1//===- SourceManager.cpp - Track and cache source files -------------------===//
2//
3// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4// See https://llvm.org/LICENSE.txt for license information.
5// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6//
7//===----------------------------------------------------------------------===//
8//
9// This file implements the SourceManager interface.
10//
11//===----------------------------------------------------------------------===//
12
13#include "clang/Basic/SourceManager.h"
14#include "clang/Basic/Diagnostic.h"
15#include "clang/Basic/FileManager.h"
16#include "clang/Basic/LLVM.h"
17#include "clang/Basic/SourceLocation.h"
18#include "clang/Basic/SourceManagerInternals.h"
19#include "llvm/ADT/DenseMap.h"
20#include "llvm/ADT/MapVector.h"
21#include "llvm/ADT/STLExtras.h"
22#include "llvm/ADT/SmallVector.h"
23#include "llvm/ADT/Statistic.h"
24#include "llvm/ADT/StringRef.h"
25#include "llvm/ADT/StringSwitch.h"
26#include "llvm/Support/Allocator.h"
27#include "llvm/Support/AutoConvert.h"
28#include "llvm/Support/Capacity.h"
29#include "llvm/Support/Compiler.h"
30#include "llvm/Support/Endian.h"
31#include "llvm/Support/ErrorHandling.h"
32#include "llvm/Support/MemoryBuffer.h"
33#include "llvm/Support/raw_ostream.h"
34#include <algorithm>
35#include <cassert>
36#include <cstddef>
37#include <cstdint>
38#include <memory>
39#include <optional>
40#include <tuple>
41#include <utility>
42#include <vector>
43
44using namespace clang;
45using namespace SrcMgr;
46using llvm::MemoryBuffer;
47
48#define DEBUG_TYPE "source-manager"
49
50// Reaching a limit of 2^31 results in a hard error. This metric allows to track
51// if particular invocation of the compiler is close to it.
52STATISTIC(MaxUsedSLocBytes, "Maximum number of bytes used by source locations "
53 "(both loaded and local).");
54
55//===----------------------------------------------------------------------===//
56// SourceManager Helper Classes
57//===----------------------------------------------------------------------===//
58
59/// getSizeBytesMapped - Returns the number of bytes actually mapped for this
60/// ContentCache. This can be 0 if the MemBuffer was not actually expanded.
61unsigned ContentCache::getSizeBytesMapped() const {
62 return Buffer ? Buffer->getBufferSize() : 0;
63}
64
65/// Returns the kind of memory used to back the memory buffer for
66/// this content cache. This is used for performance analysis.
67llvm::MemoryBuffer::BufferKind ContentCache::getMemoryBufferKind() const {
68 if (Buffer == nullptr) {
69 assert(0 && "Buffer should never be null");
70 return llvm::MemoryBuffer::MemoryBuffer_Malloc;
71 }
72 return Buffer->getBufferKind();
73}
74
75/// getSize - Returns the size of the content encapsulated by this ContentCache.
76/// This can be the size of the source file or the size of an arbitrary
77/// scratch buffer. If the ContentCache encapsulates a source file, that
78/// file is not lazily brought in from disk to satisfy this query.
79unsigned ContentCache::getSize() const {
80 return Buffer ? (unsigned)Buffer->getBufferSize()
81 : (unsigned)ContentsEntry->getSize();
82}
83
84const char *ContentCache::getInvalidBOM(StringRef BufStr) {
85 // If the buffer is valid, check to see if it has a UTF Byte Order Mark
86 // (BOM). We only support UTF-8 with and without a BOM right now. See
87 // http://en.wikipedia.org/wiki/Byte_order_mark for more information.
88 const char *InvalidBOM =
89 llvm::StringSwitch<const char *>(BufStr)
90 .StartsWith(S: llvm::StringLiteral::withInnerNUL(Str: "\x00\x00\xFE\xFF"),
91 Value: "UTF-32 (BE)")
92 .StartsWith(S: llvm::StringLiteral::withInnerNUL(Str: "\xFF\xFE\x00\x00"),
93 Value: "UTF-32 (LE)")
94 .StartsWith(S: "\xFE\xFF", Value: "UTF-16 (BE)")
95 .StartsWith(S: "\xFF\xFE", Value: "UTF-16 (LE)")
96 .StartsWith(S: "\x2B\x2F\x76", Value: "UTF-7")
97 .StartsWith(S: "\xF7\x64\x4C", Value: "UTF-1")
98 .StartsWith(S: "\xDD\x73\x66\x73", Value: "UTF-EBCDIC")
99 .StartsWith(S: "\x0E\xFE\xFF", Value: "SCSU")
100 .StartsWith(S: "\xFB\xEE\x28", Value: "BOCU-1")
101 .StartsWith(S: "\x84\x31\x95\x33", Value: "GB-18030")
102 .Default(Value: nullptr);
103
104 return InvalidBOM;
105}
106
107std::optional<llvm::MemoryBufferRef>
108ContentCache::getBufferOrNone(DiagnosticsEngine &Diag, FileManager &FM,
109 SourceLocation Loc) const {
110 // Lazily create the Buffer for ContentCaches that wrap files. If we already
111 // computed it, just return what we have.
112 if (IsBufferInvalid)
113 return std::nullopt;
114 if (Buffer)
115 return Buffer->getMemBufferRef();
116 if (!ContentsEntry)
117 return std::nullopt;
118
119 // Start with the assumption that the buffer is invalid to simplify early
120 // return paths.
121 IsBufferInvalid = true;
122
123 auto BufferOrError = FM.getBufferForFile(Entry: *ContentsEntry, isVolatile: IsFileVolatile);
124
125 // If we were unable to open the file, then we are in an inconsistent
126 // situation where the content cache referenced a file which no longer
127 // exists. Most likely, we were using a stat cache with an invalid entry but
128 // the file could also have been removed during processing. Since we can't
129 // really deal with this situation, just create an empty buffer.
130 if (!BufferOrError) {
131 Diag.Report(Loc, DiagID: diag::err_cannot_open_file)
132 << ContentsEntry->getName() << BufferOrError.getError().message();
133
134 return std::nullopt;
135 }
136
137 Buffer = std::move(*BufferOrError);
138
139 // Check that the file's size fits in an 'unsigned' (with room for a
140 // past-the-end value). This is deeply regrettable, but various parts of
141 // Clang (including elsewhere in this file!) use 'unsigned' to represent file
142 // offsets, line numbers, string literal lengths, and so on, and fail
143 // miserably on large source files.
144 //
145 // Note: ContentsEntry could be a named pipe, in which case
146 // ContentsEntry::getSize() could have the wrong size. Use
147 // MemoryBuffer::getBufferSize() instead.
148 if (Buffer->getBufferSize() >= std::numeric_limits<unsigned>::max()) {
149 Diag.Report(Loc, DiagID: diag::err_file_too_large) << ContentsEntry->getName();
150
151 return std::nullopt;
152 }
153
154 // Unless this is a named pipe (in which case we can handle a mismatch),
155 // check that the file's size is the same as in the file entry (which may
156 // have come from a stat cache).
157 // The buffer will always be larger than the file size on z/OS in the presence
158 // of characters outside the base character set.
159 assert(Buffer->getBufferSize() >= (size_t)ContentsEntry->getSize());
160 if (!ContentsEntry->isNamedPipe() &&
161 Buffer->getBufferSize() < (size_t)ContentsEntry->getSize()) {
162 Diag.Report(Loc, DiagID: diag::err_file_modified) << ContentsEntry->getName();
163
164 return std::nullopt;
165 }
166
167 // If the buffer is valid, check to see if it has a UTF Byte Order Mark
168 // (BOM). We only support UTF-8 with and without a BOM right now. See
169 // http://en.wikipedia.org/wiki/Byte_order_mark for more information.
170 StringRef BufStr = Buffer->getBuffer();
171 const char *InvalidBOM = getInvalidBOM(BufStr);
172
173 if (InvalidBOM) {
174 Diag.Report(Loc, DiagID: diag::err_unsupported_bom)
175 << InvalidBOM << ContentsEntry->getName();
176 return std::nullopt;
177 }
178
179 // Buffer has been validated.
180 IsBufferInvalid = false;
181 return Buffer->getMemBufferRef();
182}
183
184unsigned LineTableInfo::getLineTableFilenameID(StringRef Name) {
185 auto IterBool = FilenameIDs.try_emplace(Key: Name, Args: FilenamesByID.size());
186 if (IterBool.second)
187 FilenamesByID.push_back(x: &*IterBool.first);
188 return IterBool.first->second;
189}
190
191/// Add a line note to the line table that indicates that there is a \#line or
192/// GNU line marker at the specified FID/Offset location which changes the
193/// presumed location to LineNo/FilenameID. If EntryExit is 0, then this doesn't
194/// change the presumed \#include stack. If it is 1, this is a file entry, if
195/// it is 2 then this is a file exit. FileKind specifies whether this is a
196/// system header or extern C system header.
197void LineTableInfo::AddLineNote(FileID FID, unsigned Offset, unsigned LineNo,
198 int FilenameID, unsigned EntryExit,
199 SrcMgr::CharacteristicKind FileKind) {
200 std::vector<LineEntry> &Entries = LineEntries[FID];
201
202 assert((Entries.empty() || Entries.back().FileOffset < Offset) &&
203 "Adding line entries out of order!");
204
205 unsigned IncludeOffset = 0;
206 if (EntryExit == 1) {
207 // Push #include
208 IncludeOffset = Offset-1;
209 } else {
210 const auto *PrevEntry = Entries.empty() ? nullptr : &Entries.back();
211 if (EntryExit == 2) {
212 // Pop #include
213 assert(PrevEntry && PrevEntry->IncludeOffset &&
214 "PPDirectives should have caught case when popping empty include "
215 "stack");
216 PrevEntry = FindNearestLineEntry(FID, Offset: PrevEntry->IncludeOffset);
217 }
218 if (PrevEntry) {
219 IncludeOffset = PrevEntry->IncludeOffset;
220 if (FilenameID == -1) {
221 // An unspecified FilenameID means use the previous (or containing)
222 // filename if available, or the main source file otherwise.
223 FilenameID = PrevEntry->FilenameID;
224 }
225 }
226 }
227
228 Entries.push_back(x: LineEntry::get(Offs: Offset, Line: LineNo, Filename: FilenameID, FileKind,
229 IncludeOffset));
230}
231
232/// FindNearestLineEntry - Find the line entry nearest to FID that is before
233/// it. If there is no line entry before Offset in FID, return null.
234const LineEntry *LineTableInfo::FindNearestLineEntry(FileID FID,
235 unsigned Offset) {
236 const std::vector<LineEntry> &Entries = LineEntries[FID];
237 assert(!Entries.empty() && "No #line entries for this FID after all!");
238
239 // It is very common for the query to be after the last #line, check this
240 // first.
241 if (Entries.back().FileOffset <= Offset)
242 return &Entries.back();
243
244 // Do a binary search to find the maximal element that is still before Offset.
245 std::vector<LineEntry>::const_iterator I = llvm::upper_bound(Range: Entries, Value&: Offset);
246 if (I == Entries.begin())
247 return nullptr;
248 return &*--I;
249}
250
251/// Add a new line entry that has already been encoded into
252/// the internal representation of the line table.
253void LineTableInfo::AddEntry(FileID FID,
254 const std::vector<LineEntry> &Entries) {
255 LineEntries[FID] = Entries;
256}
257
258/// getLineTableFilenameID - Return the uniqued ID for the specified filename.
259unsigned SourceManager::getLineTableFilenameID(StringRef Name) {
260 return getLineTable().getLineTableFilenameID(Name);
261}
262
263/// AddLineNote - Add a line note to the line table for the FileID and offset
264/// specified by Loc. If FilenameID is -1, it is considered to be
265/// unspecified.
266void SourceManager::AddLineNote(SourceLocation Loc, unsigned LineNo,
267 int FilenameID, bool IsFileEntry,
268 bool IsFileExit,
269 SrcMgr::CharacteristicKind FileKind) {
270 FileIDAndOffset LocInfo = getDecomposedExpansionLoc(Loc);
271
272 bool Invalid = false;
273 SLocEntry &Entry = getSLocEntry(FID: LocInfo.first, Invalid: &Invalid);
274 if (!Entry.isFile() || Invalid)
275 return;
276
277 SrcMgr::FileInfo &FileInfo = Entry.getFile();
278
279 // Remember that this file has #line directives now if it doesn't already.
280 FileInfo.setHasLineDirectives();
281
282 (void) getLineTable();
283
284 unsigned EntryExit = 0;
285 if (IsFileEntry)
286 EntryExit = 1;
287 else if (IsFileExit)
288 EntryExit = 2;
289
290 LineTable->AddLineNote(FID: LocInfo.first, Offset: LocInfo.second, LineNo, FilenameID,
291 EntryExit, FileKind);
292}
293
294LineTableInfo &SourceManager::getLineTable() {
295 if (!LineTable)
296 LineTable.reset(p: new LineTableInfo());
297 return *LineTable;
298}
299
300//===----------------------------------------------------------------------===//
301// Private 'Create' methods.
302//===----------------------------------------------------------------------===//
303
304SourceManager::SourceManager(DiagnosticsEngine &Diag, FileManager &FileMgr,
305 bool UserFilesAreVolatile)
306 : Diag(Diag), FileMgr(FileMgr), UserFilesAreVolatile(UserFilesAreVolatile) {
307 clearIDTables();
308 Diag.setSourceManager(this);
309}
310
311SourceManager::~SourceManager() {
312 // Delete FileEntry objects corresponding to content caches. Since the actual
313 // content cache objects are bump pointer allocated, we just have to run the
314 // dtors, but we call the deallocate method for completeness.
