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/Capacity.h"
28#include "llvm/Support/Compiler.h"
29#include "llvm/Support/Endian.h"
30#include "llvm/Support/ErrorHandling.h"
31#include "llvm/Support/FileSystem.h"
32#include "llvm/Support/MathExtras.h"
33#include "llvm/Support/MemoryBuffer.h"
34#include "llvm/Support/Path.h"
35#include "llvm/Support/raw_ostream.h"
36#include <algorithm>
37#include <cassert>
38#include <cstddef>
39#include <cstdint>
40#include <memory>
41#include <optional>
42#include <tuple>
43#include <utility>
44#include <vector>
45
46using namespace clang;
47using namespace SrcMgr;
48using llvm::MemoryBuffer;
49
50#define DEBUG_TYPE "source-manager"
51
52// Reaching a limit of 2^31 results in a hard error. This metric allows to track
53// if particular invocation of the compiler is close to it.
54STATISTIC(MaxUsedSLocBytes, "Maximum number of bytes used by source locations "
55 "(both loaded and local).");
56
57//===----------------------------------------------------------------------===//
58// SourceManager Helper Classes
59//===----------------------------------------------------------------------===//
60
61/// getSizeBytesMapped - Returns the number of bytes actually mapped for this
62/// ContentCache. This can be 0 if the MemBuffer was not actually expanded.
63unsigned ContentCache::getSizeBytesMapped() const {
64 return Buffer ? Buffer->getBufferSize() : 0;
65}
66
67/// Returns the kind of memory used to back the memory buffer for
68/// this content cache. This is used for performance analysis.
69llvm::MemoryBuffer::BufferKind ContentCache::getMemoryBufferKind() const {
70 if (Buffer == nullptr) {
71 assert(0 && "Buffer should never be null");
72 return llvm::MemoryBuffer::MemoryBuffer_Malloc;
73 }
74 return Buffer->getBufferKind();
75}
76
77/// getSize - Returns the size of the content encapsulated by this ContentCache.
78/// This can be the size of the source file or the size of an arbitrary
79/// scratch buffer. If the ContentCache encapsulates a source file, that
80/// file is not lazily brought in from disk to satisfy this query.
81unsigned ContentCache::getSize() const {
82 return Buffer ? (unsigned)Buffer->getBufferSize()
83 : (unsigned)ContentsEntry->getSize();
84}
85
86const char *ContentCache::getInvalidBOM(StringRef BufStr) {
87 // If the buffer is valid, check to see if it has a UTF Byte Order Mark
88 // (BOM). We only support UTF-8 with and without a BOM right now. See
89 // http://en.wikipedia.org/wiki/Byte_order_mark for more information.
90 const char *InvalidBOM =
91 llvm::StringSwitch<const char *>(BufStr)
92 .StartsWith(S: llvm::StringLiteral::withInnerNUL(Str: "\x00\x00\xFE\xFF"),
93 Value: "UTF-32 (BE)")
94 .StartsWith(S: llvm::StringLiteral::withInnerNUL(Str: "\xFF\xFE\x00\x00"),
95 Value: "UTF-32 (LE)")
96 .StartsWith(S: "\xFE\xFF", Value: "UTF-16 (BE)")
97 .StartsWith(S: "\xFF\xFE", Value: "UTF-16 (LE)")
98 .StartsWith(S: "\x2B\x2F\x76", Value: "UTF-7")
99 .StartsWith(S: "\xF7\x64\x4C", Value: "UTF-1")
100 .StartsWith(S: "\xDD\x73\x66\x73", Value: "UTF-EBCDIC")
101 .StartsWith(S: "\x0E\xFE\xFF", Value: "SCSU")
102 .StartsWith(S: "\xFB\xEE\x28", Value: "BOCU-1")
103 .StartsWith(S: "\x84\x31\x95\x33", Value: "GB-18030")
104 .Default(Value: nullptr);
105
106 return InvalidBOM;
107}
108
109std::optional<llvm::MemoryBufferRef>
110ContentCache::getBufferOrNone(DiagnosticsEngine &Diag, FileManager &FM,
111 SourceLocation Loc) const {
112 // Lazily create the Buffer for ContentCaches that wrap files. If we already
113 // computed it, just return what we have.
114 if (IsBufferInvalid)
115 return std::nullopt;
116 if (Buffer)
117 return Buffer->getMemBufferRef();
118 if (!ContentsEntry)
119 return std::nullopt;
120
121 // Start with the assumption that the buffer is invalid to simplify early
122 // return paths.
123 IsBufferInvalid = true;
124
125 auto BufferOrError = FM.getBufferForFile(Entry: *ContentsEntry, isVolatile: IsFileVolatile);
126
127 // If we were unable to open the file, then we are in an inconsistent
128 // situation where the content cache referenced a file which no longer
129 // exists. Most likely, we were using a stat cache with an invalid entry but
130 // the file could also have been removed during processing. Since we can't
131 // really deal with this situation, just create an empty buffer.
132 if (!BufferOrError) {
133 if (Diag.isDiagnosticInFlight())
134 Diag.SetDelayedDiagnostic(DiagID: diag::err_cannot_open_file,
135 Arg1: ContentsEntry->getName(),
136 Arg2: BufferOrError.getError().message());
137 else
138 Diag.Report(Loc, DiagID: diag::err_cannot_open_file)
139 << ContentsEntry->getName() << BufferOrError.getError().message();
140
141 return std::nullopt;
142 }
143
144 Buffer = std::move(*BufferOrError);
145
146 // Check that the file's size fits in an 'unsigned' (with room for a
147 // past-the-end value). This is deeply regrettable, but various parts of
148 // Clang (including elsewhere in this file!) use 'unsigned' to represent file
149 // offsets, line numbers, string literal lengths, and so on, and fail
150 // miserably on large source files.
151 //
152 // Note: ContentsEntry could be a named pipe, in which case
153 // ContentsEntry::getSize() could have the wrong size. Use
154 // MemoryBuffer::getBufferSize() instead.
155 if (Buffer->getBufferSize() >= std::numeric_limits<unsigned>::max()) {
156 if (Diag.isDiagnosticInFlight())
157 Diag.SetDelayedDiagnostic(DiagID: diag::err_file_too_large,
158 Arg1: ContentsEntry->getName());
159 else
160 Diag.Report(Loc, DiagID: diag::err_file_too_large)
161 << ContentsEntry->getName();
162
163 return std::nullopt;
164 }
165
166 // Unless this is a named pipe (in which case we can handle a mismatch),
167 // check that the file's size is the same as in the file entry (which may
168 // have come from a stat cache).
169 if (!ContentsEntry->isNamedPipe() &&
170 Buffer->getBufferSize() != (size_t)ContentsEntry->getSize()) {
171 if (Diag.isDiagnosticInFlight())
172 Diag.SetDelayedDiagnostic(DiagID: diag::err_file_modified,
173 Arg1: ContentsEntry->getName());
174 else
175 Diag.Report(Loc, DiagID: diag::err_file_modified)
176 << ContentsEntry->getName();
177
178 return std::nullopt;
179 }
180
181 // If the buffer is valid, check to see if it has a UTF Byte Order Mark
182 // (BOM). We only support UTF-8 with and without a BOM right now. See
183 // http://en.wikipedia.org/wiki/Byte_order_mark for more information.
