOffset);
801 }
802
803 /// \brief Create a new FileID that represents the specified memory buffer.
804 ///
805 /// This does no caching of the buffer and takes ownership of the
806 /// MemoryBuffer, so only pass a MemoryBuffer to this once.
807 FileID createFileID(std::unique_ptr<llvm::MemoryBuffer> Buffer,
808 SrcMgr::CharacteristicKind FileCharacter = SrcMgr::C_User,
809 int LoadedID = 0, unsigned LoadedOffset = 0,
810 SourceLocation IncludeLoc = SourceLocation()) {
811 return createFileID(
812 createMemBufferContentCache(Buffer.release(), /*DoNotFree*/ false),
813 IncludeLoc, FileCharacter, LoadedID, LoadedOffset);
814 }
815
816 enum UnownedTag { Unowned };
817
818 /// \brief Create a new FileID that represents the specified memory buffer.
819 ///
820 /// This does no caching of the buffer and takes ownership of the
821 /// MemoryBuffer, so only pass a MemoryBuffer to this once.
822 FileID createFileID(UnownedTag, llvm::MemoryBuffer *Buffer,
823 SrcMgr::CharacteristicKind FileCharacter = SrcMgr::C_User,
824 int LoadedID = 0, unsigned LoadedOffset = 0,
825 SourceLocation IncludeLoc = SourceLocation()) {
826 return createFileID(createMemBufferContentCache(Buffer, /*DoNotFree*/true),
827 IncludeLoc, FileCharacter, LoadedID, LoadedOffset);
828 }
829
830 /// \brief Get the FileID for \p SourceFile if it exists. Otherwise, create a
831 /// new FileID for the \p SourceFile.
832 FileID getOrCreateFileID(const FileEntry *SourceFile,
833 SrcMgr::CharacteristicKind FileCharacter) {
834 FileID ID = translateFile(SourceFile);
835 return ID.isValid() ? ID : createFileID(SourceFile, SourceLocation(),
836 FileCharacter);
837 }
838
839 /// \brief Return a new SourceLocation that encodes the
840 /// fact that a token from SpellingLoc should actually be referenced from
841 /// ExpansionLoc, and that it represents the expansion of a macro argument
842 /// into the function-like macro body.
843 SourceLocation createMacroArgExpansionLoc(SourceLocation Loc,
844 SourceLocation ExpansionLoc,
845 unsigned TokLength);
846
847 /// \brief Return a new SourceLocation that encodes the fact
848 /// that a token from SpellingLoc should actually be referenced from
849 /// ExpansionLoc.
850 SourceLocation createExpansionLoc(SourceLocation Loc,
851 SourceLocation ExpansionLocStart,
852 SourceLocation ExpansionLocEnd,
853 unsigned TokLength,
854 int LoadedID = 0,
855 unsigned LoadedOffset = 0);
856
857 /// \brief Retrieve the memory buffer associated with the given file.
858 ///
859 /// \param Invalid If non-NULL, will be set \c true if an error
860 /// occurs while retrieving the memory buffer.
861 llvm::MemoryBuffer *getMemoryBufferForFile(const FileEntry *File,
862 bool *Invalid = nullptr);
863
864 /// \brief Override the contents of the given source file by providing an
865 /// already-allocated buffer.
866 ///
867 /// \param SourceFile the source file whose contents will be overridden.
868 ///
869 /// \param Buffer the memory buffer whose contents will be used as the
870 /// data in the given source file.
871 ///
872 /// \param DoNotFree If true, then the buffer will not be freed when the
873 /// source manager is destroyed.
874 void overrideFileContents(const FileEntry *SourceFile,
875 llvm::MemoryBuffer *Buffer, bool DoNotFree);
876 void overrideFileContents(const FileEntry *SourceFile,
877 std::unique_ptr<llvm::MemoryBuffer> Buffer) {
878 overrideFileContents(SourceFile, Buffer.release(), /*DoNotFree*/ false);
879 }
880
881 /// \brief Override the given source file with another one.
882 ///
883 /// \param SourceFile the source file which will be overridden.
884 ///
885 /// \param NewFile the file whose contents will be used as the
886 /// data instead of the contents of the given source file.
887 void overrideFileContents(const FileEntry *SourceFile,
888 const FileEntry *NewFile);
889
890 /// \brief Returns true if the file contents have been overridden.
891 bool isFileOverridden(const FileEntry *File) const {
892 if (OverriddenFilesInfo) {
893 if (OverriddenFilesInfo->OverriddenFilesWithBuffer.count(File))
894 return true;
895 if (OverriddenFilesInfo->OverriddenFiles.find(File) !=
896 OverriddenFilesInfo->OverriddenFiles.end())
897 return true;
898 }
899 return false;
900 }
901
902 /// \brief Disable overridding the contents of a file, previously enabled
903 /// with #overrideFileContents.
904 ///
905 /// This should be called before parsing has begun.
906 void disableFileContentsOverride(const FileEntry *File);
907
908 /// \brief Specify that a file is transient.
909 void setFileIsTransient(const FileEntry *SourceFile);
910
911 /// \brief Specify that all files that are read during this compilation are
912 /// transient.
913 void setAllFilesAreTransient(bool Transient) {
914 FilesAreTransient = Transient;
915 }
916
917 //===--------------------------------------------------------------------===//
918 // FileID manipulation methods.
919 //===--------------------------------------------------------------------===//
920
921 /// \brief Return the buffer for the specified FileID.