315 for (unsigned i = 0, e = MemBufferInfos.size(); i != e; ++i) {
316 if (MemBufferInfos[i]) {
317 MemBufferInfos[i]->~ContentCache();
318 ContentCacheAlloc.Deallocate(Ptr: MemBufferInfos[i]);
319 }
320 }
321 for (auto I = FileInfos.begin(), E = FileInfos.end(); I != E; ++I) {
322 if (I->second) {
323 I->second->~ContentCache();
324 ContentCacheAlloc.Deallocate(Ptr: I->second);
325 }
326 }
327}
328
329void SourceManager::clearIDTables() {
330 MainFileID = FileID();
331 LocalSLocEntryTable.clear();
332 LocalLocOffsetTable.clear();
333 LoadedSLocEntryTable.clear();
334 SLocEntryLoaded.clear();
335 SLocEntryOffsetLoaded.clear();
336 LastLineNoFileIDQuery = FileID();
337 LastLineNoContentCache = nullptr;
338 LastFileIDLookup = FileID();
339 LastLookupStartOffset = LastLookupEndOffset = 0;
340
341 IncludedLocMap.clear();
342 if (LineTable)
343 LineTable->clear();
344
345 // Use up FileID #0 as an invalid expansion.
346 NextLocalOffset = 0;
347 CurrentLoadedOffset = MaxLoadedOffset;
348 createExpansionLoc(SpellingLoc: SourceLocation(), ExpansionLocStart: SourceLocation(), ExpansionLocEnd: SourceLocation(), Length: 1);
349 // Diagnostics engine keeps some references to fileids, mostly for dealing
350 // with diagnostic pragmas, make sure they're reset as well.
351 Diag.ResetPragmas();
352}
353
354bool SourceManager::isMainFile(const FileEntry &SourceFile) {
355 assert(MainFileID.isValid() && "expected initialized SourceManager");
356 if (auto *FE = getFileEntryForID(FID: MainFileID))
357 return FE->getUID() == SourceFile.getUID();
358 return false;
359}
360
361void SourceManager::initializeForReplay(const SourceManager &Old) {
362 assert(MainFileID.isInvalid() && "expected uninitialized SourceManager");
363
364 auto CloneContentCache = [&](const ContentCache *Cache) -> ContentCache * {
365 auto *Clone = new (ContentCacheAlloc.Allocate<ContentCache>()) ContentCache;
366 Clone->OrigEntry = Cache->OrigEntry;
367 Clone->ContentsEntry = Cache->ContentsEntry;
368 Clone->BufferOverridden = Cache->BufferOverridden;
369 Clone->IsFileVolatile = Cache->IsFileVolatile;
370 Clone->IsTransient = Cache->IsTransient;
371 Clone->setUnownedBuffer(Cache->getBufferIfLoaded());
372 return Clone;
373 };
374
375 // Ensure all SLocEntries are loaded from the external source.
376 for (unsigned I = 0, N = Old.LoadedSLocEntryTable.size(); I != N; ++I)
377 if (!Old.SLocEntryLoaded[I])
378 Old.loadSLocEntry(Index: I, Invalid: nullptr);
379
380 // Inherit any content cache data from the old source manager.
381 for (auto &FileInfo : Old.FileInfos) {
382 SrcMgr::ContentCache *&Slot = FileInfos[FileInfo.first];
383 if (Slot)
384 continue;
385 Slot = CloneContentCache(FileInfo.second);
386 }
387}
388
389ContentCache &SourceManager::getOrCreateContentCache(FileEntryRef FileEnt,
390 bool isSystemFile) {
391 // Do we already have information about this file?
392 ContentCache *&Entry = FileInfos[FileEnt];
393 if (Entry)
394 return *Entry;
395
396 // Nope, create a new Cache entry.
397 Entry = ContentCacheAlloc.Allocate<ContentCache>();
398
399 if (OverriddenFilesInfo) {
400 // If the file contents are overridden with contents from another file,
401 // pass that file to ContentCache.
402 auto overI = OverriddenFilesInfo->OverriddenFiles.find(Val: FileEnt);
403 if (overI == OverriddenFilesInfo->OverriddenFiles.end())
404 new (Entry) ContentCache(FileEnt);
405 else
406 new (Entry) ContentCache(OverridenFilesKeepOriginalName ? FileEnt
407 : overI->second,
408 overI->second);
409 } else {
410 new (Entry) ContentCache(FileEnt);
411 }
412
413 Entry->IsFileVolatile = UserFilesAreVolatile && !isSystemFile;
414 Entry->IsTransient = FilesAreTransient;
415 Entry->BufferOverridden |= FileEnt.isNamedPipe();
416
417 return *Entry;
418}
419
420/// Create a new ContentCache for the specified memory buffer.
421/// This does no caching.
422ContentCache &SourceManager::createMemBufferContentCache(
423 std::unique_ptr<llvm::MemoryBuffer> Buffer) {
424 // Add a new ContentCache to the MemBufferInfos list and return it.
425 ContentCache *Entry = ContentCacheAlloc.Allocate<ContentCache>();
426 new (Entry) ContentCache();
427 MemBufferInfos.push_back(x: Entry);
428 Entry->setBuffer(std::move(Buffer));
429 return *Entry;
430}
431
432const SrcMgr::SLocEntry &SourceManager::loadSLocEntry(unsigned Index,
433 bool *Invalid) const {
434 return const_cast<SourceManager *>(this)->loadSLocEntry(Index, Invalid);
435}
436
437SrcMgr::SLocEntry &SourceManager::loadSLocEntry(unsigned Index, bool *Invalid) {
438 assert(!SLocEntryLoaded[Index]);
439 if (ExternalSLocEntries->ReadSLocEntry(ID: -(static_cast<int>(Index) + 2))) {
440 if (Invalid)
441 *Invalid = true;
442 // If the file of the SLocEntry changed we could still have loaded it.
443 if (!SLocEntryLoaded[Index]) {
444 // Try to recover; create a SLocEntry so the rest of clang can handle it.
445 if (!FakeSLocEntryForRecovery)
446 FakeSLocEntryForRecovery = std::make_unique<SLocEntry>(args: SLocEntry::get(
447 Offset: 0, FI: FileInfo::get(IL: SourceLocation(), Con&: getFakeContentCacheForRecovery(),
448 FileCharacter: SrcMgr::C_User, Filename: "")));
449 return *FakeSLocEntryForRecovery;
450 }
451 }
452
453 return LoadedSLocEntryTable[Index];
454}
455
456std::pair<int, SourceLocation::UIntTy>
457SourceManager::AllocateLoadedSLocEntries(unsigned NumSLocEntries,
458 SourceLocation::UIntTy TotalSize) {
459 assert(ExternalSLocEntries && "Don't have an external sloc source");
460 // Make sure we're not about to run out of source locations.
461 if (CurrentLoadedOffset < TotalSize ||
462 CurrentLoadedOffset - TotalSize < NextLocalOffset) {
463 return std::make_pair(x: 0, y: 0);
464 }
465 LoadedSLocEntryTable.resize(NewSize: LoadedSLocEntryTable.size() + NumSLocEntries);
466 SLocEntryLoaded.resize(N: LoadedSLocEntryTable.size());
467 SLocEntryOffsetLoaded.resize(N: LoadedSLocEntryTable.size());
468 CurrentLoadedOffset -= TotalSize;
469 updateSlocUsageStats();
470 int BaseID = -int(LoadedSLocEntryTable.size()) - 1;
471 LoadedSLocEntryAllocBegin.push_back(Elt: FileID::get(V: BaseID));
472 return std::make_pair(x&: BaseID, y&: CurrentLoadedOffset);
473}
474
475/// As part of recovering from missing or changed content, produce a
476/// fake, non-empty buffer.
477llvm::MemoryBufferRef SourceManager::getFakeBufferForRecovery() const {
478 if (!FakeBufferForRecovery)
479 FakeBufferForRecovery =
480 llvm::MemoryBuffer::getMemBuffer(InputData: "<<<INVALID BUFFER>>");
481
482 return *FakeBufferForRecovery;
483}
484
485/// As part of recovering from missing or changed content, produce a
486/// fake content cache.
487SrcMgr::ContentCache &SourceManager::getFakeContentCacheForRecovery() const {
488 if (!FakeContentCacheForRecovery) {
489 FakeContentCacheForRecovery = std::make_unique<SrcMgr::ContentCache>();
490 FakeContentCacheForRecovery->setUnownedBuffer(getFakeBufferForRecovery());
491 }
492 return *FakeContentCacheForRecovery;
493}
494
495/// Returns the previous in-order FileID or an invalid FileID if there
496/// is no previous one.
497FileID SourceManager::getPreviousFileID(FileID FID) const {
498 if (FID.isInvalid())
499 return FileID();
500
501 int ID = FID.ID;
502 if (ID == -1)
503 return FileID();
504
505 if (ID > 0) {
506 if (ID-1 == 0)
507 return FileID();
508 } else if (unsigned(-(ID-1) - 2) >= LoadedSLocEntryTable.size()) {
509 return FileID();
510 }
511
512 return FileID::get(V: ID-1);
513}
514
515/// Returns the next in-order FileID or an invalid FileID if there is
516/// no next one.
517FileID SourceManager::getNextFileID(FileID FID) const {
518 if (FID.isInvalid())
519 return FileID();
520
521 int ID = FID.ID;
522 if (ID > 0) {
523 if (unsigned(ID+1) >= local_sloc_entry_size())
524 return FileID();
525 } else if (ID+1 >= -1) {
526 return FileID();
527 }
528
529 return FileID::get(V: ID+1);
530}
531
532//===----------------------------------------------------------------------===//
533// Methods to create new FileID's and macro expansions.
534//===----------------------------------------------------------------------===//
535
536/// Create a new FileID that represents the specified file
537/// being \#included from the specified IncludePosition.
538FileID SourceManager::createFileID(FileEntryRef SourceFile,
539 SourceLocation IncludePos,
540 SrcMgr::CharacteristicKind FileCharacter,
541 int LoadedID,
542 SourceLocation::UIntTy LoadedOffset) {
543 SrcMgr::ContentCache &IR = getOrCreateContentCache(FileEnt: SourceFile,
544 isSystemFile: isSystem(CK: FileCharacter));
545
546 // If this is a named pipe, immediately load the buffer to ensure subsequent
547 // calls to ContentCache::getSize() are accurate.
548 if (IR.ContentsEntry->isNamedPipe())
549 (void)IR.getBufferOrNone(Diag, FM&: getFileManager(), Loc: SourceLocation());
550
551 return createFileIDImpl(File&: IR, Filename: SourceFile.getName(), IncludePos, DirCharacter: FileCharacter,
552 LoadedID, LoadedOffset);
553}
554
555/// Create a new FileID that represents the specified memory buffer.
556///
557/// This does no caching of the buffer and takes ownership of the
558/// MemoryBuffer, so only pass a MemoryBuffer to this once.
559FileID SourceManager::createFileID(std::unique_ptr<llvm::MemoryBuffer> Buffer,
560 SrcMgr::CharacteristicKind FileCharacter,
561 int LoadedID,
562 SourceLocation::UIntTy LoadedOffset,
563 SourceLocation IncludeLoc) {
564 StringRef Name = Buffer->getBufferIdentifier();
565 return createFileIDImpl(File&: createMemBufferContentCache(Buffer: std::move(Buffer)), Filename: Name,
566 IncludePos: IncludeLoc, DirCharacter: FileCharacter, LoadedID, LoadedOffset);
567}
568
569/// Create a new FileID that represents the specified memory buffer.
570///
571/// This does not take ownership of the MemoryBuffer. The memory buffer must
572/// outlive the SourceManager.
573FileID SourceManager::createFileID(const llvm::MemoryBufferRef &Buffer,
574 SrcMgr::CharacteristicKind FileCharacter,
575 int LoadedID,
576 SourceLocation::UIntTy LoadedOffset,
577 SourceLocation IncludeLoc) {
578 return createFileID(Buffer: llvm::MemoryBuffer::getMemBuffer(Ref: Buffer), FileCharacter,
579 LoadedID, LoadedOffset, IncludeLoc);
580}
581
582/// Get the FileID for \p SourceFile if it exists. Otherwise, create a
583/// new FileID for the \p SourceFile.
584FileID
585SourceManager::getOrCreateFileID(FileEntryRef SourceFile,
586 SrcMgr::CharacteristicKind FileCharacter) {
587 FileID ID = translateFile(SourceFile);
588 return ID.isValid() ? ID : createFileID(SourceFile, IncludePos: SourceLocation(),
589 FileCharacter);
590}
591
592/// createFileID - Create a new FileID for the specified ContentCache and
593/// include position. This works regardless of whether the ContentCache
594/// corresponds to a file or some other input source.
595FileID SourceManager::createFileIDImpl(ContentCache &File, StringRef Filename,
596 SourceLocation IncludePos,
597 SrcMgr::CharacteristicKind FileCharacter,
598 int LoadedID,
599 SourceLocation::UIntTy LoadedOffset) {
600 if (LoadedID < 0) {
601 assert(LoadedID != -1 && "Loading sentinel FileID");
602 unsigned Index = unsigned(-LoadedID) - 2;
603 assert(Index < LoadedSLocEntryTable.size() && "FileID out of range");
604 assert(!SLocEntryLoaded[Index] && "FileID already loaded");
605 LoadedSLocEntryTable[Index] = SLocEntry::get(
606 Offset: LoadedOffset, FI: FileInfo::get(IL: IncludePos, Con&: File, FileCharacter, Filename));
607 SLocEntryLoaded[Index] = SLocEntryOffsetLoaded[Index] = true;
608 return FileID::get(V: LoadedID);
609 }
610 unsigned FileSize = File.getSize();
611 llvm::ErrorOr<bool> NeedConversion =
612 llvm::needConversion(FileName: Filename.str().c_str());
613 if (NeedConversion && *NeedConversion) {
614 // Buffer size may increase due to potential z/OS EBCDIC to UTF-8
615 // conversion.