184 StringRef BufStr = Buffer->getBuffer();
185 const char *InvalidBOM = getInvalidBOM(BufStr);
186
187 if (InvalidBOM) {
188 Diag.Report(Loc, DiagID: diag::err_unsupported_bom)
189 << InvalidBOM << ContentsEntry->getName();
190 return std::nullopt;
191 }
192
193 // Buffer has been validated.
194 IsBufferInvalid = false;
195 return Buffer->getMemBufferRef();
196}
197
198unsigned LineTableInfo::getLineTableFilenameID(StringRef Name) {
199 auto IterBool = FilenameIDs.try_emplace(Key: Name, Args: FilenamesByID.size());
200 if (IterBool.second)
201 FilenamesByID.push_back(x: &*IterBool.first);
202 return IterBool.first->second;
203}
204
205/// Add a line note to the line table that indicates that there is a \#line or
206/// GNU line marker at the specified FID/Offset location which changes the
207/// presumed location to LineNo/FilenameID. If EntryExit is 0, then this doesn't
208/// change the presumed \#include stack. If it is 1, this is a file entry, if
209/// it is 2 then this is a file exit. FileKind specifies whether this is a
210/// system header or extern C system header.
211void LineTableInfo::AddLineNote(FileID FID, unsigned Offset, unsigned LineNo,
212 int FilenameID, unsigned EntryExit,
213 SrcMgr::CharacteristicKind FileKind) {
214 std::vector<LineEntry> &Entries = LineEntries[FID];
215
216 assert((Entries.empty() || Entries.back().FileOffset < Offset) &&
217 "Adding line entries out of order!");
218
219 unsigned IncludeOffset = 0;
220 if (EntryExit == 1) {
221 // Push #include
222 IncludeOffset = Offset-1;
223 } else {
224 const auto *PrevEntry = Entries.empty() ? nullptr : &Entries.back();
225 if (EntryExit == 2) {
226 // Pop #include
227 assert(PrevEntry && PrevEntry->IncludeOffset &&
228 "PPDirectives should have caught case when popping empty include "
229 "stack");
230 PrevEntry = FindNearestLineEntry(FID, Offset: PrevEntry->IncludeOffset);
231 }
232 if (PrevEntry) {
233 IncludeOffset = PrevEntry->IncludeOffset;
234 if (FilenameID == -1) {
235 // An unspecified FilenameID means use the previous (or containing)
236 // filename if available, or the main source file otherwise.
237 FilenameID = PrevEntry->FilenameID;
238 }
239 }
240 }
241
242 Entries.push_back(x: LineEntry::get(Offs: Offset, Line: LineNo, Filename: FilenameID, FileKind,
243 IncludeOffset));
244}
245
246/// FindNearestLineEntry - Find the line entry nearest to FID that is before
247/// it. If there is no line entry before Offset in FID, return null.
248const LineEntry *LineTableInfo::FindNearestLineEntry(FileID FID,
249 unsigned Offset) {
250 const std::vector<LineEntry> &Entries = LineEntries[FID];
251 assert(!Entries.empty() && "No #line entries for this FID after all!");
252
253 // It is very common for the query to be after the last #line, check this
254 // first.
255 if (Entries.back().FileOffset <= Offset)
256 return &Entries.back();
257
258 // Do a binary search to find the maximal element that is still before Offset.
259 std::vector<LineEntry>::const_iterator I = llvm::upper_bound(Range: Entries, Value&: Offset);
260 if (I == Entries.begin())
261 return nullptr;
262 return &*--I;
263}
264
265/// Add a new line entry that has already been encoded into
266/// the internal representation of the line table.
267void LineTableInfo::AddEntry(FileID FID,
268 const std::vector<LineEntry> &Entries) {
269 LineEntries[FID] = Entries;
270}
271
272/// getLineTableFilenameID - Return the uniqued ID for the specified filename.
273unsigned SourceManager::getLineTableFilenameID(StringRef Name) {
274 return getLineTable().getLineTableFilenameID(Name);
275}
276
277/// AddLineNote - Add a line note to the line table for the FileID and offset
278/// specified by Loc. If FilenameID is -1, it is considered to be
279/// unspecified.
280void SourceManager::AddLineNote(SourceLocation Loc, unsigned LineNo,
281 int FilenameID, bool IsFileEntry,
282 bool IsFileExit,
283 SrcMgr::CharacteristicKind FileKind) {
284 std::pair<FileID, unsigned> LocInfo = getDecomposedExpansionLoc(Loc);
285
286 bool Invalid = false;
287 SLocEntry &Entry = getSLocEntry(FID: LocInfo.first, Invalid: &Invalid);
288 if (!Entry.isFile() || Invalid)
289 return;
290
291 SrcMgr::FileInfo &FileInfo = Entry.getFile();
292
293 // Remember that this file has #line directives now if it doesn't already.
294 FileInfo.setHasLineDirectives();
295
296 (void) getLineTable();
297
298 unsigned EntryExit = 0;
299 if (IsFileEntry)
300 EntryExit = 1;
301 else if (IsFileExit)
302 EntryExit = 2;
303
304 LineTable->AddLineNote(FID: LocInfo.first, Offset: LocInfo.second, LineNo, FilenameID,
305 EntryExit, FileKind);
306}
307
308LineTableInfo &SourceManager::getLineTable() {
309 if (!LineTable)
310 LineTable.reset(p: new LineTableInfo());
311 return *LineTable;
312}
313
314//===----------------------------------------------------------------------===//
315// Private 'Create' methods.
316//===----------------------------------------------------------------------===//
317
318SourceManager::SourceManager(DiagnosticsEngine &Diag, FileManager &FileMgr,
319 bool UserFilesAreVolatile)
320 : Diag(Diag), FileMgr(FileMgr), UserFilesAreVolatile(UserFilesAreVolatile) {
321 clearIDTables();
322 Diag.setSourceManager(this);
323}
324
325SourceManager::~SourceManager() {
326 // Delete FileEntry objects corresponding to content caches. Since the actual
327 // content cache objects are bump pointer allocated, we just have to run the
328 // dtors, but we call the deallocate method for completeness.
329 for (unsigned i = 0, e = MemBufferInfos.size(); i != e; ++i) {
330 if (MemBufferInfos[i]) {
331 MemBufferInfos[i]->~ContentCache();
332 ContentCacheAlloc.Deallocate(Ptr: MemBufferInfos[i]);
333 }
334 }
335 for (auto I = FileInfos.begin(), E = FileInfos.end(); I != E; ++I) {
336 if (I->second) {
337 I->second->~ContentCache();
338 ContentCacheAlloc.Deallocate(Ptr: I->second);
339 }
340 }
341}
342
343void SourceManager::clearIDTables() {
344 MainFileID = FileID();
345 LocalSLocEntryTable.clear();
346 LoadedSLocEntryTable.clear();
347 SLocEntryLoaded.clear();
348 SLocEntryOffsetLoaded.clear();
349 LastLineNoFileIDQuery = FileID();
350 LastLineNoContentCache = nullptr;
351 LastFileIDLookup = FileID();
352
353 if (LineTable)
354 LineTable->clear();
355
356 // Use up FileID #0 as an invalid expansion.