922 ///
923 /// If there is an error opening this buffer the first time, this
924 /// manufactures a temporary buffer and returns a non-empty error string.
925 llvm::MemoryBuffer *getBuffer(FileID FID, SourceLocation Loc,
926 bool *Invalid = nullptr) const {
927 bool MyInvalid = false;
928 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &MyInvalid);
929 if (MyInvalid || !Entry.isFile()) {
930 if (Invalid)
931 *Invalid = true;
932
933 return getFakeBufferForRecovery();
934 }
935
936 return Entry.getFile().getContentCache()->getBuffer(Diag, *this, Loc,
937 Invalid);
938 }
939
940 llvm::MemoryBuffer *getBuffer(FileID FID, bool *Invalid = nullptr) const {
941 bool MyInvalid = false;
942 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &MyInvalid);
943 if (MyInvalid || !Entry.isFile()) {
944 if (Invalid)
945 *Invalid = true;
946
947 return getFakeBufferForRecovery();
948 }
949
950 return Entry.getFile().getContentCache()->getBuffer(Diag, *this,
951 SourceLocation(),
952 Invalid);
953 }
954
955 /// \brief Returns the FileEntry record for the provided FileID.
956 const FileEntry *getFileEntryForID(FileID FID) const {
957 bool MyInvalid = false;
958 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &MyInvalid);
959 if (MyInvalid || !Entry.isFile())
960 return nullptr;
961
962 const SrcMgr::ContentCache *Content = Entry.getFile().getContentCache();
963 if (!Content)
964 return nullptr;
965 return Content->OrigEntry;
966 }
967
968 /// \brief Returns the FileEntry record for the provided SLocEntry.
969 const FileEntry *getFileEntryForSLocEntry(const SrcMgr::SLocEntry &sloc) const
970 {
971 const SrcMgr::ContentCache *Content = sloc.getFile().getContentCache();
972 if (!Content)
973 return nullptr;
974 return Content->OrigEntry;
975 }
976
977 /// \brief Return a StringRef to the source buffer data for the
978 /// specified FileID.
979 ///
980 /// \param FID The file ID whose contents will be returned.
981 /// \param Invalid If non-NULL, will be set true if an error occurred.
982 StringRef getBufferData(FileID FID, bool *Invalid = nullptr) const;
983
984 /// \brief Get the number of FileIDs (files and macros) that were created
985 /// during preprocessing of \p FID, including it.
986 unsigned getNumCreatedFIDsForFileID(FileID FID) const {
987 bool Invalid = false;
988 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
989 if (Invalid || !Entry.isFile())
990 return 0;
991
992 return Entry.getFile().NumCreatedFIDs;
993 }
994
995 /// \brief Set the number of FileIDs (files and macros) that were created
996 /// during preprocessing of \p FID, including it.
997 void setNumCreatedFIDsForFileID(FileID FID, unsigned NumFIDs) const {
998 bool Invalid = false;
999 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
1000 if (Invalid || !Entry.isFile())
1001 return;
1002
1003 assert(Entry.getFile().NumCreatedFIDs == 0 && "Already set!");
1004 const_cast<SrcMgr::FileInfo &>(Entry.getFile()).NumCreatedFIDs = NumFIDs;
1005 }
1006
1007 //===--------------------------------------------------------------------===//
1008 // SourceLocation manipulation methods.
1009 //===--------------------------------------------------------------------===//
1010
1011 /// \brief Return the FileID for a SourceLocation.
1012 ///
1013 /// This is a very hot method that is used for all SourceManager queries
1014 /// that start with a SourceLocation object. It is responsible for finding
1015 /// the entry in SLocEntryTable which contains the specified location.
1016 ///
1017 FileID getFileID(SourceLocation SpellingLoc) const {
1018 unsigned SLocOffset = SpellingLoc.getOffset();
1019
1020 // If our one-entry cache covers this offset, just return it.
1021 if (isOffsetInFileID(LastFileIDLookup, SLocOffset))
1022 return LastFileIDLookup;
1023
1024 return getFileIDSlow(SLocOffset);
1025 }
1026
1027 /// \brief Return the filename of the file containing a SourceLocation.
1028 StringRef getFilename(SourceLocation SpellingLoc) const {
1029 if (const FileEntry *F = getFileEntryForID(getFileID(SpellingLoc)))
1030 return F->getName();
1031 return StringRef();
1032 }
1033
1034 /// \brief Return the source location corresponding to the first byte of
1035 /// the specified file.
1036 SourceLocation getLocForStartOfFile(FileID FID) const {
1037 bool Invalid = false;
1038 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
1039 if (Invalid || !Entry.isFile())
1040 return SourceLocation();
1041
1042 unsigned FileOffset = Entry.getOffset();
1043 return SourceLocation::getFileLoc(FileOffset);
1044 }
1045
1046 /// \brief Return the source location corresponding to the last byte of the
1047 /// specified file.
1048 SourceLocation getLocForEndOfFile(FileID FID) const {
1049 bool Invalid = false;
1050 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
1051 if (Invalid || !Entry.isFile())
1052 return SourceLocation();
1053
1054 unsigned FileOffset = Entry.getOffset();
1055 return SourceLocation::getFileLoc(FileOffset + getFileIDSize(FID));
1056 }
1057
1058 /// \brief Returns the include location if \p FID is a \#include'd file
1059 /// otherwise it returns an invalid location.