616 if (std::optional<llvm::MemoryBufferRef> Buffer =
617 File.getBufferOrNone(Diag, FM&: getFileManager())) {
618 unsigned BufSize = Buffer->getBufferSize();
619 if (BufSize > FileSize) {
620 if (File.ContentsEntry.has_value())
621 File.ContentsEntry->updateFileEntryBufferSize(BufferSize: BufSize);
622 FileSize = BufSize;
623 }
624 }
625 }
626 if (!(NextLocalOffset + FileSize + 1 > NextLocalOffset &&
627 NextLocalOffset + FileSize + 1 <= CurrentLoadedOffset)) {
628 Diag.Report(Loc: IncludePos, DiagID: diag::err_sloc_space_too_large);
629 noteSLocAddressSpaceUsage(Diag);
630 return FileID();
631 }
632 assert(LocalSLocEntryTable.size() == LocalLocOffsetTable.size());
633 LocalSLocEntryTable.push_back(
634 Elt: SLocEntry::get(Offset: NextLocalOffset,
635 FI: FileInfo::get(IL: IncludePos, Con&: File, FileCharacter, Filename)));
636 LocalLocOffsetTable.push_back(Elt: NextLocalOffset);
637 LastLookupStartOffset = NextLocalOffset;
638 // We do a +1 here because we want a SourceLocation that means "the end of the
639 // file", e.g. for the "no newline at the end of the file" diagnostic.
640 NextLocalOffset += FileSize + 1;
641 LastLookupEndOffset = NextLocalOffset;
642 updateSlocUsageStats();
643
644 // Set LastFileIDLookup to the newly created file. The next getFileID call is
645 // almost guaranteed to be from that file.
646 FileID FID = FileID::get(V: LocalSLocEntryTable.size()-1);
647 return LastFileIDLookup = FID;
648}
649
650SourceLocation SourceManager::createMacroArgExpansionLoc(
651 SourceLocation SpellingLoc, SourceLocation ExpansionLoc, unsigned Length) {
652 ExpansionInfo Info = ExpansionInfo::createForMacroArg(SpellingLoc,
653 ExpansionLoc);
654 return createExpansionLocImpl(Expansion: Info, Length);
655}
656
657SourceLocation SourceManager::createExpansionLoc(
658 SourceLocation SpellingLoc, SourceLocation ExpansionLocStart,
659 SourceLocation ExpansionLocEnd, unsigned Length,
660 bool ExpansionIsTokenRange, int LoadedID,
661 SourceLocation::UIntTy LoadedOffset) {
662 ExpansionInfo Info = ExpansionInfo::create(
663 SpellingLoc, Start: ExpansionLocStart, End: ExpansionLocEnd, ExpansionIsTokenRange);
664 return createExpansionLocImpl(Expansion: Info, Length, LoadedID, LoadedOffset);
665}
666
667SourceLocation SourceManager::createTokenSplitLoc(SourceLocation Spelling,
668 SourceLocation TokenStart,
669 SourceLocation TokenEnd) {
670 assert(getFileID(TokenStart) == getFileID(TokenEnd) &&
671 "token spans multiple files");
672 return createExpansionLocImpl(
673 Expansion: ExpansionInfo::createForTokenSplit(SpellingLoc: Spelling, Start: TokenStart, End: TokenEnd),
674 Length: TokenEnd.getOffset() - TokenStart.getOffset());
675}
676
677SourceLocation
678SourceManager::createExpansionLocImpl(const ExpansionInfo &Info,
679 unsigned Length, int LoadedID,
680 SourceLocation::UIntTy LoadedOffset) {
681 if (LoadedID < 0) {
682 assert(LoadedID != -1 && "Loading sentinel FileID");
683 unsigned Index = unsigned(-LoadedID) - 2;
684 assert(Index < LoadedSLocEntryTable.size() && "FileID out of range");
685 assert(!SLocEntryLoaded[Index] && "FileID already loaded");
686 LoadedSLocEntryTable[Index] = SLocEntry::get(Offset: LoadedOffset, Expansion: Info);
687 SLocEntryLoaded[Index] = SLocEntryOffsetLoaded[Index] = true;
688 return SourceLocation::getMacroLoc(ID: LoadedOffset);
689 }
690 assert(LocalSLocEntryTable.size() == LocalLocOffsetTable.size());
691 LocalSLocEntryTable.push_back(Elt: SLocEntry::get(Offset: NextLocalOffset, Expansion: Info));
692 LocalLocOffsetTable.push_back(Elt: NextLocalOffset);
693 if (NextLocalOffset + Length + 1 <= NextLocalOffset ||
694 NextLocalOffset + Length + 1 > CurrentLoadedOffset) {
695 Diag.Report(DiagID: diag::err_sloc_space_too_large);
696 // FIXME: call `noteSLocAddressSpaceUsage` to report details to users and
697 // use a source location from `Info` to point at an error.
698 // Currently, both cause Clang to run indefinitely, this needs to be fixed.
699 // FIXME: return an error instead of crashing. Returning invalid source
700 // locations causes compiler to run indefinitely.
701 llvm::report_fatal_error(reason: "ran out of source locations");
702 }
703 // See createFileID for that +1.
704 NextLocalOffset += Length + 1;
705 updateSlocUsageStats();
706 return SourceLocation::getMacroLoc(ID: NextLocalOffset - (Length + 1));
707}
708
709std::optional<llvm::MemoryBufferRef>
710SourceManager::getMemoryBufferForFileOrNone(FileEntryRef File) {
711 SrcMgr::ContentCache &IR = getOrCreateContentCache(FileEnt: File);
712 return IR.getBufferOrNone(Diag, FM&: getFileManager(), Loc: SourceLocation());
713}
714
715void SourceManager::overrideFileContents(
716 FileEntryRef SourceFile, std::unique_ptr<llvm::MemoryBuffer> Buffer) {
717 SrcMgr::ContentCache &IR = getOrCreateContentCache(FileEnt: SourceFile);
718
719 IR.setBuffer(std::move(Buffer));
720 IR.BufferOverridden = true;
721
722 getOverriddenFilesInfo().OverriddenFilesWithBuffer.insert(V: SourceFile);
723}
724
725void SourceManager::overrideFileContents(const FileEntry *SourceFile,
726 FileEntryRef NewFile) {
727 assert(SourceFile->getSize() == NewFile.getSize() &&
728 "Different sizes, use the FileManager to create a virtual file with "
729 "the correct size");
730 assert(FileInfos.find_as(SourceFile) == FileInfos.end() &&
731 "This function should be called at the initialization stage, before "
732 "any parsing occurs.");
733 // FileEntryRef is not default-constructible.
734 auto Pair = getOverriddenFilesInfo().OverriddenFiles.insert(
735 KV: std::make_pair(x&: SourceFile, y&: NewFile));
736 if (!Pair.second)
737 Pair.first->second = NewFile;
738}
739
740OptionalFileEntryRef
741SourceManager::bypassFileContentsOverride(FileEntryRef File) {
742 assert(isFileOverridden(&File.getFileEntry()));
743 OptionalFileEntryRef BypassFile = FileMgr.getBypassFile(VFE: File);
744
745 // If the file can't be found in the FS, give up.
746 if (!BypassFile)
747 return std::nullopt;
748
749 (void)getOrCreateContentCache(FileEnt: *BypassFile);
750 return BypassFile;
751}
752
753void SourceManager::setFileIsTransient(FileEntryRef File) {
754 getOrCreateContentCache(FileEnt: File).IsTransient = true;
755}
756
757std::optional<StringRef>
758SourceManager::getNonBuiltinFilenameForID(FileID FID) const {
759 if (const SrcMgr::SLocEntry *Entry = getSLocEntryForFile(FID))
760 if (Entry->getFile().getContentCache().OrigEntry)
761 return Entry->getFile().getName();
762 return std::nullopt;
763}
764
765StringRef SourceManager::getBufferData(FileID FID, bool *Invalid) const {
766 auto B = getBufferDataOrNone(FID);
767 if (Invalid)
768 *Invalid = !B;
769 return B ? *B : "<<<<<INVALID SOURCE LOCATION>>>>>";
770}
771
772std::optional<StringRef>
773SourceManager::getBufferDataIfLoaded(FileID FID) const {
774 if (const SrcMgr::SLocEntry *Entry = getSLocEntryForFile(FID))
775 return Entry->getFile().getContentCache().getBufferDataIfLoaded();
776 return std::nullopt;
777}
778
779std::optional<StringRef> SourceManager::getBufferDataOrNone(FileID FID) const {
780 if (const SrcMgr::SLocEntry *Entry = getSLocEntryForFile(FID))
781 if (auto B = Entry->getFile().getContentCache().getBufferOrNone(
782 Diag, FM&: getFileManager(), Loc: SourceLocation()))
783 return B->getBuffer();
784 return std::nullopt;
785}
786
787//===----------------------------------------------------------------------===//
788// SourceLocation manipulation methods.
789//===----------------------------------------------------------------------===//
790
791/// Return the FileID for a SourceLocation.
792///
793/// This is the cache-miss path of getFileID. Not as hot as that function, but
794/// still very important. It is responsible for finding the entry in the
795/// SLocEntry tables that contains the specified location.
796FileID SourceManager::getFileIDSlow(SourceLocation::UIntTy SLocOffset) const {
797 if (!SLocOffset)
798 return FileID::get(V: 0);
799
800 // Now it is time to search for the correct file. See where the SLocOffset
801 // sits in the global view and consult local or loaded buffers for it.
802 if (SLocOffset < NextLocalOffset)
803 return getFileIDLocal(SLocOffset);
804 return getFileIDLoaded(SLocOffset);
805}
806
807/// Return the FileID for a SourceLocation with a low offset.
808///
809/// This function knows that the SourceLocation is in a local buffer, not a
810/// loaded one.
811FileID SourceManager::getFileIDLocal(SourceLocation::UIntTy SLocOffset) const {
812 assert(SLocOffset < NextLocalOffset && "Bad function choice");
813 assert(SLocOffset >= LocalSLocEntryTable[0].getOffset() && SLocOffset > 0 &&
814 "Invalid SLocOffset");
815 assert(LocalSLocEntryTable.size() == LocalLocOffsetTable.size());
816 assert(LastFileIDLookup.ID >= 0 && "Only cache local file sloc entry");
817
818 // After the first and second level caches, I see two common sorts of
819 // behavior: 1) a lot of searched FileID's are "near" the cached file
820 // location or are "near" the cached expansion location. 2) others are just
821 // completely random and may be a very long way away.
822 //
823 // To handle this, we do a linear search for up to 8 steps to catch #1 quickly
824 // then we fall back to a less cache efficient, but more scalable, binary
825 // search to find the location.
826
827 // See if this is near the file point - worst case we start scanning from the
828 // most newly created FileID.
829
830 // LessIndex - This is the lower bound of the range that we're searching.
831 // We know that the offset corresponding to the FileID is less than
832 // SLocOffset.
833 unsigned LessIndex = 0;
834 // upper bound of the search range.
835 unsigned GreaterIndex = LocalLocOffsetTable.size();
836 // Use the LastFileIDLookup to prune the search space.
837 if (LastLookupStartOffset < SLocOffset)
838 LessIndex = LastFileIDLookup.ID;
839 else
840 GreaterIndex = LastFileIDLookup.ID;
841
842 // Find the FileID that contains this.
843 unsigned NumProbes = 0;
844 while (true) {
845 --GreaterIndex;
846 assert(GreaterIndex < LocalLocOffsetTable.size());
847 if (LocalLocOffsetTable[GreaterIndex] <= SLocOffset) {
848 FileID Res = FileID::get(V: int(GreaterIndex));
849 // Remember it. We have good locality across FileID lookups.
850 LastFileIDLookup = Res;
851 LastLookupStartOffset = LocalLocOffsetTable[GreaterIndex];
852 LastLookupEndOffset =
853 GreaterIndex + 1 >= LocalLocOffsetTable.size()
854 ? NextLocalOffset
855 : LocalLocOffsetTable[GreaterIndex + 1];
856 NumLinearScans += NumProbes + 1;
857 return Res;
858 }
859 if (++NumProbes == 8)
860 break;
861 }
862
863 while (LessIndex < GreaterIndex) {
864 ++NumBinaryProbes;
865
866 unsigned MiddleIndex = LessIndex + (GreaterIndex - LessIndex) / 2;
867 if (LocalLocOffsetTable[MiddleIndex] <= SLocOffset)
868 LessIndex = MiddleIndex + 1;
869 else
870 GreaterIndex = MiddleIndex;
871 }
872
873 // At this point, LessIndex is the index of the *first element greater than*
874 // SLocOffset. The element we are actually looking for is the one immediately
875 // before it.
876 LastLookupStartOffset = LocalLocOffsetTable[LessIndex - 1];
877 LastLookupEndOffset = LocalLocOffsetTable[LessIndex];
878 return LastFileIDLookup = FileID::get(V: LessIndex - 1);
879}
880
881/// Return the FileID for a SourceLocation with a high offset.
882///
883/// This function knows that the SourceLocation is in a loaded buffer, not a
884/// local one.
885FileID SourceManager::getFileIDLoaded(SourceLocation::UIntTy SLocOffset) const {
886 if (SLocOffset < CurrentLoadedOffset) {
887 assert(0 && "Invalid SLocOffset or bad function choice");
888 return FileID();
889 }
890
891 return FileID::get(V: ExternalSLocEntries->getSLocEntryID(SLocOffset));
892}
893
894SourceLocation SourceManager::
895getExpansionLocSlowCase(SourceLocation Loc) const {
896 do {
897 // Note: If Loc indicates an offset into a token that came from a macro
898 // expansion (e.g. the 5th character of the token) we do not want to add
899 // this offset when going to the expansion location. The expansion
900 // location is the macro invocation, which the offset has nothing to do
901 // with. This is unlike when we get the spelling loc, because the offset
902 // directly correspond to the token whose spelling we're inspecting.