357 NextLocalOffset = 0;
358 CurrentLoadedOffset = MaxLoadedOffset;
359 createExpansionLoc(SpellingLoc: SourceLocation(), ExpansionLocStart: SourceLocation(), ExpansionLocEnd: SourceLocation(), Length: 1);
360}
361
362bool SourceManager::isMainFile(const FileEntry &SourceFile) {
363 assert(MainFileID.isValid() && "expected initialized SourceManager");
364 if (auto *FE = getFileEntryForID(FID: MainFileID))
365 return FE->getUID() == SourceFile.getUID();
366 return false;
367}
368
369void SourceManager::initializeForReplay(const SourceManager &Old) {
370 assert(MainFileID.isInvalid() && "expected uninitialized SourceManager");
371
372 auto CloneContentCache = [&](const ContentCache *Cache) -> ContentCache * {
373 auto *Clone = new (ContentCacheAlloc.Allocate<ContentCache>()) ContentCache;
374 Clone->OrigEntry = Cache->OrigEntry;
375 Clone->ContentsEntry = Cache->ContentsEntry;
376 Clone->BufferOverridden = Cache->BufferOverridden;
377 Clone->IsFileVolatile = Cache->IsFileVolatile;
378 Clone->IsTransient = Cache->IsTransient;
379 Clone->setUnownedBuffer(Cache->getBufferIfLoaded());
380 return Clone;
381 };
382
383 // Ensure all SLocEntries are loaded from the external source.
384 for (unsigned I = 0, N = Old.LoadedSLocEntryTable.size(); I != N; ++I)
385 if (!Old.SLocEntryLoaded[I])
386 Old.loadSLocEntry(Index: I, Invalid: nullptr);
387
388 // Inherit any content cache data from the old source manager.
389 for (auto &FileInfo : Old.FileInfos) {
390 SrcMgr::ContentCache *&Slot = FileInfos[FileInfo.first];
391 if (Slot)
392 continue;
393 Slot = CloneContentCache(FileInfo.second);
394 }
395}
396
397ContentCache &SourceManager::getOrCreateContentCache(FileEntryRef FileEnt,
398 bool isSystemFile) {
399 // Do we already have information about this file?
400 ContentCache *&Entry = FileInfos[FileEnt];
401 if (Entry)
402 return *Entry;
403
404 // Nope, create a new Cache entry.
405 Entry = ContentCacheAlloc.Allocate<ContentCache>();
406
407 if (OverriddenFilesInfo) {
408 // If the file contents are overridden with contents from another file,
409 // pass that file to ContentCache.
410 auto overI = OverriddenFilesInfo->OverriddenFiles.find(Val: FileEnt);
411 if (overI == OverriddenFilesInfo->OverriddenFiles.end())
412 new (Entry) ContentCache(FileEnt);
413 else
414 new (Entry) ContentCache(OverridenFilesKeepOriginalName ? FileEnt
415 : overI->second,
416 overI->second);
417 } else {
418 new (Entry) ContentCache(FileEnt);
419 }
420
421 Entry->IsFileVolatile = UserFilesAreVolatile && !isSystemFile;
422 Entry->IsTransient = FilesAreTransient;
423 Entry->BufferOverridden |= FileEnt.isNamedPipe();
424
425 return *Entry;
426}
427
428/// Create a new ContentCache for the specified memory buffer.
429/// This does no caching.
430ContentCache &SourceManager::createMemBufferContentCache(
431 std::unique_ptr<llvm::MemoryBuffer> Buffer) {
432 // Add a new ContentCache to the MemBufferInfos list and return it.
433 ContentCache *Entry = ContentCacheAlloc.Allocate<ContentCache>();
434 new (Entry) ContentCache();
435 MemBufferInfos.push_back(x: Entry);
436 Entry->setBuffer(std::move(Buffer));
437 return *Entry;
438}
439
440const SrcMgr::SLocEntry &SourceManager::loadSLocEntry(unsigned Index,
441 bool *Invalid) const {
442 return const_cast<SourceManager *>(this)->loadSLocEntry(Index, Invalid);
443}
444
445SrcMgr::SLocEntry &SourceManager::loadSLocEntry(unsigned Index, bool *Invalid) {
446 assert(!SLocEntryLoaded[Index]);
447 if (ExternalSLocEntries->ReadSLocEntry(ID: -(static_cast<int>(Index) + 2))) {
448 if (Invalid)
449 *Invalid = true;
450 // If the file of the SLocEntry changed we could still have loaded it.
451 if (!SLocEntryLoaded[Index]) {
452 // Try to recover; create a SLocEntry so the rest of clang can handle it.
453 if (!FakeSLocEntryForRecovery)
454 FakeSLocEntryForRecovery = std::make_unique<SLocEntry>(args: SLocEntry::get(
455 Offset: 0, FI: FileInfo::get(IL: SourceLocation(), Con&: getFakeContentCacheForRecovery(),
456 FileCharacter: SrcMgr::C_User, Filename: "")));
457 return *FakeSLocEntryForRecovery;
458 }
459 }
460
461 return LoadedSLocEntryTable[Index];
462}
463
464std::pair<int, SourceLocation::UIntTy>
465SourceManager::AllocateLoadedSLocEntries(unsigned NumSLocEntries,
466 SourceLocation::UIntTy TotalSize) {
467 assert(ExternalSLocEntries && "Don't have an external sloc source");
468 // Make sure we're not about to run out of source locations.
469 if (CurrentLoadedOffset < TotalSize ||
470 CurrentLoadedOffset - TotalSize < NextLocalOffset) {
471 return std::make_pair(x: 0, y: 0);
472 }
473 LoadedSLocEntryTable.resize(NewSize: LoadedSLocEntryTable.size() + NumSLocEntries);
474 SLocEntryLoaded.resize(N: LoadedSLocEntryTable.size());
475 SLocEntryOffsetLoaded.resize(N: LoadedSLocEntryTable.size());
476 CurrentLoadedOffset -= TotalSize;
477 updateSlocUsageStats();
478 int BaseID = -int(LoadedSLocEntryTable.size()) - 1;
479 LoadedSLocEntryAllocBegin.push_back(Elt: FileID::get(V: BaseID));
480 return std::make_pair(x&: BaseID, y&: CurrentLoadedOffset);
481}
482
483/// As part of recovering from missing or changed content, produce a
484/// fake, non-empty buffer.
485llvm::MemoryBufferRef SourceManager::getFakeBufferForRecovery() const {
486 if (!FakeBufferForRecovery)
487 FakeBufferForRecovery =
488 llvm::MemoryBuffer::getMemBuffer(InputData: "<<<INVALID BUFFER>>");
489
490 return *FakeBufferForRecovery;
491}
492
493/// As part of recovering from missing or changed content, produce a
494/// fake content cache.
495SrcMgr::ContentCache &SourceManager::getFakeContentCacheForRecovery() const {
496 if (!FakeContentCacheForRecovery) {
497 FakeContentCacheForRecovery = std::make_unique<SrcMgr::ContentCache>();
498 FakeContentCacheForRecovery->setUnownedBuffer(getFakeBufferForRecovery());
499 }
500 return *FakeContentCacheForRecovery;
501}
502
503/// Returns the previous in-order FileID or an invalid FileID if there
504/// is no previous one.