1060 SourceLocation getIncludeLoc(FileID FID) const {
1061 bool Invalid = false;
1062 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, &Invalid);
1063 if (Invalid || !Entry.isFile())
1064 return SourceLocation();
1065
1066 return Entry.getFile().getIncludeLoc();
1067 }
1068
1069 // \brief Returns the import location if the given source location is
1070 // located within a module, or an invalid location if the source location
1071 // is within the current translation unit.
1072 std::pair<SourceLocation, StringRef>
1073 getModuleImportLoc(SourceLocation Loc) const {
1074 FileID FID = getFileID(Loc);
1075
1076 // Positive file IDs are in the current translation unit, and -1 is a
1077 // placeholder.
1078 if (FID.ID >= -1)
1079 return std::make_pair(SourceLocation(), "");
1080
1081 return ExternalSLocEntries->getModuleImportLoc(FID.ID);
1082 }
1083
1084 /// \brief Given a SourceLocation object \p Loc, return the expansion
1085 /// location referenced by the ID.
1086 SourceLocation getExpansionLoc(SourceLocation Loc) const {
1087 // Handle the non-mapped case inline, defer to out of line code to handle
1088 // expansions.
1089 if (Loc.isFileID()) return Loc;
1090 return getExpansionLocSlowCase(Loc);
1091 }
1092
1093 /// \brief Given \p Loc, if it is a macro location return the expansion
1094 /// location or the spelling location, depending on if it comes from a
1095 /// macro argument or not.
1096 SourceLocation getFileLoc(SourceLocation Loc) const {
1097 if (Loc.isFileID()) return Loc;
1098 return getFileLocSlowCase(Loc);
1099 }
1100
1101 /// \brief Return the start/end of the expansion information for an
1102 /// expansion location.
1103 ///
1104 /// \pre \p Loc is required to be an expansion location.
1105 std::pair<SourceLocation,SourceLocation>
1106 getImmediateExpansionRange(SourceLocation Loc) const;
1107
1108 /// \brief Given a SourceLocation object, return the range of
1109 /// tokens covered by the expansion in the ultimate file.
1110 std::pair<SourceLocation,SourceLocation>
1111 getExpansionRange(SourceLocation Loc) const;
1112
1113 /// \brief Given a SourceRange object, return the range of
1114 /// tokens covered by the expansion in the ultimate file.
1115 SourceRange getExpansionRange(SourceRange Range) const {
1116 return SourceRange(getExpansionRange(Range.getBegin()).first,
1117 getExpansionRange(Range.getEnd()).second);
1118 }
1119
1120 /// \brief Given a SourceLocation object, return the spelling
1121 /// location referenced by the ID.
1122 ///
1123 /// This is the place where the characters that make up the lexed token
1124 /// can be found.
1125 SourceLocation getSpellingLoc(SourceLocation Loc) const {
1126 // Handle the non-mapped case inline, defer to out of line code to handle
1127 // expansions.
1128 if (Loc.isFileID()) return Loc;
1129 return getSpellingLocSlowCase(Loc);
1130 }
1131
1132 /// \brief Given a SourceLocation object, return the spelling location
1133 /// referenced by the ID.
1134 ///
1135 /// This is the first level down towards the place where the characters
1136 /// that make up the lexed token can be found. This should not generally
1137 /// be used by clients.
1138 SourceLocation getImmediateSpellingLoc(SourceLocation Loc) const;
1139
1140 /// \brief Decompose the specified location into a raw FileID + Offset pair.
1141 ///
1142 /// The first element is the FileID, the second is the offset from the
1143 /// start of the buffer of the location.
1144 std::pair<FileID, unsigned> getDecomposedLoc(SourceLocation Loc) const {
1145 FileID FID = getFileID(Loc);
1146 bool Invalid = false;
1147 const SrcMgr::SLocEntry &E = getSLocEntry(FID, &Invalid);
1148 if (Invalid)
1149 return std::make_pair(FileID(), 0);
1150 return std::make_pair(FID, Loc.getOffset()-E.getOffset());
1151 }
1152
1153 /// \brief Decompose the specified location into a raw FileID + Offset pair.
1154 ///
1155 /// If the location is an expansion record, walk through it until we find
1156 /// the final location expanded.
1157 std::pair<FileID, unsigned>
1158 getDecomposedExpansionLoc(SourceLocation Loc) const {
1159 FileID FID = getFileID(Loc);
1160 bool Invalid = false;
1161 const SrcMgr::SLocEntry *E = &getSLocEntry(FID, &Invalid);
1162 if (Invalid)
1163 return std::make_pair(FileID(), 0);
1164
1165 unsigned Offset = Loc.getOffset()-E->getOffset();
1166 if (Loc.isFileID())
1167 return std::make_pair(FID, Offset);
1168
1169 return getDecomposedExpansionLocSlowCase(E);
1170 }
1171
1172 /// \brief Decompose the specified location into a raw FileID + Offset pair.
1173 ///
1174 /// If the location is an expansion record, walk through it until we find
1175 /// its spelling record.