903 Loc = getSLocEntry(FID: getFileID(SpellingLoc: Loc)).getExpansion().getExpansionLocStart();
904 } while (!Loc.isFileID());
905
906 return Loc;
907}
908
909SourceLocation SourceManager::getSpellingLocSlowCase(SourceLocation Loc) const {
910 do {
911 FileIDAndOffset LocInfo = getDecomposedLoc(Loc);
912 Loc = getSLocEntry(FID: LocInfo.first).getExpansion().getSpellingLoc();
913 Loc = Loc.getLocWithOffset(Offset: LocInfo.second);
914 } while (!Loc.isFileID());
915 return Loc;
916}
917
918SourceLocation SourceManager::getFileLocSlowCase(SourceLocation Loc) const {
919 do {
920 if (isMacroArgExpansion(Loc))
921 Loc = getImmediateSpellingLoc(Loc);
922 else
923 Loc = getImmediateExpansionRange(Loc).getBegin();
924 } while (!Loc.isFileID());
925 return Loc;
926}
927
928FileIDAndOffset SourceManager::getDecomposedExpansionLocSlowCase(
929 const SrcMgr::SLocEntry *E) const {
930 // If this is an expansion record, walk through all the expansion points.
931 FileID FID;
932 SourceLocation Loc;
933 unsigned Offset;
934 do {
935 Loc = E->getExpansion().getExpansionLocStart();
936
937 FID = getFileID(SpellingLoc: Loc);
938 E = &getSLocEntry(FID);
939 Offset = Loc.getOffset()-E->getOffset();
940 } while (!Loc.isFileID());
941
942 return std::make_pair(x&: FID, y&: Offset);
943}
944
945FileIDAndOffset
946SourceManager::getDecomposedSpellingLocSlowCase(const SrcMgr::SLocEntry *E,
947 unsigned Offset) const {
948 // If this is an expansion record, walk through all the expansion points.
949 FileID FID;
950 SourceLocation Loc;
951 do {
952 Loc = E->getExpansion().getSpellingLoc();
953 Loc = Loc.getLocWithOffset(Offset);
954
955 FID = getFileID(SpellingLoc: Loc);
956 E = &getSLocEntry(FID);
957 Offset = Loc.getOffset()-E->getOffset();
958 } while (!Loc.isFileID());
959
960 return std::make_pair(x&: FID, y&: Offset);
961}
962
963/// getImmediateSpellingLoc - Given a SourceLocation object, return the
964/// spelling location referenced by the ID. This is the first level down
965/// towards the place where the characters that make up the lexed token can be
966/// found. This should not generally be used by clients.
967SourceLocation SourceManager::getImmediateSpellingLoc(SourceLocation Loc) const{
968 if (Loc.isFileID()) return Loc;
969 FileIDAndOffset LocInfo = getDecomposedLoc(Loc);
970 Loc = getSLocEntry(FID: LocInfo.first).getExpansion().getSpellingLoc();
971 return Loc.getLocWithOffset(Offset: LocInfo.second);
972}
973
974/// Return the filename of the file containing a SourceLocation.
975StringRef SourceManager::getFilename(SourceLocation SpellingLoc) const {
976 if (OptionalFileEntryRef F = getFileEntryRefForID(FID: getFileID(SpellingLoc)))
977 return F->getName();
978 return StringRef();
979}
980
981/// getImmediateExpansionRange - Loc is required to be an expansion location.
982/// Return the start/end of the expansion information.
983CharSourceRange
984SourceManager::getImmediateExpansionRange(SourceLocation Loc) const {
985 assert(Loc.isMacroID() && "Not a macro expansion loc!");
986 const ExpansionInfo &Expansion = getSLocEntry(FID: getFileID(SpellingLoc: Loc)).getExpansion();
987 return Expansion.getExpansionLocRange();
988}
989
990SourceLocation SourceManager::getTopMacroCallerLoc(SourceLocation Loc) const {
991 while (isMacroArgExpansion(Loc))
992 Loc = getImmediateSpellingLoc(Loc);
993 return Loc;
994}
995
996/// getExpansionRange - Given a SourceLocation object, return the range of
997/// tokens covered by the expansion in the ultimate file.
998CharSourceRange SourceManager::getExpansionRange(SourceLocation Loc) const {
999 if (Loc.isFileID())
1000 return CharSourceRange(SourceRange(Loc, Loc), true);
1001
1002 CharSourceRange Res = getImmediateExpansionRange(Loc);
1003
1004 // Fully resolve the start and end locations to their ultimate expansion
1005 // points.
1006 while (!Res.getBegin().isFileID())
1007 Res.setBegin(getImmediateExpansionRange(Loc: Res.getBegin()).getBegin());
1008 while (!Res.getEnd().isFileID()) {
1009 CharSourceRange EndRange = getImmediateExpansionRange(Loc: Res.getEnd());
1010 Res.setEnd(EndRange.getEnd());
1011 Res.setTokenRange(EndRange.isTokenRange());
1012 }
1013 return Res;
1014}
1015
1016bool SourceManager::isMacroArgExpansion(SourceLocation Loc,
1017 SourceLocation *StartLoc) const {
1018 if (!Loc.isMacroID()) return false;
1019
1020 FileID FID = getFileID(SpellingLoc: Loc);
1021 const SrcMgr::ExpansionInfo &Expansion = getSLocEntry(FID).getExpansion();
1022 if (!Expansion.isMacroArgExpansion()) return false;
1023
1024 if (StartLoc)
1025 *StartLoc = Expansion.getExpansionLocStart();
1026 return true;
1027}
1028
1029bool SourceManager::isMacroBodyExpansion(SourceLocation Loc) const {
1030 if (!Loc.isMacroID()) return false;
1031
1032 FileID FID = getFileID(SpellingLoc: Loc);
1033 const SrcMgr::ExpansionInfo &Expansion = getSLocEntry(FID).getExpansion();
1034 return Expansion.isMacroBodyExpansion();
1035}
1036
1037bool SourceManager::isAtStartOfImmediateMacroExpansion(SourceLocation Loc,
1038 SourceLocation *MacroBegin) const {
1039 assert(Loc.isValid() && Loc.isMacroID() && "Expected a valid macro loc");
1040
1041 FileIDAndOffset DecompLoc = getDecomposedLoc(Loc);
1042 if (DecompLoc.second > 0)
1043 return false; // Does not point at the start of expansion range.
1044
1045 bool Invalid = false;
1046 const SrcMgr::ExpansionInfo &ExpInfo =
1047 getSLocEntry(FID: DecompLoc.first, Invalid: &Invalid).getExpansion();
1048 if (Invalid)
1049 return false;
1050 SourceLocation ExpLoc = ExpInfo.getExpansionLocStart();
1051
1052 if (ExpInfo.isMacroArgExpansion()) {
1053 // For macro argument expansions, check if the previous FileID is part of
1054 // the same argument expansion, in which case this Loc is not at the
1055 // beginning of the expansion.
1056 FileID PrevFID = getPreviousFileID(FID: DecompLoc.first);
1057 if (!PrevFID.isInvalid()) {
1058 const SrcMgr::SLocEntry &PrevEntry = getSLocEntry(FID: PrevFID, Invalid: &Invalid);
1059 if (Invalid)
1060 return false;
1061 if (PrevEntry.isExpansion() &&
1062 PrevEntry.getExpansion().getExpansionLocStart() == ExpLoc)
1063 return false;
1064 }
1065 }
1066
1067 if (MacroBegin)
1068 *MacroBegin = ExpLoc;
1069 return true;
1070}
1071
1072bool SourceManager::isAtEndOfImmediateMacroExpansion(SourceLocation Loc,
1073 SourceLocation *MacroEnd) const {
1074 assert(Loc.isValid() && Loc.isMacroID() && "Expected a valid macro loc");
1075
1076 FileID FID = getFileID(SpellingLoc: Loc);
1077 SourceLocation NextLoc = Loc.getLocWithOffset(Offset: 1);
1078 if (isInFileID(Loc: NextLoc, FID))
1079 return false; // Does not point at the end of expansion range.
1080
1081 bool Invalid = false;
1082 const SrcMgr::ExpansionInfo &ExpInfo =
1083 getSLocEntry(FID, Invalid: &Invalid).getExpansion();
1084 if (Invalid)
1085 return false;
1086
1087 if (ExpInfo.isMacroArgExpansion()) {
1088 // For macro argument expansions, check if the next FileID is part of the
1089 // same argument expansion, in which case this Loc is not at the end of the
1090 // expansion.
1091 FileID NextFID = getNextFileID(FID);
1092 if (!NextFID.isInvalid()) {
1093 const SrcMgr::SLocEntry &NextEntry = getSLocEntry(FID: NextFID, Invalid: &Invalid);
1094 if (Invalid)
1095 return false;
1096 if (NextEntry.isExpansion() &&
1097 NextEntry.getExpansion().getExpansionLocStart() ==
1098 ExpInfo.getExpansionLocStart())
1099 return false;
1100 }
1101 }
1102
1103 if (MacroEnd)
1104 *MacroEnd = ExpInfo.getExpansionLocEnd();
1105 return true;
1106}
1107
1108//===----------------------------------------------------------------------===//
1109// Queries about the code at a SourceLocation.
1110//===----------------------------------------------------------------------===//
1111
1112/// getCharacterData - Return a pointer to the start of the specified location
1113/// in the appropriate MemoryBuffer.
1114const char *SourceManager::getCharacterData(SourceLocation SL,
1115 bool *Invalid) const {
1116 // Note that this is a hot function in the getSpelling() path, which is
1117 // heavily used by -E mode.
1118 FileIDAndOffset LocInfo = getDecomposedSpellingLoc(Loc: SL);
1119
1120 // Note that calling 'getBuffer()' may lazily page in a source file.
1121 bool CharDataInvalid = false;
1122 const SLocEntry &Entry = getSLocEntry(FID: LocInfo.first, Invalid: &CharDataInvalid);
1123 if (CharDataInvalid || !Entry.isFile()) {
1124 if (Invalid)
1125 *Invalid = true;
1126
1127 return "<<<<INVALID BUFFER>>>>";
1128 }
1129 std::optional<llvm::MemoryBufferRef> Buffer =
1130 Entry.getFile().getContentCache().getBufferOrNone(Diag, FM&: getFileManager(),
1131 Loc: SourceLocation());
1132 if (Invalid)
1133 *Invalid = !Buffer;
1134 return Buffer ? Buffer->getBufferStart() + LocInfo.second
1135 : "<<<<INVALID BUFFER>>>>";
1136}
1137
1138/// getColumnNumber - Return the column # for the specified file position.
1139/// this is significantly cheaper to compute than the line number.
1140unsigned SourceManager::getColumnNumber(FileID FID, unsigned FilePos,
1141 bool *Invalid) const {
1142 std::optional<llvm::MemoryBufferRef> MemBuf = getBufferOrNone(FID);
1143 if (Invalid)
1144 *Invalid = !MemBuf;
1145
1146 if (!MemBuf)
1147 return 1;
1148
1149 // It is okay to request a position just past the end of the buffer.
1150 if (FilePos > MemBuf->getBufferSize()) {
1151 if (Invalid)
1152 *Invalid = true;
1153 return 1;
1154 }
1155
1156 const char *Buf = MemBuf->getBufferStart();
1157 // See if we just calculated the line number for this FilePos and can use
1158 // that to lookup the start of the line instead of searching for it.
1159 if (LastLineNoFileIDQuery == FID && LastLineNoContentCache->SourceLineCache &&
1160 LastLineNoResult < LastLineNoContentCache->SourceLineCache.size()) {
1161 const unsigned *SourceLineCache =
1162 LastLineNoContentCache->SourceLineCache.begin();
1163 unsigned LineStart = SourceLineCache[LastLineNoResult - 1];
1164 unsigned LineEnd = SourceLineCache[LastLineNoResult];
1165 if (FilePos >= LineStart && FilePos < LineEnd) {
1166 // LineEnd is the LineStart of the next line.
1167 // A line ends with separator LF or CR+LF on Windows.
1168 // FilePos might point to the last separator,
1169 // but we need a column number at most 1 + the last column.
1170 if (FilePos + 1 == LineEnd && FilePos > LineStart) {
1171 if (Buf[FilePos - 1] == '\r' || Buf[FilePos - 1] == '\n')
1172 --FilePos;
1173 }
1174 return FilePos - LineStart + 1;
1175 }
1176 }
1177
1178 unsigned LineStart = FilePos;
1179 while (LineStart && Buf[LineStart-1] != '\n' && Buf[LineStart-1] != '\r')
1180 --LineStart;
1181 return FilePos-LineStart+1;
1182}
1183
1184// isInvalid - Return the result of calling loc.isInvalid(), and
1185// if Invalid is not null, set its value to same.
1186template<typename LocType>
1187static bool isInvalid(LocType Loc, bool *Invalid) {
1188 bool MyInvalid = Loc.isInvalid();
1189 if (Invalid)
1190 *Invalid = MyInvalid;
1191 return MyInvalid;
1192}
1193
1194unsigned SourceManager::getSpellingColumnNumber(SourceLocation Loc,
1195 bool *Invalid) const {
1196 if (isInvalid(Loc, Invalid)) return 0;
1197 FileIDAndOffset LocInfo = getDecomposedSpellingLoc(Loc);
1198 return getColumnNumber(FID: LocInfo.first, FilePos: LocInfo.second, Invalid);
1199}
1200
1201unsigned SourceManager::getExpansionColumnNumber(SourceLocation Loc,
1202 bool *Invalid) const {
1203 if (isInvalid(Loc, Invalid)) return 0;
1204 FileIDAndOffset LocInfo = getDecomposedExpansionLoc(Loc);
1205 return getColumnNumber(FID: LocInfo.first, FilePos: LocInfo.second, Invalid);
1206}
1207
1208unsigned SourceManager::getPresumedColumnNumber(SourceLocation Loc,
1209 bool *Invalid) const {
1210 PresumedLoc PLoc = getPresumedLoc(Loc);
1211 if (isInvalid(Loc: PLoc, Invalid)) return 0;
1212 return PLoc.getColumn();
1213}
1214
1215// Check if multi-byte word x has bytes between m and n, included. This may also
1216// catch bytes equal to n + 1.
1217// The returned value holds a 0x80 at each byte position that holds a match.