505FileID SourceManager::getPreviousFileID(FileID FID) const {
506 if (FID.isInvalid())
507 return FileID();
508
509 int ID = FID.ID;
510 if (ID == -1)
511 return FileID();
512
513 if (ID > 0) {
514 if (ID-1 == 0)
515 return FileID();
516 } else if (unsigned(-(ID-1) - 2) >= LoadedSLocEntryTable.size()) {
517 return FileID();
518 }
519
520 return FileID::get(V: ID-1);
521}
522
523/// Returns the next in-order FileID or an invalid FileID if there is
524/// no next one.
525FileID SourceManager::getNextFileID(FileID FID) const {
526 if (FID.isInvalid())
527 return FileID();
528
529 int ID = FID.ID;
530 if (ID > 0) {
531 if (unsigned(ID+1) >= local_sloc_entry_size())
532 return FileID();
533 } else if (ID+1 >= -1) {
534 return FileID();
535 }
536
537 return FileID::get(V: ID+1);
538}
539
540//===----------------------------------------------------------------------===//
541// Methods to create new FileID's and macro expansions.
542//===----------------------------------------------------------------------===//
543
544/// Create a new FileID that represents the specified file
545/// being \#included from the specified IncludePosition.
546FileID SourceManager::createFileID(FileEntryRef SourceFile,
547 SourceLocation IncludePos,
548 SrcMgr::CharacteristicKind FileCharacter,
549 int LoadedID,
550 SourceLocation::UIntTy LoadedOffset) {
551 SrcMgr::ContentCache &IR = getOrCreateContentCache(FileEnt: SourceFile,
552 isSystemFile: isSystem(CK: FileCharacter));
553
554 // If this is a named pipe, immediately load the buffer to ensure subsequent
555 // calls to ContentCache::getSize() are accurate.
556 if (IR.ContentsEntry->isNamedPipe())
557 (void)IR.getBufferOrNone(Diag, FM&: getFileManager(), Loc: SourceLocation());
558
559 return createFileIDImpl(File&: IR, Filename: SourceFile.getName(), IncludePos, DirCharacter: FileCharacter,
560 LoadedID, LoadedOffset);
561}
562
563/// Create a new FileID that represents the specified memory buffer.
564///
565/// This does no caching of the buffer and takes ownership of the
566/// MemoryBuffer, so only pass a MemoryBuffer to this once.
567FileID SourceManager::createFileID(std::unique_ptr<llvm::MemoryBuffer> Buffer,
568 SrcMgr::CharacteristicKind FileCharacter,
569 int LoadedID,
570 SourceLocation::UIntTy LoadedOffset,
571 SourceLocation IncludeLoc) {
572 StringRef Name = Buffer->getBufferIdentifier();
573 return createFileIDImpl(File&: createMemBufferContentCache(Buffer: std::move(Buffer)), Filename: Name,
574 IncludePos: IncludeLoc, DirCharacter: FileCharacter, LoadedID, LoadedOffset);
575}
576
577/// Create a new FileID that represents the specified memory buffer.
578///
579/// This does not take ownership of the MemoryBuffer. The memory buffer must
580/// outlive the SourceManager.
581FileID SourceManager::createFileID(const llvm::MemoryBufferRef &Buffer,
582 SrcMgr::CharacteristicKind FileCharacter,
583 int LoadedID,
584 SourceLocation::UIntTy LoadedOffset,
585 SourceLocation IncludeLoc) {
586 return createFileID(Buffer: llvm::MemoryBuffer::getMemBuffer(Ref: Buffer), FileCharacter,
587 LoadedID, LoadedOffset, IncludeLoc);
588}
589
590/// Get the FileID for \p SourceFile if it exists. Otherwise, create a
591/// new FileID for the \p SourceFile.
592FileID
593SourceManager::getOrCreateFileID(FileEntryRef SourceFile,
594 SrcMgr::CharacteristicKind FileCharacter) {
595 FileID ID = translateFile(SourceFile);
596 return ID.isValid() ? ID : createFileID(SourceFile, IncludePos: SourceLocation(),
597 FileCharacter);
598}
599
600/// createFileID - Create a new FileID for the specified ContentCache and
601/// include position. This works regardless of whether the ContentCache
602/// corresponds to a file or some other input source.
603FileID SourceManager::createFileIDImpl(ContentCache &File, StringRef Filename,
604 SourceLocation IncludePos,
605 SrcMgr::CharacteristicKind FileCharacter,
606 int LoadedID,
607 SourceLocation::UIntTy LoadedOffset) {
608 if (LoadedID < 0) {
609 assert(LoadedID != -1 && "Loading sentinel FileID");
610 unsigned Index = unsigned(-LoadedID) - 2;
611 assert(Index < LoadedSLocEntryTable.size() && "FileID out of range");
612 assert(!SLocEntryLoaded[Index] && "FileID already loaded");
613 LoadedSLocEntryTable[Index] = SLocEntry::get(
614 Offset: LoadedOffset, FI: FileInfo::get(IL: IncludePos, Con&: File, FileCharacter, Filename));
615 SLocEntryLoaded[Index] = SLocEntryOffsetLoaded[Index] = true;
616 return FileID::get(V: LoadedID);
617 }
618 unsigned FileSize = File.getSize();
619 if (!(NextLocalOffset + FileSize + 1 > NextLocalOffset &&
620 NextLocalOffset + FileSize + 1 <= CurrentLoadedOffset)) {
621 Diag.Report(Loc: IncludePos, DiagID: diag::err_sloc_space_too_large);
622 noteSLocAddressSpaceUsage(Diag);
623 return FileID();
624 }
625 LocalSLocEntryTable.push_back(
626 Elt: SLocEntry::get(Offset: NextLocalOffset,
627 FI: FileInfo::get(IL: IncludePos, Con&: File, FileCharacter, Filename)));
628 // We do a +1 here because we want a SourceLocation that means "the end of the
629 // file", e.g. for the "no newline at the end of the file" diagnostic.
630 NextLocalOffset += FileSize + 1;
631 updateSlocUsageStats();
632
633 // Set LastFileIDLookup to the newly created file. The next getFileID call is
634 // almost guaranteed to be from that file.