1176 std::pair<FileID, unsigned>
1177 getDecomposedSpellingLoc(SourceLocation Loc) const {
1178 FileID FID = getFileID(Loc);
1179 bool Invalid = false;
1180 const SrcMgr::SLocEntry *E = &getSLocEntry(FID, &Invalid);
1181 if (Invalid)
1182 return std::make_pair(FileID(), 0);
1183
1184 unsigned Offset = Loc.getOffset()-E->getOffset();
1185 if (Loc.isFileID())
1186 return std::make_pair(FID, Offset);
1187 return getDecomposedSpellingLocSlowCase(E, Offset);
1188 }
1189
1190 /// \brief Returns the "included/expanded in" decomposed location of the given
1191 /// FileID.
1192 std::pair<FileID, unsigned> getDecomposedIncludedLoc(FileID FID) const;
1193
1194 /// \brief Returns the offset from the start of the file that the
1195 /// specified SourceLocation represents.
1196 ///
1197 /// This is not very meaningful for a macro ID.
1198 unsigned getFileOffset(SourceLocation SpellingLoc) const {
1199 return getDecomposedLoc(SpellingLoc).second;
1200 }
1201
1202 /// \brief Tests whether the given source location represents a macro
1203 /// argument's expansion into the function-like macro definition.
1204 ///
1205 /// \param StartLoc If non-null and function returns true, it is set to the
1206 /// start location of the macro argument expansion.
1207 ///
1208 /// Such source locations only appear inside of the expansion
1209 /// locations representing where a particular function-like macro was
1210 /// expanded.
1211 bool isMacroArgExpansion(SourceLocation Loc,
1212 SourceLocation *StartLoc = nullptr) const;
1213
1214 /// \brief Tests whether the given source location represents the expansion of
1215 /// a macro body.
1216 ///
1217 /// This is equivalent to testing whether the location is part of a macro
1218 /// expansion but not the expansion of an argument to a function-like macro.
1219 bool isMacroBodyExpansion(SourceLocation Loc) const;
1220
1221 /// \brief Returns true if the given MacroID location points at the beginning
1222 /// of the immediate macro expansion.
1223 ///
1224 /// \param MacroBegin If non-null and function returns true, it is set to the
1225 /// begin location of the immediate macro expansion.
1226 bool isAtStartOfImmediateMacroExpansion(SourceLocation Loc,
1227 SourceLocation *MacroBegin = nullptr) const;
1228
1229 /// \brief Returns true if the given MacroID location points at the character
1230 /// end of the immediate macro expansion.
1231 ///
1232 /// \param MacroEnd If non-null and function returns true, it is set to the
1233 /// character end location of the immediate macro expansion.
1234 bool
1235 isAtEndOfImmediateMacroExpansion(SourceLocation Loc,
1236 SourceLocation *MacroEnd = nullptr) const;
1237
1238 /// \brief Returns true if \p Loc is inside the [\p Start, +\p Length)
1239 /// chunk of the source location address space.
1240 ///
1241 /// If it's true and \p RelativeOffset is non-null, it will be set to the
1242 /// relative offset of \p Loc inside the chunk.
1243 bool isInSLocAddrSpace(SourceLocation Loc,
1244 SourceLocation Start, unsigned Length,
1245 unsigned *RelativeOffset = nullptr) const {
1246 assert(((Start.getOffset() < NextLocalOffset &&
1247 Start.getOffset()+Length <= NextLocalOffset) ||
1248 (Start.getOffset() >= CurrentLoadedOffset &&
1249 Start.getOffset()+Length < MaxLoadedOffset)) &&
1250 "Chunk is not valid SLoc address space");
1251 unsigned LocOffs = Loc.getOffset();
1252 unsigned BeginOffs = Start.getOffset();
1253 unsigned EndOffs = BeginOffs + Length;
1254 if (LocOffs >= BeginOffs && LocOffs < EndOffs) {
1255 if (RelativeOffset)
1256 *RelativeOffset = LocOffs - BeginOffs;
1257 return true;
1258 }
1259
1260 return false;
1261 }
1262
1263 /// \brief Return true if both \p LHS and \p RHS are in the local source
1264 /// location address space or the loaded one.
1265 ///
1266 /// If it's true and \p RelativeOffset is non-null, it will be set to the
1267 /// offset of \p RHS relative to \p LHS.
1268 bool isInSameSLocAddrSpace(SourceLocation LHS, SourceLocation RHS,
1269 int *RelativeOffset) const {
1270 unsigned LHSOffs = LHS.getOffset(), RHSOffs = RHS.getOffset();
1271 bool LHSLoaded = LHSOffs >= CurrentLoadedOffset;
1272 bool RHSLoaded = RHSOffs >= CurrentLoadedOffset;
1273
1274 if (LHSLoaded == RHSLoaded) {
1275 if (RelativeOffset)
1276 *RelativeOffset = RHSOffs - LHSOffs;
1277 return true;
1278 }
1279
1280 return false;
1281 }
1282
1283 //===--------------------------------------------------------------------===//
1284 // Queries about the code at a SourceLocation.
1285 //===--------------------------------------------------------------------===//
1286
1287 /// \brief Return a pointer to the start of the specified location
1288 /// in the appropriate spelling MemoryBuffer.
1289 ///
1290 /// \param Invalid If non-NULL, will be set \c true if an error occurs.
1291 const char *getCharacterData(SourceLocation SL,
1292 bool *Invalid = nullptr) const;
1293
1294 /// \brief Return the column # for the specified file position.
1295 ///
1296 /// This is significantly cheaper to compute than the line number. This
1297 /// returns zero if the column number isn't known. This may only be called
1298 /// on a file sloc, so you must choose a spelling or expansion location
1299 /// before calling this method.