1218// see http://graphics.stanford.edu/~seander/bithacks.html#HasBetweenInWord
1219template <class T>
1220static constexpr inline T likelyhasbetween(T x, unsigned char m,
1221 unsigned char n) {
1222 return ((x - ~static_cast<T>(0) / 255 * (n + 1)) & ~x &
1223 ((x & ~static_cast<T>(0) / 255 * 127) +
1224 (~static_cast<T>(0) / 255 * (127 - (m - 1))))) &
1225 ~static_cast<T>(0) / 255 * 128;
1226}
1227
1228LineOffsetMapping LineOffsetMapping::get(llvm::MemoryBufferRef Buffer,
1229 llvm::BumpPtrAllocator &Alloc) {
1230
1231 // Find the file offsets of all of the *physical* source lines. This does
1232 // not look at trigraphs, escaped newlines, or anything else tricky.
1233 SmallVector<unsigned, 256> LineOffsets;
1234
1235 // Line #1 starts at char 0.
1236 LineOffsets.push_back(Elt: 0);
1237
1238 const unsigned char *Start = (const unsigned char *)Buffer.getBufferStart();
1239 const unsigned char *End = (const unsigned char *)Buffer.getBufferEnd();
1240 const unsigned char *Buf = Start;
1241
1242 uint64_t Word;
1243
1244 // scan sizeof(Word) bytes at a time for new lines.
1245 // This is much faster than scanning each byte independently.
1246 if ((unsigned long)(End - Start) > sizeof(Word)) {
1247 do {
1248 Word = llvm::support::endian::read64(P: Buf, E: llvm::endianness::little);
1249 // no new line => jump over sizeof(Word) bytes.
1250 auto Mask = likelyhasbetween(x: Word, m: '\n', n: '\r');
1251 if (!Mask) {
1252 Buf += sizeof(Word);
1253 continue;
1254 }
1255
1256 // At that point, Mask contains 0x80 set at each byte that holds a value
1257 // in [\n, \r + 1 [
1258
1259 // Scan for the next newline - it's very likely there's one.
1260 unsigned N = llvm::countr_zero(Val: Mask) - 7; // -7 because 0x80 is the marker
1261 Word >>= N;
1262 Buf += N / 8 + 1;
1263 unsigned char Byte = Word;
1264 switch (Byte) {
1265 case '\r':
1266 // If this is \r\n, skip both characters.
1267 if (*Buf == '\n') {
1268 ++Buf;
1269 }
1270 [[fallthrough]];
1271 case '\n':
1272 LineOffsets.push_back(Elt: Buf - Start);
1273 };
1274 } while (Buf < End - sizeof(Word) - 1);
1275 }
1276
1277 // Handle tail using a regular check.
1278 while (Buf < End) {
1279 if (*Buf == '\n') {
1280 LineOffsets.push_back(Elt: Buf - Start + 1);
1281 } else if (*Buf == '\r') {
1282 // If this is \r\n, skip both characters.
1283 if (Buf + 1 < End && Buf[1] == '\n') {
1284 ++Buf;
1285 }
1286 LineOffsets.push_back(Elt: Buf - Start + 1);
1287 }
1288 ++Buf;
1289 }
1290
1291 return LineOffsetMapping(LineOffsets, Alloc);
1292}
1293
1294LineOffsetMapping::LineOffsetMapping(ArrayRef<unsigned> LineOffsets,
1295 llvm::BumpPtrAllocator &Alloc)
1296 : Storage(Alloc.Allocate<unsigned>(Num: LineOffsets.size() + 1)) {
1297 Storage[0] = LineOffsets.size();
1298 std::copy(first: LineOffsets.begin(), last: LineOffsets.end(), result: Storage + 1);
1299}
1300
1301/// getLineNumber - Given a SourceLocation, return the spelling line number
1302/// for the position indicated. This requires building and caching a table of
1303/// line offsets for the MemoryBuffer, so this is not cheap: use only when
1304/// about to emit a diagnostic.
1305unsigned SourceManager::getLineNumber(FileID FID, unsigned FilePos,
1306 bool *Invalid) const {
1307 if (FID.isInvalid()) {
1308 if (Invalid)
1309 *Invalid = true;
1310 return 1;
1311 }
1312
1313 const ContentCache *Content;
1314 if (LastLineNoFileIDQuery == FID)
1315 Content = LastLineNoContentCache;
1316 else {
1317 bool MyInvalid = false;
1318 const SLocEntry &Entry = getSLocEntry(FID, Invalid: &MyInvalid);
1319 if (MyInvalid || !Entry.isFile()) {
1320 if (Invalid)
1321 *Invalid = true;
1322 return 1;
1323 }
1324
1325 Content = &Entry.getFile().getContentCache();
1326 }
1327
1328 // If this is the first use of line information for this buffer, compute the
1329 // SourceLineCache for it on demand.
1330 if (!Content->SourceLineCache) {
1331 std::optional<llvm::MemoryBufferRef> Buffer =
1332 Content->getBufferOrNone(Diag, FM&: getFileManager(), Loc: SourceLocation());
1333 if (Invalid)
1334 *Invalid = !Buffer;
1335 if (!Buffer)
1336 return 1;
1337
1338 Content->SourceLineCache =
1339 LineOffsetMapping::get(Buffer: *Buffer, Alloc&: ContentCacheAlloc);
1340 } else if (Invalid)
1341 *Invalid = false;
1342
1343 // Okay, we know we have a line number table. Do a binary search to find the
1344 // line number that this character position lands on.
1345 const unsigned *SourceLineCache = Content->SourceLineCache.begin();
1346 const unsigned *SourceLineCacheStart = SourceLineCache;
1347 const unsigned *SourceLineCacheEnd = Content->SourceLineCache.end();
1348
1349 unsigned QueriedFilePos = FilePos+1;
1350
1351 // FIXME: I would like to be convinced that this code is worth being as
1352 // complicated as it is, binary search isn't that slow.
1353 //
1354 // If it is worth being optimized, then in my opinion it could be more
1355 // performant, simpler, and more obviously correct by just "galloping" outward
1356 // from the queried file position. In fact, this could be incorporated into a
1357 // generic algorithm such as lower_bound_with_hint.
1358 //
1359 // If someone gives me a test case where this matters, and I will do it! - DWD
1360
1361 // If the previous query was to the same file, we know both the file pos from
1362 // that query and the line number returned. This allows us to narrow the
1363 // search space from the entire file to something near the match.
1364 if (LastLineNoFileIDQuery == FID) {
1365 if (QueriedFilePos >= LastLineNoFilePos) {
1366 // FIXME: Potential overflow?
1367 SourceLineCache = SourceLineCache+LastLineNoResult-1;
1368
1369 // The query is likely to be nearby the previous one. Here we check to
1370 // see if it is within 5, 10 or 20 lines. It can be far away in cases
1371 // where big comment blocks and vertical whitespace eat up lines but
1372 // contribute no tokens.
1373 if (SourceLineCache+5 < SourceLineCacheEnd) {
1374 if (SourceLineCache[5] > QueriedFilePos)
1375 SourceLineCacheEnd = SourceLineCache+5;
1376 else if (SourceLineCache+10 < SourceLineCacheEnd) {
1377 if (SourceLineCache[10] > QueriedFilePos)
1378 SourceLineCacheEnd = SourceLineCache+10;
1379 else if (SourceLineCache+20 < SourceLineCacheEnd) {
1380 if (SourceLineCache[20] > QueriedFilePos)
1381 SourceLineCacheEnd = SourceLineCache+20;
1382 }
1383 }
1384 }
1385 } else {
1386 if (LastLineNoResult < Content->SourceLineCache.size())
1387 SourceLineCacheEnd = SourceLineCache+LastLineNoResult+1;
1388 }
1389 }
1390
1391 const unsigned *Pos =
1392 std::lower_bound(first: SourceLineCache, last: SourceLineCacheEnd, val: QueriedFilePos);
1393 unsigned LineNo = Pos-SourceLineCacheStart;
1394
1395 LastLineNoFileIDQuery = FID;
1396 LastLineNoContentCache = Content;
1397 LastLineNoFilePos = QueriedFilePos;
1398 LastLineNoResult = LineNo;
1399 return LineNo;
1400}
1401
1402unsigned SourceManager::getSpellingLineNumber(SourceLocation Loc,
1403 bool *Invalid) const {
1404 if (isInvalid(Loc, Invalid)) return 0;
1405 FileIDAndOffset LocInfo = getDecomposedSpellingLoc(Loc);
1406 return getLineNumber(FID: LocInfo.first, FilePos: LocInfo.second);
1407}
1408unsigned SourceManager::getExpansionLineNumber(SourceLocation Loc,
1409 bool *Invalid) const {
1410 if (isInvalid(Loc, Invalid)) return 0;
1411 FileIDAndOffset LocInfo = getDecomposedExpansionLoc(Loc);
1412 return getLineNumber(FID: LocInfo.first, FilePos: LocInfo.second);
1413}
1414unsigned SourceManager::getPresumedLineNumber(SourceLocation Loc,
1415 bool *Invalid) const {
1416 PresumedLoc PLoc = getPresumedLoc(Loc);
1417 if (isInvalid(Loc: PLoc, Invalid)) return 0;
1418 return PLoc.getLine();
1419}
1420
1421/// getFileCharacteristic - return the file characteristic of the specified
1422/// source location, indicating whether this is a normal file, a system
1423/// header, or an "implicit extern C" system header.
1424///
1425/// This state can be modified with flags on GNU linemarker directives like:
1426/// # 4 "foo.h" 3
1427/// which changes all source locations in the current file after that to be
1428/// considered to be from a system header.
1429SrcMgr::CharacteristicKind
1430SourceManager::getFileCharacteristic(SourceLocation Loc) const {
1431 assert(Loc.isValid() && "Can't get file characteristic of invalid loc!");
1432 FileIDAndOffset LocInfo = getDecomposedExpansionLoc(Loc);
1433 const SLocEntry *SEntry = getSLocEntryForFile(FID: LocInfo.first);
1434 if (!SEntry)
1435 return C_User;
1436
1437 const SrcMgr::FileInfo &FI = SEntry->getFile();
1438
1439 // If there are no #line directives in this file, just return the whole-file
1440 // state.
1441 if (!FI.hasLineDirectives())
1442 return FI.getFileCharacteristic();
1443
1444 assert(LineTable && "Can't have linetable entries without a LineTable!");
1445 // See if there is a #line directive before the location.
1446 const LineEntry *Entry =
1447 LineTable->FindNearestLineEntry(FID: LocInfo.first, Offset: LocInfo.second);
1448
1449 // If this is before the first line marker, use the file characteristic.
1450 if (!Entry)
1451 return FI.getFileCharacteristic();
1452
1453 return Entry->FileKind;
1454}
1455
1456/// Return the filename or buffer identifier of the buffer the location is in.
1457/// Note that this name does not respect \#line directives. Use getPresumedLoc
1458/// for normal clients.
1459StringRef SourceManager::getBufferName(SourceLocation Loc,
1460 bool *Invalid) const {
1461 if (isInvalid(Loc, Invalid)) return "<invalid loc>";
1462
1463 auto B = getBufferOrNone(FID: getFileID(SpellingLoc: Loc));
1464 if (Invalid)
1465 *Invalid = !B;
1466 return B ? B->getBufferIdentifier() : "<invalid buffer>";
1467}
1468
1469/// getPresumedLoc - This method returns the "presumed" location of a
1470/// SourceLocation specifies. A "presumed location" can be modified by \#line
1471/// or GNU line marker directives. This provides a view on the data that a
1472/// user should see in diagnostics, for example.
1473///
1474/// Note that a presumed location is always given as the expansion point of an
1475/// expansion location, not at the spelling location.
1476PresumedLoc SourceManager::getPresumedLoc(SourceLocation Loc,
1477 bool UseLineDirectives) const {
1478 if (Loc.isInvalid()) return PresumedLoc();
1479
1480 // Presumed locations are always for expansion points.
1481 FileIDAndOffset LocInfo = getDecomposedExpansionLoc(Loc);
1482
1483 bool Invalid = false;
1484 const SLocEntry &Entry = getSLocEntry(FID: LocInfo.first, Invalid: &Invalid);
1485 if (Invalid || !Entry.isFile())
1486 return PresumedLoc();
1487
1488 const SrcMgr::FileInfo &FI = Entry.getFile();
1489 const SrcMgr::ContentCache *C = &FI.getContentCache();
1490
1491 // To get the source name, first consult the FileEntry (if one exists)
1492 // before the MemBuffer as this will avoid unnecessarily paging in the
1493 // MemBuffer.
1494 FileID FID = LocInfo.first;
1495 StringRef Filename;
1496 if (C->OrigEntry)
1497 Filename = C->OrigEntry->getName();
1498 else if (auto Buffer = C->getBufferOrNone(Diag, FM&: getFileManager()))
1499 Filename = Buffer->getBufferIdentifier();
1500
1501 unsigned LineNo = getLineNumber(FID: LocInfo.first, FilePos: LocInfo.second, Invalid: &Invalid);
1502 if (Invalid)
1503 return PresumedLoc();
1504 unsigned ColNo = getColumnNumber(FID: LocInfo.first, FilePos: LocInfo.second, Invalid: &Invalid);
1505 if (Invalid)
1506 return PresumedLoc();
1507
1508 SourceLocation IncludeLoc = FI.getIncludeLoc();
1509
1510 // If we have #line directives in this file, update and overwrite the physical
1511 // location info if appropriate.
1512 if (UseLineDirectives && FI.hasLineDirectives()) {
1513 assert(LineTable && "Can't have linetable entries without a LineTable!");
1514 // See if there is a #line directive before this. If so, get it.
1515 if (const LineEntry *Entry =
1516 LineTable->FindNearestLineEntry(FID: LocInfo.first, Offset: LocInfo.second)) {
1517 // If the LineEntry indicates a filename, use it.
1518 if (Entry->FilenameID != -1) {
1519 Filename = LineTable->getFilename(ID: Entry->FilenameID);
1520 // The contents of files referenced by #line are not in the
1521 // SourceManager
1522 FID = FileID::get(V: 0);
1523 }
1524
1525 // Use the line number specified by the LineEntry. This line number may
1526 // be multiple lines down from the line entry. Add the difference in
1527 // physical line numbers from the query point and the line marker to the
1528 // total.