635 FileID FID = FileID::get(V: LocalSLocEntryTable.size()-1);
636 return LastFileIDLookup = FID;
637}
638
639SourceLocation SourceManager::createMacroArgExpansionLoc(
640 SourceLocation SpellingLoc, SourceLocation ExpansionLoc, unsigned Length) {
641 ExpansionInfo Info = ExpansionInfo::createForMacroArg(SpellingLoc,
642 ExpansionLoc);
643 return createExpansionLocImpl(Expansion: Info, Length);
644}
645
646SourceLocation SourceManager::createExpansionLoc(
647 SourceLocation SpellingLoc, SourceLocation ExpansionLocStart,
648 SourceLocation ExpansionLocEnd, unsigned Length,
649 bool ExpansionIsTokenRange, int LoadedID,
650 SourceLocation::UIntTy LoadedOffset) {
651 ExpansionInfo Info = ExpansionInfo::create(
652 SpellingLoc, Start: ExpansionLocStart, End: ExpansionLocEnd, ExpansionIsTokenRange);
653 return createExpansionLocImpl(Expansion: Info, Length, LoadedID, LoadedOffset);
654}
655
656SourceLocation SourceManager::createTokenSplitLoc(SourceLocation Spelling,
657 SourceLocation TokenStart,
658 SourceLocation TokenEnd) {
659 assert(getFileID(TokenStart) == getFileID(TokenEnd) &&
660 "token spans multiple files");
661 return createExpansionLocImpl(
662 Expansion: ExpansionInfo::createForTokenSplit(SpellingLoc: Spelling, Start: TokenStart, End: TokenEnd),
663 Length: TokenEnd.getOffset() - TokenStart.getOffset());
664}
665
666SourceLocation
667SourceManager::createExpansionLocImpl(const ExpansionInfo &Info,
668 unsigned Length, int LoadedID,
669 SourceLocation::UIntTy LoadedOffset) {
670 if (LoadedID < 0) {
671 assert(LoadedID != -1 && "Loading sentinel FileID");
672 unsigned Index = unsigned(-LoadedID) - 2;
673 assert(Index < LoadedSLocEntryTable.size() && "FileID out of range");
674 assert(!SLocEntryLoaded[Index] && "FileID already loaded");
675 LoadedSLocEntryTable[Index] = SLocEntry::get(Offset: LoadedOffset, Expansion: Info);
676 SLocEntryLoaded[Index] = SLocEntryOffsetLoaded[Index] = true;
677 return SourceLocation::getMacroLoc(ID: LoadedOffset);
678 }
679 LocalSLocEntryTable.push_back(Elt: SLocEntry::get(Offset: NextLocalOffset, Expansion: Info));
680 if (NextLocalOffset + Length + 1 <= NextLocalOffset ||
681 NextLocalOffset + Length + 1 > CurrentLoadedOffset) {
682 Diag.Report(Loc: SourceLocation(), DiagID: diag::err_sloc_space_too_large);
683 // FIXME: call `noteSLocAddressSpaceUsage` to report details to users and
684 // use a source location from `Info` to point at an error.
685 // Currently, both cause Clang to run indefinitely, this needs to be fixed.
686 // FIXME: return an error instead of crashing. Returning invalid source
687 // locations causes compiler to run indefinitely.
688 llvm::report_fatal_error(reason: "ran out of source locations");
689 }
690 // See createFileID for that +1.
691 NextLocalOffset += Length + 1;
692 updateSlocUsageStats();
693 return SourceLocation::getMacroLoc(ID: NextLocalOffset - (Length + 1));
694}
695
696std::optional<llvm::MemoryBufferRef>
697SourceManager::getMemoryBufferForFileOrNone(FileEntryRef File) {
698 SrcMgr::ContentCache &IR = getOrCreateContentCache(FileEnt: File);
699 return IR.getBufferOrNone(Diag, FM&: getFileManager(), Loc: SourceLocation());
700}
701
702void SourceManager::overrideFileContents(
703 FileEntryRef SourceFile, std::unique_ptr<llvm::MemoryBuffer> Buffer) {
704 SrcMgr::ContentCache &IR = getOrCreateContentCache(FileEnt: SourceFile);
705
706 IR.setBuffer(std::move(Buffer));
707 IR.BufferOverridden = true;
708
709 getOverriddenFilesInfo().OverriddenFilesWithBuffer.insert(V: SourceFile);
710}
711
712void SourceManager::overrideFileContents(const FileEntry *SourceFile,
713 FileEntryRef NewFile) {
714 assert(SourceFile->getSize() == NewFile.getSize() &&
715 "Different sizes, use the FileManager to create a virtual file with "
716 "the correct size");
717 assert(FileInfos.find_as(SourceFile) == FileInfos.end() &&
718 "This function should be called at the initialization stage, before "
719 "any parsing occurs.");
720 // FileEntryRef is not default-constructible.
721 auto Pair = getOverriddenFilesInfo().OverriddenFiles.insert(
722 KV: std::make_pair(x&: SourceFile, y&: NewFile));
723 if (!Pair.second)
724 Pair.first->second = NewFile;
725}
726
727OptionalFileEntryRef
728SourceManager::bypassFileContentsOverride(FileEntryRef File) {
729 assert(isFileOverridden(&File.getFileEntry()));
730 OptionalFileEntryRef BypassFile = FileMgr.getBypassFile(VFE: File);
731
732 // If the file can't be found in the FS, give up.
733 if (!BypassFile)
734 return std::nullopt;
735
736 (void)getOrCreateContentCache(FileEnt: *BypassFile);
737 return BypassFile;
738}
739
740void SourceManager::setFileIsTransient(FileEntryRef File) {
741 getOrCreateContentCache(FileEnt: File).IsTransient = true;
742}
743
744std::optional<StringRef>
745SourceManager::getNonBuiltinFilenameForID(FileID FID) const {
746 if (const SrcMgr::SLocEntry *Entry = getSLocEntryForFile(FID))
747 if (Entry->getFile().getContentCache().OrigEntry)
748 return Entry->getFile().getName();
749 return std::nullopt;
750}
751
752StringRef SourceManager::getBufferData(FileID FID, bool *Invalid) const {
753 auto B = getBufferDataOrNone(FID);
754 if (Invalid)
755 *Invalid = !B;
756 return B ? *B : "<<<<<INVALID SOURCE LOCATION>>>>>";
757}
758
759std::optional<StringRef>
760SourceManager::getBufferDataIfLoaded(FileID FID) const {
761 if (const SrcMgr::SLocEntry *Entry = getSLocEntryForFile(FID))
762 return Entry->getFile().getContentCache().getBufferDataIfLoaded();
763 return std::nullopt;
764}
765
766std::optional<StringRef> SourceManager::getBufferDataOrNone(FileID FID) const {
767 if (const SrcMgr::SLocEntry *Entry = getSLocEntryForFile(FID))
768 if (auto B = Entry->getFile().getContentCache().getBufferOrNone(
769 Diag, FM&: getFileManager(), Loc: SourceLocation()))
770 return B->getBuffer();
771 return std::nullopt;
772}
773
774//===----------------------------------------------------------------------===//
775// SourceLocation manipulation methods.
776//===----------------------------------------------------------------------===//
777
778/// Return the FileID for a SourceLocation.
779///
780/// This is the cache-miss path of getFileID. Not as hot as that function, but
781/// still very important. It is responsible for finding the entry in the
782/// SLocEntry tables that contains the specified location.
783FileID SourceManager::getFileIDSlow(SourceLocation::UIntTy SLocOffset) const {
784 if (!SLocOffset)
785 return FileID::get(V: 0);
786
787 // Now it is time to search for the correct file. See where the SLocOffset
788 // sits in the global view and consult local or loaded buffers for it.
789 if (SLocOffset < NextLocalOffset)
790 return getFileIDLocal(SLocOffset);
791 return getFileIDLoaded(SLocOffset);
792}
793
794/// Return the FileID for a SourceLocation with a low offset.
795///
796/// This function knows that the SourceLocation is in a local buffer, not a
797/// loaded one.