1300 unsigned getColumnNumber(FileID FID, unsigned FilePos,
1301 bool *Invalid = nullptr) const;
1302 unsigned getSpellingColumnNumber(SourceLocation Loc,
1303 bool *Invalid = nullptr) const;
1304 unsigned getExpansionColumnNumber(SourceLocation Loc,
1305 bool *Invalid = nullptr) const;
1306 unsigned getPresumedColumnNumber(SourceLocation Loc,
1307 bool *Invalid = nullptr) const;
1308
1309 /// \brief Given a SourceLocation, return the spelling line number
1310 /// for the position indicated.
1311 ///
1312 /// This requires building and caching a table of line offsets for the
1313 /// MemoryBuffer, so this is not cheap: use only when about to emit a
1314 /// diagnostic.
1315 unsigned getLineNumber(FileID FID, unsigned FilePos, bool *Invalid = nullptr) const;
1316 unsigned getSpellingLineNumber(SourceLocation Loc, bool *Invalid = nullptr) const;
1317 unsigned getExpansionLineNumber(SourceLocation Loc, bool *Invalid = nullptr) const;
1318 unsigned getPresumedLineNumber(SourceLocation Loc, bool *Invalid = nullptr) const;
1319
1320 /// \brief Return the filename or buffer identifier of the buffer the
1321 /// location is in.
1322 ///
1323 /// Note that this name does not respect \#line directives. Use
1324 /// getPresumedLoc for normal clients.
1325 StringRef getBufferName(SourceLocation Loc, bool *Invalid = nullptr) const;
1326
1327 /// \brief Return the file characteristic of the specified source
1328 /// location, indicating whether this is a normal file, a system
1329 /// header, or an "implicit extern C" system header.
1330 ///
1331 /// This state can be modified with flags on GNU linemarker directives like:
1332 /// \code
1333 /// # 4 "foo.h" 3
1334 /// \endcode
1335 /// which changes all source locations in the current file after that to be
1336 /// considered to be from a system header.
1337 SrcMgr::CharacteristicKind getFileCharacteristic(SourceLocation Loc) const;
1338
1339 /// \brief Returns the "presumed" location of a SourceLocation specifies.
1340 ///
1341 /// A "presumed location" can be modified by \#line or GNU line marker
1342 /// directives. This provides a view on the data that a user should see
1343 /// in diagnostics, for example.
1344 ///
1345 /// Note that a presumed location is always given as the expansion point of
1346 /// an expansion location, not at the spelling location.
1347 ///
1348 /// \returns The presumed location of the specified SourceLocation. If the
1349 /// presumed location cannot be calculated (e.g., because \p Loc is invalid
1350 /// or the file containing \p Loc has changed on disk), returns an invalid
1351 /// presumed location.
1352 PresumedLoc getPresumedLoc(SourceLocation Loc,
1353 bool UseLineDirectives = true) const;
1354
1355 /// \brief Returns whether the PresumedLoc for a given SourceLocation is
1356 /// in the main file.
1357 ///
1358 /// This computes the "presumed" location for a SourceLocation, then checks
1359 /// whether it came from a file other than the main file. This is different
1360 /// from isWrittenInMainFile() because it takes line marker directives into
1361 /// account.
1362 bool isInMainFile(SourceLocation Loc) const;
1363
1364 /// \brief Returns true if the spelling locations for both SourceLocations
1365 /// are part of the same file buffer.
1366 ///
1367 /// This check ignores line marker directives.
1368 bool isWrittenInSameFile(SourceLocation Loc1, SourceLocation Loc2) const {
1369 return getFileID(Loc1) == getFileID(Loc2);
1370 }
1371
1372 /// \brief Returns true if the spelling location for the given location
1373 /// is in the main file buffer.
1374 ///
1375 /// This check ignores line marker directives.
1376 bool isWrittenInMainFile(SourceLocation Loc) const {
1377 return getFileID(Loc) == getMainFileID();
1378 }
1379
1380 /// \brief Returns if a SourceLocation is in a system header.
1381 bool isInSystemHeader(SourceLocation Loc) const {
1382 return isSystem(getFileCharacteristic(Loc));
1383 }
1384
1385 /// \brief Returns if a SourceLocation is in an "extern C" system header.
1386 bool isInExternCSystemHeader(SourceLocation Loc) const {
1387 return getFileCharacteristic(Loc) == SrcMgr::C_ExternCSystem;
1388 }
1389
1390 /// \brief Returns whether \p Loc is expanded from a macro in a system header.
1391 bool isInSystemMacro(SourceLocation loc) const {
1392 return loc.isMacroID() && isInSystemHeader(getSpellingLoc(loc));
1393 }
1394
1395 /// \brief The size of the SLocEntry that \p FID represents.
1396 unsigned getFileIDSize(FileID FID) const;
1397
1398 /// \brief Given a specific FileID, returns true if \p Loc is inside that
1399 /// FileID chunk and sets relative offset (offset of \p Loc from beginning
1400 /// of FileID) to \p relativeOffset.
1401 bool isInFileID(SourceLocation Loc, FileID FID,
1402 unsigned *RelativeOffset = nullptr) const {
1403 unsigned Offs = Loc.getOffset();
1404 if (isOffsetInFileID(FID, Offs)) {
1405 if (RelativeOffset)
1406 *RelativeOffset = Offs - getSLocEntry(FID).getOffset();
1407 return true;
1408 }
1409
1410 return false;
1411 }
1412
1413 //===--------------------------------------------------------------------===//
1414 // Line Table Manipulation Routines
1415 //===--------------------------------------------------------------------===//
1416
1417 /// \brief Return the uniqued ID for the specified filename.