1529 unsigned MarkerLineNo = getLineNumber(FID: LocInfo.first, FilePos: Entry->FileOffset);
1530 LineNo = Entry->LineNo + (LineNo-MarkerLineNo-1);
1531
1532 // Note that column numbers are not molested by line markers.
1533
1534 // Handle virtual #include manipulation.
1535 if (Entry->IncludeOffset) {
1536 IncludeLoc = getLocForStartOfFile(FID: LocInfo.first);
1537 IncludeLoc = IncludeLoc.getLocWithOffset(Offset: Entry->IncludeOffset);
1538 }
1539 }
1540 }
1541
1542 return PresumedLoc(Filename.data(), FID, LineNo, ColNo, IncludeLoc);
1543}
1544
1545/// Returns whether the PresumedLoc for a given SourceLocation is
1546/// in the main file.
1547///
1548/// This computes the "presumed" location for a SourceLocation, then checks
1549/// whether it came from a file other than the main file. This is different
1550/// from isWrittenInMainFile() because it takes line marker directives into
1551/// account.
1552bool SourceManager::isInMainFile(SourceLocation Loc) const {
1553 if (Loc.isInvalid()) return false;
1554
1555 // Presumed locations are always for expansion points.
1556 FileIDAndOffset LocInfo = getDecomposedExpansionLoc(Loc);
1557
1558 const SLocEntry *Entry = getSLocEntryForFile(FID: LocInfo.first);
1559 if (!Entry)
1560 return false;
1561
1562 const SrcMgr::FileInfo &FI = Entry->getFile();
1563
1564 // Check if there is a line directive for this location.
1565 if (FI.hasLineDirectives())
1566 if (const LineEntry *Entry =
1567 LineTable->FindNearestLineEntry(FID: LocInfo.first, Offset: LocInfo.second))
1568 if (Entry->IncludeOffset)
1569 return false;
1570
1571 return FI.getIncludeLoc().isInvalid();
1572}
1573
1574/// The size of the SLocEntry that \p FID represents.
1575unsigned SourceManager::getFileIDSize(FileID FID) const {
1576 bool Invalid = false;
1577 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, Invalid: &Invalid);
1578 if (Invalid)
1579 return 0;
1580
1581 int ID = FID.ID;
1582 SourceLocation::UIntTy NextOffset;
1583 if ((ID > 0 && unsigned(ID+1) == local_sloc_entry_size()))
1584 NextOffset = getNextLocalOffset();
1585 else if (ID+1 == -1)
1586 NextOffset = MaxLoadedOffset;
1587 else
1588 NextOffset = getSLocEntry(FID: FileID::get(V: ID+1)).getOffset();
1589
1590 return NextOffset - Entry.getOffset() - 1;
1591}
1592
1593//===----------------------------------------------------------------------===//
1594// Other miscellaneous methods.
1595//===----------------------------------------------------------------------===//
1596
1597/// Get the source location for the given file:line:col triplet.
1598///
1599/// If the source file is included multiple times, the source location will
1600/// be based upon an arbitrary inclusion.
1601SourceLocation SourceManager::translateFileLineCol(const FileEntry *SourceFile,
1602 unsigned Line,
1603 unsigned Col) const {
1604 assert(SourceFile && "Null source file!");
1605 assert(Line && Col && "Line and column should start from 1!");
1606
1607 FileID FirstFID = translateFile(SourceFile);
1608 return translateLineCol(FID: FirstFID, Line, Col);
1609}
1610
1611/// Get the FileID for the given file.
1612///
1613/// If the source file is included multiple times, the FileID will be the
1614/// first inclusion.
1615FileID SourceManager::translateFile(const FileEntry *SourceFile) const {
1616 assert(SourceFile && "Null source file!");
1617
1618 // First, check the main file ID, since it is common to look for a
1619 // location in the main file.
1620 if (MainFileID.isValid()) {
1621 bool Invalid = false;
1622 const SLocEntry &MainSLoc = getSLocEntry(FID: MainFileID, Invalid: &Invalid);
1623 if (Invalid)
1624 return FileID();
1625
1626 if (MainSLoc.isFile()) {
1627 if (MainSLoc.getFile().getContentCache().OrigEntry == SourceFile)
1628 return MainFileID;
1629 }
1630 }
1631
1632 // The location we're looking for isn't in the main file; look
1633 // through all of the local source locations.
1634 for (unsigned I = 0, N = local_sloc_entry_size(); I != N; ++I) {
1635 const SLocEntry &SLoc = getLocalSLocEntry(Index: I);
1636 if (SLoc.isFile() &&
1637 SLoc.getFile().getContentCache().OrigEntry == SourceFile)
1638 return FileID::get(V: I);
1639 }
1640
1641 // If that still didn't help, try the modules.
1642 for (unsigned I = 0, N = loaded_sloc_entry_size(); I != N; ++I) {
1643 const SLocEntry &SLoc = getLoadedSLocEntry(Index: I);
1644 if (SLoc.isFile() &&
1645 SLoc.getFile().getContentCache().OrigEntry == SourceFile)
1646 return FileID::get(V: -int(I) - 2);
1647 }
1648
1649 return FileID();
1650}
1651
1652/// Get the source location in \arg FID for the given line:col.
1653/// Returns null location if \arg FID is not a file SLocEntry.
1654SourceLocation SourceManager::translateLineCol(FileID FID,
1655 unsigned Line,
1656 unsigned Col) const {
1657 // Lines are used as a one-based index into a zero-based array. This assert
1658 // checks for possible buffer underruns.
1659 assert(Line && Col && "Line and column should start from 1!");
1660
1661 if (FID.isInvalid())
1662 return SourceLocation();
1663
1664 bool Invalid = false;
1665 const SLocEntry &Entry = getSLocEntry(FID, Invalid: &Invalid);
1666 if (Invalid)
1667 return SourceLocation();
1668
1669 if (!Entry.isFile())
1670 return SourceLocation();
1671
1672 SourceLocation FileLoc = SourceLocation::getFileLoc(ID: Entry.getOffset());
1673
1674 if (Line == 1 && Col == 1)
1675 return FileLoc;
1676
1677 const ContentCache *Content = &Entry.getFile().getContentCache();
1678
1679 // If this is the first use of line information for this buffer, compute the
1680 // SourceLineCache for it on demand.
1681 std::optional<llvm::MemoryBufferRef> Buffer =
1682 Content->getBufferOrNone(Diag, FM&: getFileManager());
1683 if (!Buffer)
1684 return SourceLocation();
1685 if (!Content->SourceLineCache)
1686 Content->SourceLineCache =
1687 LineOffsetMapping::get(Buffer: *Buffer, Alloc&: ContentCacheAlloc);
1688
1689 if (Line > Content->SourceLineCache.size()) {
1690 unsigned Size = Buffer->getBufferSize();
1691 if (Size > 0)
1692 --Size;
1693 return FileLoc.getLocWithOffset(Offset: Size);
1694 }
1695
1696 unsigned FilePos = Content->SourceLineCache[Line - 1];
1697 const char *Buf = Buffer->getBufferStart() + FilePos;
1698 unsigned BufLength = Buffer->getBufferSize() - FilePos;
1699 if (BufLength == 0)
1700 return FileLoc.getLocWithOffset(Offset: FilePos);
1701
1702 unsigned i = 0;
1703
1704 // Check that the given column is valid.
1705 while (i < BufLength-1 && i < Col-1 && Buf[i] != '\n' && Buf[i] != '\r')
1706 ++i;
1707 return FileLoc.getLocWithOffset(Offset: FilePos + i);
1708}
1709
1710/// Compute a map of macro argument chunks to their expanded source
1711/// location. Chunks that are not part of a macro argument will map to an
1712/// invalid source location. e.g. if a file contains one macro argument at
1713/// offset 100 with length 10, this is how the map will be formed:
1714/// 0 -> SourceLocation()
1715/// 100 -> Expanded macro arg location
1716/// 110 -> SourceLocation()
1717void SourceManager::computeMacroArgsCache(MacroArgsMap &MacroArgsCache,
1718 FileID FID) const {
1719 assert(FID.isValid());
1720
1721 // Initially no macro argument chunk is present.
1722 MacroArgsCache.try_emplace(k: 0);
1723
1724 int ID = FID.ID;
1725 while (true) {
1726 ++ID;
1727 // Stop if there are no more FileIDs to check.
1728 if (ID > 0) {
1729 if (unsigned(ID) >= local_sloc_entry_size())
1730 return;
1731 } else if (ID == -1) {
1732 return;
1733 }
1734
1735 bool Invalid = false;
1736 const SrcMgr::SLocEntry &Entry = getSLocEntryByID(ID, Invalid: &Invalid);
1737 if (Invalid)
1738 return;
1739 if (Entry.isFile()) {
1740 auto& File = Entry.getFile();
1741 if (File.getFileCharacteristic() == C_User_ModuleMap ||
1742 File.getFileCharacteristic() == C_System_ModuleMap)
1743 continue;
1744
1745 SourceLocation IncludeLoc = File.getIncludeLoc();
1746 bool IncludedInFID =
1747 (IncludeLoc.isValid() && isInFileID(Loc: IncludeLoc, FID)) ||
1748 // Predefined header doesn't have a valid include location in main
1749 // file, but any files created by it should still be skipped when
1750 // computing macro args expanded in the main file.
1751 (FID == MainFileID && Entry.getFile().getName() == "<built-in>");
1752 if (IncludedInFID) {
1753 // Skip the files/macros of the #include'd file, we only care about
1754 // macros that lexed macro arguments from our file.
1755 if (Entry.getFile().NumCreatedFIDs)
1756 ID += Entry.getFile().NumCreatedFIDs - 1 /*because of next ++ID*/;
1757 continue;
1758 }
1759 // If file was included but not from FID, there is no more files/macros
1760 // that may be "contained" in this file.
1761 if (IncludeLoc.isValid())
1762 return;
1763 continue;
1764 }
1765
1766 const ExpansionInfo &ExpInfo = Entry.getExpansion();
1767
1768 if (ExpInfo.getExpansionLocStart().isFileID()) {
1769 if (!isInFileID(Loc: ExpInfo.getExpansionLocStart(), FID))
1770 return; // No more files/macros that may be "contained" in this file.
1771 }
1772
1773 if (!ExpInfo.isMacroArgExpansion())
1774 continue;
1775
1776 associateFileChunkWithMacroArgExp(MacroArgsCache, FID,
1777 SpellLoc: ExpInfo.getSpellingLoc(),
1778 ExpansionLoc: SourceLocation::getMacroLoc(ID: Entry.getOffset()),
1779 ExpansionLength: getFileIDSize(FID: FileID::get(V: ID)));
1780 }
1781}
1782
1783void SourceManager::associateFileChunkWithMacroArgExp(
1784 MacroArgsMap &MacroArgsCache,
1785 FileID FID,
1786 SourceLocation SpellLoc,
1787 SourceLocation ExpansionLoc,
1788 unsigned ExpansionLength) const {
1789 if (!SpellLoc.isFileID()) {
1790 SourceLocation::UIntTy SpellBeginOffs = SpellLoc.getOffset();
1791 SourceLocation::UIntTy SpellEndOffs = SpellBeginOffs + ExpansionLength;
1792
1793 // The spelling range for this macro argument expansion can span multiple
1794 // consecutive FileID entries. Go through each entry contained in the
1795 // spelling range and if one is itself a macro argument expansion, recurse
1796 // and associate the file chunk that it represents.
1797
1798 // Current FileID in the spelling range.
1799 auto [SpellFID, SpellRelativeOffs] = getDecomposedLoc(Loc: SpellLoc);
1800 while (true) {
1801 const SLocEntry &Entry = getSLocEntry(FID: SpellFID);
1802 SourceLocation::UIntTy SpellFIDBeginOffs = Entry.getOffset();
1803 unsigned SpellFIDSize = getFileIDSize(FID: SpellFID);
1804 SourceLocation::UIntTy SpellFIDEndOffs = SpellFIDBeginOffs + SpellFIDSize;
1805 const ExpansionInfo &Info = Entry.getExpansion();
1806 if (Info.isMacroArgExpansion()) {
1807 unsigned CurrSpellLength;
1808 if (SpellFIDEndOffs < SpellEndOffs)
1809 CurrSpellLength = SpellFIDSize - SpellRelativeOffs;
1810 else
1811 CurrSpellLength = ExpansionLength;
1812 associateFileChunkWithMacroArgExp(MacroArgsCache, FID,
1813 SpellLoc: Info.getSpellingLoc().getLocWithOffset(Offset: SpellRelativeOffs),
1814 ExpansionLoc, ExpansionLength: CurrSpellLength);
1815 }
1816
1817 if (SpellFIDEndOffs >= SpellEndOffs)
1818 return; // we covered all FileID entries in the spelling range.
1819
1820 // Move to the next FileID entry in the spelling range.
1821 unsigned advance = SpellFIDSize - SpellRelativeOffs + 1;
1822 ExpansionLoc = ExpansionLoc.getLocWithOffset(Offset: advance);
1823 ExpansionLength -= advance;
1824 ++SpellFID.ID;
1825 SpellRelativeOffs = 0;
1826 }
1827 }
1828
1829 assert(SpellLoc.isFileID());
1830
1831 unsigned BeginOffs;
1832 if (!isInFileID(Loc: SpellLoc, FID, RelativeOffset: &BeginOffs))
1833 return;
1834
1835 unsigned EndOffs = BeginOffs + ExpansionLength;
1836
1837 // Add a new chunk for this macro argument. A previous macro argument chunk
1838 // may have been lexed again, so e.g. if the map is
1839 // 0 -> SourceLocation()
1840 // 100 -> Expanded loc #1
1841 // 110 -> SourceLocation()
1842 // and we found a new macro FileID that lexed from offset 105 with length 3,
1843 // the new map will be:
1844 // 0 -> SourceLocation()
1845 // 100 -> Expanded loc #1
1846 // 105 -> Expanded loc #2
1847 // 108 -> Expanded loc #1
1848 // 110 -> SourceLocation()
1849 //
1850 // Since re-lexed macro chunks will always be the same size or less of
1851 // previous chunks, we only need to find where the ending of the new macro
1852 // chunk is mapped to and update the map with new begin/end mappings.