798FileID SourceManager::getFileIDLocal(SourceLocation::UIntTy SLocOffset) const {
799 assert(SLocOffset < NextLocalOffset && "Bad function choice");
800
801 // After the first and second level caches, I see two common sorts of
802 // behavior: 1) a lot of searched FileID's are "near" the cached file
803 // location or are "near" the cached expansion location. 2) others are just
804 // completely random and may be a very long way away.
805 //
806 // To handle this, we do a linear search for up to 8 steps to catch #1 quickly
807 // then we fall back to a less cache efficient, but more scalable, binary
808 // search to find the location.
809
810 // See if this is near the file point - worst case we start scanning from the
811 // most newly created FileID.
812
813 // LessIndex - This is the lower bound of the range that we're searching.
814 // We know that the offset corresponding to the FileID is less than
815 // SLocOffset.
816 unsigned LessIndex = 0;
817 // upper bound of the search range.
818 unsigned GreaterIndex = LocalSLocEntryTable.size();
819 if (LastFileIDLookup.ID >= 0) {
820 // Use the LastFileIDLookup to prune the search space.
821 if (LocalSLocEntryTable[LastFileIDLookup.ID].getOffset() < SLocOffset)
822 LessIndex = LastFileIDLookup.ID;
823 else
824 GreaterIndex = LastFileIDLookup.ID;
825 }
826
827 // Find the FileID that contains this.
828 unsigned NumProbes = 0;
829 while (true) {
830 --GreaterIndex;
831 assert(GreaterIndex < LocalSLocEntryTable.size());
832 if (LocalSLocEntryTable[GreaterIndex].getOffset() <= SLocOffset) {
833 FileID Res = FileID::get(V: int(GreaterIndex));
834 // Remember it. We have good locality across FileID lookups.
835 LastFileIDLookup = Res;
836 NumLinearScans += NumProbes+1;
837 return Res;
838 }
839 if (++NumProbes == 8)
840 break;
841 }
842
843 NumProbes = 0;
844 while (true) {
845 unsigned MiddleIndex = (GreaterIndex-LessIndex)/2+LessIndex;
846 SourceLocation::UIntTy MidOffset =
847 getLocalSLocEntry(Index: MiddleIndex).getOffset();
848
849 ++NumProbes;
850
851 // If the offset of the midpoint is too large, chop the high side of the
852 // range to the midpoint.
853 if (MidOffset > SLocOffset) {
854 GreaterIndex = MiddleIndex;
855 continue;
856 }
857
858 // If the middle index contains the value, succeed and return.
859 if (MiddleIndex + 1 == LocalSLocEntryTable.size() ||
860 SLocOffset < getLocalSLocEntry(Index: MiddleIndex + 1).getOffset()) {
861 FileID Res = FileID::get(V: MiddleIndex);
862
863 // Remember it. We have good locality across FileID lookups.
864 LastFileIDLookup = Res;
865 NumBinaryProbes += NumProbes;
866 return Res;
867 }
868
869 // Otherwise, move the low-side up to the middle index.
870 LessIndex = MiddleIndex;
871 }
872}
873
874/// Return the FileID for a SourceLocation with a high offset.
875///
876/// This function knows that the SourceLocation is in a loaded buffer, not a
877/// local one.
878FileID SourceManager::getFileIDLoaded(SourceLocation::UIntTy SLocOffset) const {
879 if (SLocOffset < CurrentLoadedOffset) {
880 assert(0 && "Invalid SLocOffset or bad function choice");
881 return FileID();
882 }
883
884 return FileID::get(V: ExternalSLocEntries->getSLocEntryID(SLocOffset));
885}
886
887SourceLocation SourceManager::
888getExpansionLocSlowCase(SourceLocation Loc) const {
889 do {
890 // Note: If Loc indicates an offset into a token that came from a macro
891 // expansion (e.g. the 5th character of the token) we do not want to add
892 // this offset when going to the expansion location. The expansion
893 // location is the macro invocation, which the offset has nothing to do
894 // with. This is unlike when we get the spelling loc, because the offset
895 // directly correspond to the token whose spelling we're inspecting.
896 Loc = getSLocEntry(FID: getFileID(SpellingLoc: Loc)).getExpansion().getExpansionLocStart();
897 } while (!Loc.isFileID());
898
899 return Loc;
900}
901
902SourceLocation SourceManager::getSpellingLocSlowCase(SourceLocation Loc) const {
903 do {
904 std::pair<FileID, unsigned> LocInfo = getDecomposedLoc(Loc);
905 Loc = getSLocEntry(FID: LocInfo.first).getExpansion().getSpellingLoc();
906 Loc = Loc.getLocWithOffset(Offset: LocInfo.second);
907 } while (!Loc.isFileID());
908 return Loc;
909}
910
911SourceLocation SourceManager::getFileLocSlowCase(SourceLocation Loc) const {
912 do {
913 if (isMacroArgExpansion(Loc))
914 Loc = getImmediateSpellingLoc(Loc);
915 else
916 Loc = getImmediateExpansionRange(Loc).getBegin();
917 } while (!Loc.isFileID());
918 return Loc;
919}
920
921
922std::pair<FileID, unsigned>
923SourceManager::getDecomposedExpansionLocSlowCase(
924 const SrcMgr::SLocEntry *E) const {
925 // If this is an expansion record, walk through all the expansion points.
926 FileID FID;
927 SourceLocation Loc;
928 unsigned Offset;
929 do {
930 Loc = E->getExpansion().getExpansionLocStart();
931
932 FID = getFileID(SpellingLoc: Loc);
933 E = &getSLocEntry(FID);
934 Offset = Loc.getOffset()-E->getOffset();
935 } while (!Loc.isFileID());
936
937 return std::make_pair(x&: FID, y&: Offset);
938}
939
940std::pair<FileID, unsigned>
941SourceManager::getDecomposedSpellingLocSlowCase(const SrcMgr::SLocEntry *E,
942 unsigned Offset) const {
943 // If this is an expansion record, walk through all the expansion points.
944 FileID FID;
945 SourceLocation Loc;
946 do {
947 Loc = E->getExpansion().getSpellingLoc();
948 Loc = Loc.getLocWithOffset(Offset);
949
950 FID = getFileID(SpellingLoc: Loc);
951 E = &getSLocEntry(FID);
952 Offset = Loc.getOffset()-E->getOffset();
953 } while (!Loc.isFileID());
954
955 return std::make_pair(x&: FID, y&: Offset);
956}
957
958/// getImmediateSpellingLoc - Given a SourceLocation object, return the
959/// spelling location referenced by the ID. This is the first level down
960/// towards the place where the characters that make up the lexed token can be
961/// found. This should not generally be used by clients.
962SourceLocation SourceManager::getImmediateSpellingLoc(SourceLocation Loc) const{
963 if (Loc.isFileID()) return Loc;
964 std::pair<FileID, unsigned> LocInfo = getDecomposedLoc(Loc);
965 Loc = getSLocEntry(FID: LocInfo.first).getExpansion().getSpellingLoc();
966 return Loc.getLocWithOffset(Offset: LocInfo.second);
967}
968
969/// Return the filename of the file containing a SourceLocation.