1418 unsigned getLineTableFilenameID(StringRef Str);
1419
1420 /// \brief Add a line note to the line table for the FileID and offset
1421 /// specified by Loc.
1422 ///
1423 /// If FilenameID is -1, it is considered to be unspecified.
1424 void AddLineNote(SourceLocation Loc, unsigned LineNo, int FilenameID,
1425 bool IsFileEntry, bool IsFileExit,
1426 SrcMgr::CharacteristicKind FileKind);
1427
1428 /// \brief Determine if the source manager has a line table.
1429 bool hasLineTable() const { return LineTable != nullptr; }
1430
1431 /// \brief Retrieve the stored line table.
1432 LineTableInfo &getLineTable();
1433
1434 //===--------------------------------------------------------------------===//
1435 // Queries for performance analysis.
1436 //===--------------------------------------------------------------------===//
1437
1438 /// \brief Return the total amount of physical memory allocated by the
1439 /// ContentCache allocator.
1440 size_t getContentCacheSize() const {
1441 return ContentCacheAlloc.getTotalMemory();
1442 }
1443
1444 struct MemoryBufferSizes {
1445 const size_t malloc_bytes;
1446 const size_t mmap_bytes;
1447
1448 MemoryBufferSizes(size_t malloc_bytes, size_t mmap_bytes)
1449 : malloc_bytes(malloc_bytes), mmap_bytes(mmap_bytes) {}
1450 };
1451
1452 /// \brief Return the amount of memory used by memory buffers, breaking down
1453 /// by heap-backed versus mmap'ed memory.
1454 MemoryBufferSizes getMemoryBufferSizes() const;
1455
1456 /// \brief Return the amount of memory used for various side tables and
1457 /// data structures in the SourceManager.
1458 size_t getDataStructureSizes() const;
1459
1460 //===--------------------------------------------------------------------===//
1461 // Other miscellaneous methods.
1462 //===--------------------------------------------------------------------===//
1463
1464 /// \brief Get the source location for the given file:line:col triplet.
1465 ///
1466 /// If the source file is included multiple times, the source location will
1467 /// be based upon the first inclusion.
1468 SourceLocation translateFileLineCol(const FileEntry *SourceFile,
1469 unsigned Line, unsigned Col) const;
1470
1471 /// \brief Get the FileID for the given file.
1472 ///
1473 /// If the source file is included multiple times, the FileID will be the
1474 /// first inclusion.
1475 FileID translateFile(const FileEntry *SourceFile) const;
1476
1477 /// \brief Get the source location in \p FID for the given line:col.
1478 /// Returns null location if \p FID is not a file SLocEntry.
1479 SourceLocation translateLineCol(FileID FID,
1480 unsigned Line, unsigned Col) const;
1481
1482 /// \brief If \p Loc points inside a function macro argument, the returned
1483 /// location will be the macro location in which the argument was expanded.
1484 /// If a macro argument is used multiple times, the expanded location will
1485 /// be at the first expansion of the argument.
1486 /// e.g.
1487 /// MY_MACRO(foo);
1488 /// ^
1489 /// Passing a file location pointing at 'foo', will yield a macro location
1490 /// where 'foo' was expanded into.
1491 SourceLocation getMacroArgExpandedLocation(SourceLocation Loc) const;
1492
1493 /// \brief Determines the order of 2 source locations in the translation unit.
1494 ///
1495 /// \returns true if LHS source location comes before RHS, false otherwise.
1496 bool isBeforeInTranslationUnit(SourceLocation LHS, SourceLocation RHS) const;
1497
1498 /// \brief Determines whether the two decomposed source location is in the
1499 /// same translation unit. As a byproduct, it also calculates the order
1500 /// of the source locations in case they are in the same TU.
1501 ///
1502 /// \returns Pair of bools the first component is true if the two locations
1503 /// are in the same TU. The second bool is true if the first is true
1504 /// and \p LOffs is before \p ROffs.
1505 std::pair<bool, bool>
1506 isInTheSameTranslationUnit(std::pair<FileID, unsigned> &LOffs,
1507 std::pair<FileID, unsigned> &ROffs) const;
1508
1509 /// \brief Determines the order of 2 source locations in the "source location
1510 /// address space".
1511 bool isBeforeInSLocAddrSpace(SourceLocation LHS, SourceLocation RHS) const {
1512 return isBeforeInSLocAddrSpace(LHS, RHS.getOffset());
1513 }
1514
1515 /// \brief Determines the order of a source location and a source location
1516 /// offset in the "source location address space".
1517 ///
1518 /// Note that we always consider source locations loaded from
1519 bool isBeforeInSLocAddrSpace(SourceLocation LHS, unsigned RHS) const {
1520 unsigned LHSOffset = LHS.getOffset();
1521 bool LHSLoaded = LHSOffset >= CurrentLoadedOffset;
1522 bool RHSLoaded = RHS >= CurrentLoadedOffset;
1523 if (LHSLoaded == RHSLoaded)
1524 return LHSOffset < RHS;
1525
1526 return LHSLoaded;
1527 }
1528
1529 /// Return true if the Point is within Start and End.