1853
1854 MacroArgsMap::iterator I = MacroArgsCache.upper_bound(x: EndOffs);
1855 --I;
1856 SourceLocation EndOffsMappedLoc = I->second;
1857 MacroArgsCache[BeginOffs] = ExpansionLoc;
1858 MacroArgsCache[EndOffs] = EndOffsMappedLoc;
1859}
1860
1861void SourceManager::updateSlocUsageStats() const {
1862 SourceLocation::UIntTy UsedBytes =
1863 NextLocalOffset + (MaxLoadedOffset - CurrentLoadedOffset);
1864 MaxUsedSLocBytes.updateMax(V: UsedBytes);
1865}
1866
1867/// If \arg Loc points inside a function macro argument, the returned
1868/// location will be the macro location in which the argument was expanded.
1869/// If a macro argument is used multiple times, the expanded location will
1870/// be at the first expansion of the argument.
1871/// e.g.
1872/// MY_MACRO(foo);
1873/// ^
1874/// Passing a file location pointing at 'foo', will yield a macro location
1875/// where 'foo' was expanded into.
1876SourceLocation
1877SourceManager::getMacroArgExpandedLocation(SourceLocation Loc) const {
1878 if (Loc.isInvalid() || !Loc.isFileID())
1879 return Loc;
1880
1881 auto [FID, Offset] = getDecomposedLoc(Loc);
1882 if (FID.isInvalid())
1883 return Loc;
1884
1885 std::unique_ptr<MacroArgsMap> &MacroArgsCache = MacroArgsCacheMap[FID];
1886 if (!MacroArgsCache) {
1887 MacroArgsCache = std::make_unique<MacroArgsMap>();
1888 computeMacroArgsCache(MacroArgsCache&: *MacroArgsCache, FID);
1889 }
1890
1891 assert(!MacroArgsCache->empty());
1892 MacroArgsMap::iterator I = MacroArgsCache->upper_bound(x: Offset);
1893 // In case every element in MacroArgsCache is greater than Offset we can't
1894 // decrement the iterator.
1895 if (I == MacroArgsCache->begin())
1896 return Loc;
1897
1898 --I;
1899
1900 SourceLocation::UIntTy MacroArgBeginOffs = I->first;
1901 SourceLocation MacroArgExpandedLoc = I->second;
1902 if (MacroArgExpandedLoc.isValid())
1903 return MacroArgExpandedLoc.getLocWithOffset(Offset: Offset - MacroArgBeginOffs);
1904
1905 return Loc;
1906}
1907
1908FileIDAndOffset SourceManager::getDecomposedIncludedLoc(FileID FID) const {
1909 if (FID.isInvalid())
1910 return std::make_pair(x: FileID(), y: 0);
1911
1912 // Uses IncludedLocMap to retrieve/cache the decomposed loc.
1913
1914 using DecompTy = FileIDAndOffset;
1915 auto InsertOp = IncludedLocMap.try_emplace(Key: FID);
1916 DecompTy &DecompLoc = InsertOp.first->second;
1917 if (!InsertOp.second)
1918 return DecompLoc; // already in map.
1919
1920 SourceLocation UpperLoc;
1921 bool Invalid = false;
1922 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, Invalid: &Invalid);
1923 if (!Invalid) {
1924 if (Entry.isExpansion())
1925 UpperLoc = Entry.getExpansion().getExpansionLocStart();
1926 else
1927 UpperLoc = Entry.getFile().getIncludeLoc();
1928 }
1929
1930 if (UpperLoc.isValid())
1931 DecompLoc = getDecomposedLoc(Loc: UpperLoc);
1932
1933 return DecompLoc;
1934}
1935
1936FileID SourceManager::getUniqueLoadedASTFileID(SourceLocation Loc) const {
1937 assert(isLoadedSourceLocation(Loc) &&
1938 "Must be a source location in a loaded PCH/Module file");
1939
1940 auto [FID, Ignore] = getDecomposedLoc(Loc);
1941 // `LoadedSLocEntryAllocBegin` stores the sorted lowest FID of each loaded
1942 // allocation. Later allocations have lower FileIDs. The call below is to find
1943 // the lowest FID of a loaded allocation from any FID in the same allocation.
1944 // The lowest FID is used to identify a loaded allocation.
1945 const FileID *FirstFID =
1946 llvm::lower_bound(Range: LoadedSLocEntryAllocBegin, Value&: FID, C: std::greater<FileID>{});
1947
1948 assert(FirstFID &&
1949 "The failure to find the first FileID of a "
1950 "loaded AST from a loaded source location was unexpected.");
1951 return *FirstFID;
1952}
1953
1954bool SourceManager::isInTheSameTranslationUnitImpl(
1955 const FileIDAndOffset &LOffs, const FileIDAndOffset &ROffs) const {
1956 // If one is local while the other is loaded.
1957 if (isLoadedFileID(FID: LOffs.first) != isLoadedFileID(FID: ROffs.first))
1958 return false;
1959
1960 if (isLoadedFileID(FID: LOffs.first) && isLoadedFileID(FID: ROffs.first)) {
1961 auto FindSLocEntryAlloc = [this](FileID FID) {
1962 // Loaded FileIDs are negative, we store the lowest FileID from each
1963 // allocation, later allocations have lower FileIDs.
1964 return llvm::lower_bound(Range: LoadedSLocEntryAllocBegin, Value&: FID,
1965 C: std::greater<FileID>{});
1966 };
1967
1968 // If both are loaded from different AST files.
1969 if (FindSLocEntryAlloc(LOffs.first) != FindSLocEntryAlloc(ROffs.first))
1970 return false;
1971 }
1972
1973 return true;
1974}
1975
1976/// Given a decomposed source location, move it up the include/expansion stack
1977/// to the parent source location within the same translation unit. If this is
1978/// possible, return the decomposed version of the parent in Loc and return
1979/// false. If Loc is a top-level entry, return true and don't modify it.
1980static bool MoveUpTranslationUnitIncludeHierarchy(FileIDAndOffset &Loc,
1981 const SourceManager &SM) {
1982 FileIDAndOffset UpperLoc = SM.getDecomposedIncludedLoc(FID: Loc.first);
1983 if (UpperLoc.first.isInvalid() ||
1984 !SM.isInTheSameTranslationUnitImpl(LOffs: UpperLoc, ROffs: Loc))
1985 return true; // We reached the top.
1986
1987 Loc = UpperLoc;
1988 return false;
1989}
1990
1991/// Return the cache entry for comparing the given file IDs
1992/// for isBeforeInTranslationUnit.
1993InBeforeInTUCacheEntry &SourceManager::getInBeforeInTUCache(FileID LFID,
1994 FileID RFID) const {
1995 // This is a magic number for limiting the cache size. It was experimentally
1996 // derived from a small Objective-C project (where the cache filled
1997 // out to ~250 items). We can make it larger if necessary.
1998 // FIXME: this is almost certainly full these days. Use an LRU cache?
1999 enum { MagicCacheSize = 300 };
2000 IsBeforeInTUCacheKey Key(LFID, RFID);
2001
2002 // If the cache size isn't too large, do a lookup and if necessary default
2003 // construct an entry. We can then return it to the caller for direct
2004 // use. When they update the value, the cache will get automatically
2005 // updated as well.
2006 if (IBTUCache.size() < MagicCacheSize)
2007 return IBTUCache.try_emplace(Key, Args&: LFID, Args&: RFID).first->second;
2008
2009 // Otherwise, do a lookup that will not construct a new value.
2010 InBeforeInTUCache::iterator I = IBTUCache.find(Val: Key);
2011 if (I != IBTUCache.end())
2012 return I->second;
2013
2014 // Fall back to the overflow value.
2015 IBTUCacheOverflow.setQueryFIDs(LHS: LFID, RHS: RFID);
2016 return IBTUCacheOverflow;
2017}
2018
2019/// Determines the order of 2 source locations in the translation unit.
2020///
2021/// \returns true if LHS source location comes before RHS, false otherwise.
2022bool SourceManager::isBeforeInTranslationUnit(SourceLocation LHS,
2023 SourceLocation RHS) const {
2024 assert(LHS.isValid() && RHS.isValid() && "Passed invalid source location!");
2025 if (LHS == RHS)
2026 return false;
2027
2028 FileIDAndOffset LOffs = getDecomposedLoc(Loc: LHS);
2029 FileIDAndOffset ROffs = getDecomposedLoc(Loc: RHS);
2030
2031 // getDecomposedLoc may have failed to return a valid FileID because, e.g. it
2032 // is a serialized one referring to a file that was removed after we loaded
2033 // the PCH.
2034 if (LOffs.first.isInvalid() || ROffs.first.isInvalid())
2035 return LOffs.first.isInvalid() && !ROffs.first.isInvalid();
2036
2037 std::pair<bool, bool> InSameTU = isInTheSameTranslationUnit(LOffs, ROffs);
2038 if (InSameTU.first)
2039 return InSameTU.second;
2040 // This case is used by libclang: clang_isBeforeInTranslationUnit
2041 return LOffs.first < ROffs.first;
2042}
2043
2044std::pair<bool, bool>
2045SourceManager::isInTheSameTranslationUnit(FileIDAndOffset &LOffs,
2046 FileIDAndOffset &ROffs) const {
2047 // If the source locations are not in the same TU, return early.
2048 if (!isInTheSameTranslationUnitImpl(LOffs, ROffs))
2049 return std::make_pair(x: false, y: false);
2050
2051 // If the source locations are in the same file, just compare offsets.
2052 if (LOffs.first == ROffs.first)
2053 return std::make_pair(x: true, y: LOffs.second < ROffs.second);
2054
2055 // If we are comparing a source location with multiple locations in the same
2056 // file, we get a big win by caching the result.
2057 InBeforeInTUCacheEntry &IsBeforeInTUCache =
2058 getInBeforeInTUCache(LFID: LOffs.first, RFID: ROffs.first);
2059
2060 // If we are comparing a source location with multiple locations in the same
2061 // file, we get a big win by caching the result.
2062 if (IsBeforeInTUCache.isCacheValid())
2063 return std::make_pair(
2064 x: true, y: IsBeforeInTUCache.getCachedResult(LOffset: LOffs.second, ROffset: ROffs.second));
2065
2066 // Okay, we missed in the cache, we'll compute the answer and populate it.
2067 // We need to find the common ancestor. The only way of doing this is to
2068 // build the complete include chain for one and then walking up the chain
2069 // of the other looking for a match.
2070
2071 // A location within a FileID on the path up from LOffs to the main file.
2072 struct Entry {
2073 FileIDAndOffset DecomposedLoc; // FileID redundant, but clearer.
2074 FileID ChildFID; // Used for breaking ties. Invalid for the initial loc.
2075 };
2076 llvm::SmallDenseMap<FileID, Entry, 16> LChain;
2077
2078 FileID LChild;
2079 do {
2080 LChain.try_emplace(Key: LOffs.first, Args: Entry{.DecomposedLoc: LOffs, .ChildFID: LChild});
2081 // We catch the case where LOffs is in a file included by ROffs and
2082 // quit early. The other way round unfortunately remains suboptimal.
2083 if (LOffs.first == ROffs.first)
2084 break;
2085 LChild = LOffs.first;
2086 } while (!MoveUpTranslationUnitIncludeHierarchy(Loc&: LOffs, SM: *this));
2087
2088 FileID RChild;
2089 do {
2090 auto LIt = LChain.find(Val: ROffs.first);
2091 if (LIt != LChain.end()) {
2092 // Compare the locations within the common file and cache them.
2093 LOffs = LIt->second.DecomposedLoc;
2094 LChild = LIt->second.ChildFID;
2095 // The relative order of LChild and RChild is a tiebreaker when
2096 // - locs expand to the same location (occurs in macro arg expansion)
2097 // - one loc is a parent of the other (we consider the parent as "first")
2098 // For the parent entry to be first, its invalid child file ID must
2099 // compare smaller to the valid child file ID of the other entry.
2100 // However loaded FileIDs are <0, so we perform *unsigned* comparison!
2101 // This changes the relative order of local vs loaded FileIDs, but it
2102 // doesn't matter as these are never mixed in macro expansion.
2103 unsigned LChildID = LChild.ID;
2104 unsigned RChildID = RChild.ID;
2105 assert(((LOffs.second != ROffs.second) ||
2106 (LChildID == 0 || RChildID == 0) ||
2107 isInSameSLocAddrSpace(getComposedLoc(LChild, 0),
2108 getComposedLoc(RChild, 0), nullptr)) &&
2109 "Mixed local/loaded FileIDs with same include location?");
2110 IsBeforeInTUCache.setCommonLoc(commonFID: LOffs.first, lCommonOffset: LOffs.second, rCommonOffset: ROffs.second,
2111 LParentBeforeRParent: LChildID < RChildID);
2112 return std::make_pair(
2113 x: true, y: IsBeforeInTUCache.getCachedResult(LOffset: LOffs.second, ROffset: ROffs.second));
2114 }
2115 RChild = ROffs.first;
2116 } while (!MoveUpTranslationUnitIncludeHierarchy(Loc&: ROffs, SM: *this));
2117
2118 // If we found no match, the location is either in a built-ins buffer or
2119 // associated with global inline asm. PR5662 and PR22576 are examples.
2120
2121 StringRef LB = getBufferOrFake(FID: LOffs.first).getBufferIdentifier();
2122 StringRef RB = getBufferOrFake(FID: ROffs.first).getBufferIdentifier();
2123
2124 bool LIsBuiltins = LB == "<built-in>";
2125 bool RIsBuiltins = RB == "<built-in>";
2126 // Sort built-in before non-built-in.