970StringRef SourceManager::getFilename(SourceLocation SpellingLoc) const {
971 if (OptionalFileEntryRef F = getFileEntryRefForID(FID: getFileID(SpellingLoc)))
972 return F->getName();
973 return StringRef();
974}
975
976/// getImmediateExpansionRange - Loc is required to be an expansion location.
977/// Return the start/end of the expansion information.
978CharSourceRange
979SourceManager::getImmediateExpansionRange(SourceLocation Loc) const {
980 assert(Loc.isMacroID() && "Not a macro expansion loc!");
981 const ExpansionInfo &Expansion = getSLocEntry(FID: getFileID(SpellingLoc: Loc)).getExpansion();
982 return Expansion.getExpansionLocRange();
983}
984
985SourceLocation SourceManager::getTopMacroCallerLoc(SourceLocation Loc) const {
986 while (isMacroArgExpansion(Loc))
987 Loc = getImmediateSpellingLoc(Loc);
988 return Loc;
989}
990
991/// getExpansionRange - Given a SourceLocation object, return the range of
992/// tokens covered by the expansion in the ultimate file.
993CharSourceRange SourceManager::getExpansionRange(SourceLocation Loc) const {
994 if (Loc.isFileID())
995 return CharSourceRange(SourceRange(Loc, Loc), true);
996
997 CharSourceRange Res = getImmediateExpansionRange(Loc);
998
999 // Fully resolve the start and end locations to their ultimate expansion
1000 // points.
1001 while (!Res.getBegin().isFileID())
1002 Res.setBegin(getImmediateExpansionRange(Loc: Res.getBegin()).getBegin());
1003 while (!Res.getEnd().isFileID()) {
1004 CharSourceRange EndRange = getImmediateExpansionRange(Loc: Res.getEnd());
1005 Res.setEnd(EndRange.getEnd());
1006 Res.setTokenRange(EndRange.isTokenRange());
1007 }
1008 return Res;
1009}
1010
1011bool SourceManager::isMacroArgExpansion(SourceLocation Loc,
1012 SourceLocation *StartLoc) const {
1013 if (!Loc.isMacroID()) return false;
1014
1015 FileID FID = getFileID(SpellingLoc: Loc);
1016 const SrcMgr::ExpansionInfo &Expansion = getSLocEntry(FID).getExpansion();
1017 if (!Expansion.isMacroArgExpansion()) return false;
1018
1019 if (StartLoc)
1020 *StartLoc = Expansion.getExpansionLocStart();
1021 return true;
1022}
1023
1024bool SourceManager::isMacroBodyExpansion(SourceLocation Loc) const {
1025 if (!Loc.isMacroID()) return false;
1026
1027 FileID FID = getFileID(SpellingLoc: Loc);
1028 const SrcMgr::ExpansionInfo &Expansion = getSLocEntry(FID).getExpansion();
1029 return Expansion.isMacroBodyExpansion();
1030}
1031
1032bool SourceManager::isAtStartOfImmediateMacroExpansion(SourceLocation Loc,
1033 SourceLocation *MacroBegin) const {
1034 assert(Loc.isValid() && Loc.isMacroID() && "Expected a valid macro loc");
1035
1036 std::pair<FileID, unsigned> DecompLoc = getDecomposedLoc(Loc);
1037 if (DecompLoc.second > 0)
1038 return false; // Does not point at the start of expansion range.
1039
1040 bool Invalid = false;
1041 const SrcMgr::ExpansionInfo &ExpInfo =
1042 getSLocEntry(FID: DecompLoc.first, Invalid: &Invalid).getExpansion();
1043 if (Invalid)
1044 return false;
1045 SourceLocation ExpLoc = ExpInfo.getExpansionLocStart();
1046
1047 if (ExpInfo.isMacroArgExpansion()) {
1048 // For macro argument expansions, check if the previous FileID is part of
1049 // the same argument expansion, in which case this Loc is not at the
1050 // beginning of the expansion.
1051 FileID PrevFID = getPreviousFileID(FID: DecompLoc.first);
1052 if (!PrevFID.isInvalid()) {
1053 const SrcMgr::SLocEntry &PrevEntry = getSLocEntry(FID: PrevFID, Invalid: &Invalid);
1054 if (Invalid)
1055 return false;
1056 if (PrevEntry.isExpansion() &&
1057 PrevEntry.getExpansion().getExpansionLocStart() == ExpLoc)
1058 return false;
1059 }
1060 }
1061
1062 if (MacroBegin)
1063 *MacroBegin = ExpLoc;
1064 return true;
1065}
1066
1067bool SourceManager::isAtEndOfImmediateMacroExpansion(SourceLocation Loc,
1068 SourceLocation *MacroEnd) const {
1069 assert(Loc.isValid() && Loc.isMacroID() && "Expected a valid macro loc");
1070
1071 FileID FID = getFileID(SpellingLoc: Loc);
1072 SourceLocation NextLoc = Loc.getLocWithOffset(Offset: 1);
1073 if (isInFileID(Loc: NextLoc, FID))
1074 return false; // Does not point at the end of expansion range.
1075
1076 bool Invalid = false;
1077 const SrcMgr::ExpansionInfo &ExpInfo =
1078 getSLocEntry(FID, Invalid: &Invalid).getExpansion();
1079 if (Invalid)
1080 return false;
1081
1082 if (ExpInfo.isMacroArgExpansion()) {
1083 // For macro argument expansions, check if the next FileID is part of the
1084 // same argument expansion, in which case this Loc is not at the end of the
1085 // expansion.
1086 FileID NextFID = getNextFileID(FID);
1087 if (!NextFID.isInvalid()) {
1088 const SrcMgr::SLocEntry &NextEntry = getSLocEntry(FID: NextFID, Invalid: &Invalid);
1089 if (Invalid)
1090 return false;
1091 if (NextEntry.isExpansion() &&
1092 NextEntry.getExpansion().getExpansionLocStart() ==
1093 ExpInfo.getExpansionLocStart())
1094 return false;
1095 }
1096 }
1097
1098 if (MacroEnd)
1099 *MacroEnd = ExpInfo.getExpansionLocEnd();
1100 return true;
1101}
1102
1103//===----------------------------------------------------------------------===//
1104// Queries about the code at a SourceLocation.
1105//===----------------------------------------------------------------------===//
1106
1107/// getCharacterData - Return a pointer to the start of the specified location
1108/// in the appropriate MemoryBuffer.
1109const char *SourceManager::getCharacterData(SourceLocation SL,
1110 bool *Invalid) const {
1111 // Note that this is a hot function in the getSpelling() path, which is
1112 // heavily used by -E mode.
1113 std::pair<FileID, unsigned> LocInfo = getDecomposedSpellingLoc(Loc: SL);
1114
1115 // Note that calling 'getBuffer()' may lazily page in a source file.
1116 bool CharDataInvalid = false;
1117 const SLocEntry &Entry = getSLocEntry(FID: LocInfo.first, Invalid: &CharDataInvalid);
1118 if (CharDataInvalid || !Entry.isFile()) {
1119 if (Invalid)
1120 *Invalid = true;
1121
1122 return "<<<<INVALID BUFFER>>>>";
1123 }
1124 std::optional<llvm::MemoryBufferRef> Buffer =
1125 Entry.getFile().getContentCache().getBufferOrNone(Diag, FM&: getFileManager(),
1126 Loc: SourceLocation());
1127 if (Invalid)
1128 *Invalid = !Buffer;
1129 return Buffer ? Buffer->getBufferStart() + LocInfo.second
1130 : "<<<<INVALID BUFFER>>>>";
1131}
1132
1133/// getColumnNumber - Return the column # for the specified file position.