1530 bool isPointWithin(SourceLocation Location, SourceLocation Start,
1531 SourceLocation End) const {
1532 return Location == Start || Location == End ||
1533 (isBeforeInTranslationUnit(Start, Location) &&
1534 isBeforeInTranslationUnit(Location, End));
1535 }
1536
1537 // Iterators over FileInfos.
1538 using fileinfo_iterator =
1539 llvm::DenseMap<const FileEntry*, SrcMgr::ContentCache*>::const_iterator;
1540
1541 fileinfo_iterator fileinfo_begin() const { return FileInfos.begin(); }
1542 fileinfo_iterator fileinfo_end() const { return FileInfos.end(); }
1543 bool hasFileInfo(const FileEntry *File) const {
1544 return FileInfos.find(File) != FileInfos.end();
1545 }
1546
1547 /// \brief Print statistics to stderr.
1548 void PrintStats() const;
1549
1550 void dump() const;
1551
1552 /// \brief Get the number of local SLocEntries we have.
1553 unsigned local_sloc_entry_size() const { return LocalSLocEntryTable.size(); }
1554
1555 /// \brief Get a local SLocEntry. This is exposed for indexing.
1556 const SrcMgr::SLocEntry &getLocalSLocEntry(unsigned Index,
1557 bool *Invalid = nullptr) const {
1558 assert(Index < LocalSLocEntryTable.size() && "Invalid index");
1559 return LocalSLocEntryTable[Index];
1560 }
1561
1562 /// \brief Get the number of loaded SLocEntries we have.
1563 unsigned loaded_sloc_entry_size() const { return LoadedSLocEntryTable.size();}
1564
1565 /// \brief Get a loaded SLocEntry. This is exposed for indexing.
1566 const SrcMgr::SLocEntry &getLoadedSLocEntry(unsigned Index,
1567 bool *Invalid = nullptr) const {
1568 assert(Index < LoadedSLocEntryTable.size() && "Invalid index");
1569 if (SLocEntryLoaded[Index])
1570 return LoadedSLocEntryTable[Index];
1571 return loadSLocEntry(Index, Invalid);
1572 }
1573
1574 const SrcMgr::SLocEntry &getSLocEntry(FileID FID,
1575 bool *Invalid = nullptr) const {
1576 if (FID.ID == 0 || FID.ID == -1) {
1577 if (Invalid) *Invalid = true;
1578 return LocalSLocEntryTable[0];
1579 }
1580 return getSLocEntryByID(FID.ID, Invalid);
1581 }
1582
1583 unsigned getNextLocalOffset() const { return NextLocalOffset; }
1584
1585 void setExternalSLocEntrySource(ExternalSLocEntrySource *Source) {
1586 assert(LoadedSLocEntryTable.empty() &&
1587 "Invalidating existing loaded entries");
1588 ExternalSLocEntries = Source;
1589 }
1590
1591 /// \brief Allocate a number of loaded SLocEntries, which will be actually
1592 /// loaded on demand from the external source.
1593 ///
1594 /// NumSLocEntries will be allocated, which occupy a total of TotalSize space
1595 /// in the global source view. The lowest ID and the base offset of the
1596 /// entries will be returned.
1597 std::pair<int, unsigned>
1598 AllocateLoadedSLocEntries(unsigned NumSLocEntries, unsigned TotalSize);
1599
1600 /// \brief Returns true if \p Loc came from a PCH/Module.
1601 bool isLoadedSourceLocation(SourceLocation Loc) const {
1602 return Loc.getOffset() >= CurrentLoadedOffset;
1603 }
1604
1605 /// \brief Returns true if \p Loc did not come from a PCH/Module.
1606 bool isLocalSourceLocation(SourceLocation Loc) const {
1607 return Loc.getOffset() < NextLocalOffset;
1608 }
1609
1610 /// \brief Returns true if \p FID came from a PCH/Module.
1611 bool isLoadedFileID(FileID FID) const {
1612 assert(FID.ID != -1 && "Using FileID sentinel value");
1613 return FID.ID < 0;
1614 }
1615
1616 /// \brief Returns true if \p FID did not come from a PCH/Module.
1617 bool isLocalFileID(FileID FID) const {
1618 return !isLoadedFileID(FID);
1619 }
1620
1621 /// Gets the location of the immediate macro caller, one level up the stack
1622 /// toward the initial macro typed into the source.
1623 SourceLocation getImmediateMacroCallerLoc(SourceLocation Loc) const {
1624 if (!Loc.isMacroID()) return Loc;
1625
1626 // When we have the location of (part of) an expanded parameter, its
1627 // spelling location points to the argument as expanded in the macro call,
1628 // and therefore is used to locate the macro caller.
1629 if (isMacroArgExpansion(Loc))
1630 return getImmediateSpellingLoc(Loc);
1631
1632 // Otherwise, the caller of the macro is located where this macro is
1633 // expanded (while the spelling is part of the macro definition).
1634 return getImmediateExpansionRange(Loc).first;
1635 }
1636
1637private:
1638 friend class ASTReader;
1639 friend class ASTWriter;
1640
1641 llvm::MemoryBuffer *getFakeBufferForRecovery() const;
1642 const SrcMgr::ContentCache *getFakeContentCacheForRecovery() const;
1643
1644 const SrcMgr::SLocEntry &loadSLocEntry(unsigned Index, bool *Invalid) const;
1645
1646 /// \brief Get the entry with the given unwrapped FileID.