2127 if (LIsBuiltins || RIsBuiltins) {
2128 if (LIsBuiltins != RIsBuiltins)
2129 return std::make_pair(x: true, y&: LIsBuiltins);
2130 // Both are in built-in buffers, but from different files. We just claim
2131 // that lower IDs come first.
2132 return std::make_pair(x: true, y: LOffs.first < ROffs.first);
2133 }
2134
2135 bool LIsAsm = LB == "<inline asm>";
2136 bool RIsAsm = RB == "<inline asm>";
2137 // Sort assembler after built-ins, but before the rest.
2138 if (LIsAsm || RIsAsm) {
2139 if (LIsAsm != RIsAsm)
2140 return std::make_pair(x: true, y&: RIsAsm);
2141 assert(LOffs.first == ROffs.first);
2142 return std::make_pair(x: true, y: false);
2143 }
2144
2145 bool LIsScratch = LB == "<scratch space>";
2146 bool RIsScratch = RB == "<scratch space>";
2147 // Sort scratch after inline asm, but before the rest.
2148 if (LIsScratch || RIsScratch) {
2149 if (LIsScratch != RIsScratch)
2150 return std::make_pair(x: true, y&: LIsScratch);
2151 return std::make_pair(x: true, y: LOffs.second < ROffs.second);
2152 }
2153
2154 llvm_unreachable("Unsortable locations found");
2155}
2156
2157void SourceManager::PrintStats() const {
2158 llvm::errs() << "\n*** Source Manager Stats:\n";
2159 llvm::errs() << FileInfos.size() << " files mapped, " << MemBufferInfos.size()
2160 << " mem buffers mapped.\n";
2161 llvm::errs() << LocalSLocEntryTable.size() << " local SLocEntries allocated ("
2162 << llvm::capacity_in_bytes(X: LocalSLocEntryTable)
2163 << " bytes of capacity), " << NextLocalOffset
2164 << "B of SLoc address space used.\n";
2165 llvm::errs() << LoadedSLocEntryTable.size()
2166 << " loaded SLocEntries allocated ("
2167 << llvm::capacity_in_bytes(x: LoadedSLocEntryTable)
2168 << " bytes of capacity), "
2169 << MaxLoadedOffset - CurrentLoadedOffset
2170 << "B of SLoc address space used.\n";
2171
2172 unsigned NumLineNumsComputed = 0;
2173 unsigned NumFileBytesMapped = 0;
2174 for (fileinfo_iterator I = fileinfo_begin(), E = fileinfo_end(); I != E; ++I){
2175 NumLineNumsComputed += bool(I->second->SourceLineCache);
2176 NumFileBytesMapped += I->second->getSizeBytesMapped();
2177 }
2178 unsigned NumMacroArgsComputed = MacroArgsCacheMap.size();
2179
2180 llvm::errs() << NumFileBytesMapped << " bytes of files mapped, "
2181 << NumLineNumsComputed << " files with line #'s computed, "
2182 << NumMacroArgsComputed << " files with macro args computed.\n";
2183 llvm::errs() << "FileID scans: " << NumLinearScans << " linear, "
2184 << NumBinaryProbes << " binary.\n";
2185}
2186
2187LLVM_DUMP_METHOD void SourceManager::dump() const {
2188 llvm::raw_ostream &out = llvm::errs();
2189
2190 auto DumpSLocEntry = [&](int ID, const SrcMgr::SLocEntry &Entry,
2191 std::optional<SourceLocation::UIntTy> NextStart) {
2192 out << "SLocEntry <FileID " << ID << "> " << (Entry.isFile() ? "file" : "expansion")
2193 << " <SourceLocation " << Entry.getOffset() << ":";
2194 if (NextStart)
2195 out << *NextStart << ">\n";
2196 else
2197 out << "???\?>\n";
2198 if (Entry.isFile()) {
2199 auto &FI = Entry.getFile();
2200 if (FI.NumCreatedFIDs)
2201 out << " covers <FileID " << ID << ":" << int(ID + FI.NumCreatedFIDs)
2202 << ">\n";
2203 if (FI.getIncludeLoc().isValid())
2204 out << " included from " << FI.getIncludeLoc().getOffset() << "\n";
2205 auto &CC = FI.getContentCache();
2206 out << " for " << (CC.OrigEntry ? CC.OrigEntry->getName() : "<none>")
2207 << "\n";
2208 if (CC.BufferOverridden)
2209 out << " contents overridden\n";
2210 if (CC.ContentsEntry != CC.OrigEntry) {
2211 out << " contents from "
2212 << (CC.ContentsEntry ? CC.ContentsEntry->getName() : "<none>")
2213 << "\n";
2214 }
2215 } else {
2216 auto &EI = Entry.getExpansion();
2217 out << " spelling from " << EI.getSpellingLoc().getOffset() << "\n";
2218 out << " macro " << (EI.isMacroArgExpansion() ? "arg" : "body")
2219 << " range <" << EI.getExpansionLocStart().getOffset() << ":"
2220 << EI.getExpansionLocEnd().getOffset() << ">\n";
2221 }
2222 };
2223
2224 // Dump local SLocEntries.
2225 for (unsigned ID = 0, NumIDs = LocalSLocEntryTable.size(); ID != NumIDs; ++ID) {
2226 DumpSLocEntry(ID, LocalSLocEntryTable[ID],
2227 ID == NumIDs - 1 ? NextLocalOffset
2228 : LocalSLocEntryTable[ID + 1].getOffset());
2229 }
2230 // Dump loaded SLocEntries.
2231 std::optional<SourceLocation::UIntTy> NextStart;
2232 for (unsigned Index = 0; Index != LoadedSLocEntryTable.size(); ++Index) {
2233 int ID = -(int)Index - 2;
2234 if (SLocEntryLoaded[Index]) {
2235 DumpSLocEntry(ID, LoadedSLocEntryTable[Index], NextStart);
2236 NextStart = LoadedSLocEntryTable[Index].getOffset();
2237 } else {
2238 NextStart = std::nullopt;
2239 }
2240 }
2241}
2242
2243void SourceManager::noteSLocAddressSpaceUsage(
2244 DiagnosticsEngine &Diag, std::optional<unsigned> MaxNotes) const {
2245 struct Info {
2246 // A location where this file was entered.
2247 SourceLocation Loc;
2248 // Number of times this FileEntry was entered.
2249 unsigned Inclusions = 0;
2250 // Size usage from the file itself.
2251 uint64_t DirectSize = 0;
2252 // Total size usage from the file and its macro expansions.
2253 uint64_t TotalSize = 0;
2254 };
2255 using UsageMap = llvm::MapVector<const FileEntry*, Info>;
2256
2257 UsageMap Usage;
2258 uint64_t CountedSize = 0;
2259
2260 auto AddUsageForFileID = [&](FileID ID) {
2261 // The +1 here is because getFileIDSize doesn't include the extra byte for
2262 // the one-past-the-end location.
2263 unsigned Size = getFileIDSize(FID: ID) + 1;
2264
2265 // Find the file that used this address space, either directly or by
2266 // macro expansion.
2267 SourceLocation FileStart = getFileLoc(Loc: getComposedLoc(FID: ID, Offset: 0));
2268 FileID FileLocID = getFileID(SpellingLoc: FileStart);
2269 const FileEntry *Entry = getFileEntryForID(FID: FileLocID);
2270
2271 Info &EntryInfo = Usage[Entry];
2272 if (EntryInfo.Loc.isInvalid())
2273 EntryInfo.Loc = FileStart;
2274 if (ID == FileLocID) {
2275 ++EntryInfo.Inclusions;
2276 EntryInfo.DirectSize += Size;
2277 }
2278 EntryInfo.TotalSize += Size;
2279 CountedSize += Size;
2280 };
2281
2282 // Loaded SLocEntries have indexes counting downwards from -2.
2283 for (size_t Index = 0; Index != LoadedSLocEntryTable.size(); ++Index) {
2284 AddUsageForFileID(FileID::get(V: -2 - Index));
2285 }
2286 // Local SLocEntries have indexes counting upwards from 0.
2287 for (size_t Index = 0; Index != LocalSLocEntryTable.size(); ++Index) {
2288 AddUsageForFileID(FileID::get(V: Index));
2289 }
2290
2291 // Sort the usage by size from largest to smallest. Break ties by raw source
2292 // location.
2293 auto SortedUsage = Usage.takeVector();
2294 auto Cmp = [](const UsageMap::value_type &A, const UsageMap::value_type &B) {
2295 return A.second.TotalSize > B.second.TotalSize ||
2296 (A.second.TotalSize == B.second.TotalSize &&
2297 A.second.Loc < B.second.Loc);
2298 };
2299 auto SortedEnd = SortedUsage.end();
2300 if (MaxNotes && SortedUsage.size() > *MaxNotes) {
2301 SortedEnd = SortedUsage.begin() + *MaxNotes;
2302 std::nth_element(first: SortedUsage.begin(), nth: SortedEnd, last: SortedUsage.end(), comp: Cmp);
2303 }
2304 std::sort(first: SortedUsage.begin(), last: SortedEnd, comp: Cmp);
2305
2306 // Produce note on sloc address space usage total.
2307 uint64_t LocalUsage = NextLocalOffset;
2308 uint64_t LoadedUsage = MaxLoadedOffset - CurrentLoadedOffset;
2309 int UsagePercent = static_cast<int>(100.0 * double(LocalUsage + LoadedUsage) /
2310 MaxLoadedOffset);
2311 Diag.Report(DiagID: diag::note_total_sloc_usage)
2312 << LocalUsage << LoadedUsage << (LocalUsage + LoadedUsage)
2313 << UsagePercent;
2314
2315 // Produce notes on sloc address space usage for each file with a high usage.
2316 uint64_t ReportedSize = 0;
2317 for (auto &[Entry, FileInfo] :
2318 llvm::make_range(x: SortedUsage.begin(), y: SortedEnd)) {
2319 Diag.Report(Loc: FileInfo.Loc, DiagID: diag::note_file_sloc_usage)
2320 << FileInfo.Inclusions << FileInfo.DirectSize
2321 << (FileInfo.TotalSize - FileInfo.DirectSize);
2322 ReportedSize += FileInfo.TotalSize;
2323 }
2324
2325 // Describe any remaining usage not reported in the per-file usage.
2326 if (ReportedSize != CountedSize) {
2327 Diag.Report(DiagID: diag::note_file_misc_sloc_usage)
2328 << (SortedUsage.end() - SortedEnd) << CountedSize - ReportedSize;
2329 }
2330}
2331
2332ExternalSLocEntrySource::~ExternalSLocEntrySource() = default;
2333
2334/// Return the amount of memory used by memory buffers, breaking down
2335/// by heap-backed versus mmap'ed memory.
2336SourceManager::MemoryBufferSizes SourceManager::getMemoryBufferSizes() const {
2337 size_t malloc_bytes = 0;
2338 size_t mmap_bytes = 0;
2339
2340 for (unsigned i = 0, e = MemBufferInfos.size(); i != e; ++i)
2341 if (size_t sized_mapped = MemBufferInfos[i]->getSizeBytesMapped())
2342 switch (MemBufferInfos[i]->getMemoryBufferKind()) {
2343 case llvm::MemoryBuffer::MemoryBuffer_MMap:
2344 mmap_bytes += sized_mapped;
2345 break;
2346 case llvm::MemoryBuffer::MemoryBuffer_Malloc:
2347 malloc_bytes += sized_mapped;
2348 break;
2349 }
2350
2351 return MemoryBufferSizes(malloc_bytes, mmap_bytes);
2352}
2353
2354size_t SourceManager::getDataStructureSizes() const {
2355 size_t size = llvm::capacity_in_bytes(x: MemBufferInfos) +
2356 llvm::capacity_in_bytes(X: LocalSLocEntryTable) +
2357 llvm::capacity_in_bytes(x: LoadedSLocEntryTable) +
2358 llvm::capacity_in_bytes(X: SLocEntryLoaded) +
2359 llvm::capacity_in_bytes(X: FileInfos);
2360
2361 if (OverriddenFilesInfo)
2362 size += llvm::capacity_in_bytes(X: OverriddenFilesInfo->OverriddenFiles);
2363
2364 return size;
2365}
2366
2367SourceManagerForFile::SourceManagerForFile(StringRef FileName,
2368 StringRef Content) {
2369 // This is referenced by `FileMgr` and will be released by `FileMgr` when it
2370 // is deleted.
2371 IntrusiveRefCntPtr<llvm::vfs::InMemoryFileSystem> InMemoryFileSystem(
2372 new llvm::vfs::InMemoryFileSystem);
2373 InMemoryFileSystem->addFile(
2374 Path: FileName, ModificationTime: 0,
2375 Buffer: llvm::MemoryBuffer::getMemBuffer(InputData: Content, BufferName: FileName,
2376 /*RequiresNullTerminator=*/false));
2377 // This is passed to `SM` as reference, so the pointer has to be referenced
2378 // in `Environment` so that `FileMgr` can out-live this function scope.
2379 FileMgr =
2380 std::make_unique<FileManager>(args: FileSystemOptions(), args&: InMemoryFileSystem);
2381 DiagOpts = std::make_unique<DiagnosticOptions>();
2382 // This is passed to `SM` as reference, so the pointer has to be referenced
2383 // by `Environment` due to the same reason above.
2384 Diagnostics = std::make_unique<DiagnosticsEngine>(
2385 args: IntrusiveRefCntPtr<DiagnosticIDs>(new DiagnosticIDs), args&: *DiagOpts);
2386 SourceMgr = std::make_unique<SourceManager>(args&: *Diagnostics, args&: *FileMgr);
2387 FileEntryRef FE = llvm::cantFail(ValOrErr: FileMgr->getFileRef(Filename: FileName));
2388 FileID ID =
2389 SourceMgr->createFileID(SourceFile: FE, IncludePos: SourceLocation(), FileCharacter: clang::SrcMgr::C_User);
2390 assert(ID.isValid());
2391 SourceMgr->setMainFileID(ID);
2392}
2393

source code of clang/lib/Basic/SourceManager.cpp