1647 const SrcMgr::SLocEntry &getSLocEntryByID(int ID,
1648 bool *Invalid = nullptr) const {
1649 assert(ID != -1 && "Using FileID sentinel value");
1650 if (ID < 0)
1651 return getLoadedSLocEntryByID(ID, Invalid);
1652 return getLocalSLocEntry(static_cast<unsigned>(ID), Invalid);
1653 }
1654
1655 const SrcMgr::SLocEntry &
1656 getLoadedSLocEntryByID(int ID, bool *Invalid = nullptr) const {
1657 return getLoadedSLocEntry(static_cast<unsigned>(-ID - 2), Invalid);
1658 }
1659
1660 /// Implements the common elements of storing an expansion info struct into
1661 /// the SLocEntry table and producing a source location that refers to it.
1662 SourceLocation createExpansionLocImpl(const SrcMgr::ExpansionInfo &Expansion,
1663 unsigned TokLength,
1664 int LoadedID = 0,
1665 unsigned LoadedOffset = 0);
1666
1667 /// \brief Return true if the specified FileID contains the
1668 /// specified SourceLocation offset. This is a very hot method.
1669 inline bool isOffsetInFileID(FileID FID, unsigned SLocOffset) const {
1670 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID);
1671 // If the entry is after the offset, it can't contain it.
1672 if (SLocOffset < Entry.getOffset()) return false;
1673
1674 // If this is the very last entry then it does.
1675 if (FID.ID == -2)
1676 return true;
1677
1678 // If it is the last local entry, then it does if the location is local.
1679 if (FID.ID+1 == static_cast<int>(LocalSLocEntryTable.size()))
1680 return SLocOffset < NextLocalOffset;
1681
1682 // Otherwise, the entry after it has to not include it. This works for both
1683 // local and loaded entries.
1684 return SLocOffset < getSLocEntryByID(FID.ID+1).getOffset();
1685 }
1686
1687 /// \brief Returns the previous in-order FileID or an invalid FileID if there
1688 /// is no previous one.
1689 FileID getPreviousFileID(FileID FID) const;
1690
1691 /// \brief Returns the next in-order FileID or an invalid FileID if there is
1692 /// no next one.
1693 FileID getNextFileID(FileID FID) const;
1694
1695 /// \brief Create a new fileID for the specified ContentCache and
1696 /// include position.
1697 ///
1698 /// This works regardless of whether the ContentCache corresponds to a
1699 /// file or some other input source.
1700 FileID createFileID(const SrcMgr::ContentCache* File,
1701 SourceLocation IncludePos,
1702 SrcMgr::CharacteristicKind DirCharacter,
1703 int LoadedID, unsigned LoadedOffset);
1704
1705 const SrcMgr::ContentCache *
1706 getOrCreateContentCache(const FileEntry *SourceFile,
1707 bool isSystemFile = false);
1708
1709 /// \brief Create a new ContentCache for the specified memory buffer.
1710 const SrcMgr::ContentCache *
1711 createMemBufferContentCache(llvm::MemoryBuffer *Buf, bool DoNotFree);
1712
1713 FileID getFileIDSlow(unsigned SLocOffset) const;
1714 FileID getFileIDLocal(unsigned SLocOffset) const;
1715 FileID getFileIDLoaded(unsigned SLocOffset) const;
1716
1717 SourceLocation getExpansionLocSlowCase(SourceLocation Loc) const;
1718 SourceLocation getSpellingLocSlowCase(SourceLocation Loc) const;
1719 SourceLocation getFileLocSlowCase(SourceLocation Loc) const;
1720
1721 std::pair<FileID, unsigned>
1722 getDecomposedExpansionLocSlowCase(const SrcMgr::SLocEntry *E) const;
1723 std::pair<FileID, unsigned>
1724 getDecomposedSpellingLocSlowCase(const SrcMgr::SLocEntry *E,
1725 unsigned Offset) const;
1726 void computeMacroArgsCache(MacroArgsMap &MacroArgsCache, FileID FID) const;
1727 void associateFileChunkWithMacroArgExp(MacroArgsMap &MacroArgsCache,
1728 FileID FID,
1729 SourceLocation SpellLoc,
1730 SourceLocation ExpansionLoc,
1731 unsigned ExpansionLength) const;
1732};
1733
1734/// \brief Comparison function object.
1735template<typename T>
1736class BeforeThanCompare;
1737
1738/// \brief Compare two source locations.
1739template<>
1740class BeforeThanCompare<SourceLocation> {
1741 SourceManager &SM;
1742
1743public:
1744 explicit BeforeThanCompare(SourceManager &SM) : SM(SM) {}
1745
1746 bool operator()(SourceLocation LHS, SourceLocation RHS) const {
1747 return SM.isBeforeInTranslationUnit(LHS, RHS);
1748 }
1749};
1750
1751/// \brief Compare two non-overlapping source ranges.
1752template<>
1753class BeforeThanCompare<SourceRange> {
1754 SourceManager &SM;
1755
1756public:
1757 explicit BeforeThanCompare(SourceManager &SM) : SM(SM) {}
1758
1759 bool operator()(SourceRange LHS, SourceRange RHS) const {
1760 return SM.isBeforeInTranslationUnit(LHS.getBegin(), RHS.getBegin());
1761 }
1762};
1763
1764} // namespace clang
1765
1766#endif // LLVM_CLANG_BASIC_SOURCEMANAGER_H
1767