1134/// this is significantly cheaper to compute than the line number.
1135unsigned SourceManager::getColumnNumber(FileID FID, unsigned FilePos,
1136 bool *Invalid) const {
1137 std::optional<llvm::MemoryBufferRef> MemBuf = getBufferOrNone(FID);
1138 if (Invalid)
1139 *Invalid = !MemBuf;
1140
1141 if (!MemBuf)
1142 return 1;
1143
1144 // It is okay to request a position just past the end of the buffer.
1145 if (FilePos > MemBuf->getBufferSize()) {
1146 if (Invalid)
1147 *Invalid = true;
1148 return 1;
1149 }
1150
1151 const char *Buf = MemBuf->getBufferStart();
1152 // See if we just calculated the line number for this FilePos and can use
1153 // that to lookup the start of the line instead of searching for it.
1154 if (LastLineNoFileIDQuery == FID && LastLineNoContentCache->SourceLineCache &&
1155 LastLineNoResult < LastLineNoContentCache->SourceLineCache.size()) {
1156 const unsigned *SourceLineCache =
1157 LastLineNoContentCache->SourceLineCache.begin();
1158 unsigned LineStart = SourceLineCache[LastLineNoResult - 1];
1159 unsigned LineEnd = SourceLineCache[LastLineNoResult];
1160 if (FilePos >= LineStart && FilePos < LineEnd) {
1161 // LineEnd is the LineStart of the next line.
1162 // A line ends with separator LF or CR+LF on Windows.
1163 // FilePos might point to the last separator,
1164 // but we need a column number at most 1 + the last column.
1165 if (FilePos + 1 == LineEnd && FilePos > LineStart) {
1166 if (Buf[FilePos - 1] == '\r' || Buf[FilePos - 1] == '\n')
1167 --FilePos;
1168 }
1169 return FilePos - LineStart + 1;
1170 }
1171 }
1172
1173 unsigned LineStart = FilePos;
1174 while (LineStart && Buf[LineStart-1] != '\n' && Buf[LineStart-1] != '\r')
1175 --LineStart;
1176 return FilePos-LineStart+1;
1177}
1178
1179// isInvalid - Return the result of calling loc.isInvalid(), and
1180// if Invalid is not null, set its value to same.
1181template<typename LocType>
1182static bool isInvalid(LocType Loc, bool *Invalid) {
1183 bool MyInvalid = Loc.isInvalid();
1184 if (Invalid)
1185 *Invalid = MyInvalid;
1186 return MyInvalid;
1187}
1188
1189unsigned SourceManager::getSpellingColumnNumber(SourceLocation Loc,
1190 bool *Invalid) const {
1191 if (isInvalid(Loc, Invalid)) return 0;
1192 std::pair<FileID, unsigned> LocInfo = getDecomposedSpellingLoc(Loc);
1193 return getColumnNumber(FID: LocInfo.first, FilePos: LocInfo.second, Invalid);
1194}
1195
1196unsigned SourceManager::getExpansionColumnNumber(SourceLocation Loc,
1197 bool *Invalid) const {
1198 if (isInvalid(Loc, Invalid)) return 0;
1199 std::pair<FileID, unsigned> LocInfo = getDecomposedExpansionLoc(Loc);
1200 return getColumnNumber(FID: LocInfo.first, FilePos: LocInfo.second, Invalid);
1201}
1202
1203unsigned SourceManager::getPresumedColumnNumber(SourceLocation Loc,
1204 bool *Invalid) const {
1205 PresumedLoc PLoc = getPresumedLoc(Loc);
1206 if (isInvalid(Loc: PLoc, Invalid)) return 0;
1207 return PLoc.getColumn();
1208}
1209
1210// Check if mutli-byte word x has bytes between m and n, included. This may also
1211// catch bytes equal to n + 1.
1212// The returned value holds a 0x80 at each byte position that holds a match.
1213// see http://graphics.stanford.edu/~seander/bithacks.html#HasBetweenInWord
1214template <class T>
1215static constexpr inline T likelyhasbetween(T x, unsigned char m,
1216 unsigned char n) {
1217 return ((x - ~static_cast<T>(0) / 255 * (n + 1)) & ~x &
1218 ((x & ~static_cast<T>(0) / 255 * 127) +
1219 (~static_cast<T>(0) / 255 * (127 - (m - 1))))) &
1220 ~static_cast<T>(0) / 255 * 128;
1221}
1222
1223LineOffsetMapping LineOffsetMapping::get(llvm::MemoryBufferRef Buffer,
1224 llvm::BumpPtrAllocator &Alloc) {
1225
1226 // Find the file offsets of all of the *physical* source lines. This does
1227 // not look at trigraphs, escaped newlines, or anything else tricky.
1228 SmallVector<unsigned, 256> LineOffsets;
1229
1230 // Line #1 starts at char 0.
1231 LineOffsets.push_back(Elt: 0);
1232
1233 const unsigned char *Start = (const unsigned char *)Buffer.getBufferStart();
1234 const unsigned char *End = (const unsigned char *)Buffer.getBufferEnd();
1235 const unsigned char *Buf = Start;
1236
1237 uint64_t Word;
1238
1239 // scan sizeof(Word) bytes at a time for new lines.
1240 // This is much faster than scanning each byte independently.
1241 if ((unsigned long)(End - Start) > sizeof(Word)) {
1242 do {
1243 Word = llvm::support::endian::read64(P: Buf, E: llvm::endianness::little);
1244 // no new line => jump over sizeof(Word) bytes.
1245 auto Mask = likelyhasbetween(x: Word, m: '\n', n: '\r');
1246 if (!Mask) {
1247 Buf += sizeof(Word);
1248 continue;
1249 }
1250
1251 // At that point, Mask contains 0x80 set at each byte that holds a value
1252 // in [\n, \r + 1 [
1253
1254 // Scan for the next newline - it's very likely there's one.
1255 unsigned N = llvm::countr_zero(Val: Mask) - 7; // -7 because 0x80 is the marker
1256 Word >>= N;
1257 Buf += N / 8 + 1;
1258 unsigned char Byte = Word;
1259 switch (Byte) {
1260 case '\r':
1261 // If this is \r\n, skip both characters.
1262 if (*Buf == '\n') {
1263 ++Buf;
1264 }
1265 [[fallthrough]];
1266 case '\n':
1267 LineOffsets.push_back(Elt: Buf - Start);
1268 };
1269 } while (Buf < End - sizeof(Word) - 1);
1270 }
1271
1272 // Handle tail using a regular check.
1273 while (Buf < End) {
1274 if (*Buf == '\n') {
1275 LineOffsets.push_back(Elt: Buf - Start + 1);
1276 } else if (*Buf == '\r') {
1277 // If this is \r\n, skip both characters.
1278 if (Buf + 1 < End && Buf[1] == '\n') {
1279 ++Buf;
1280 }
1281 LineOffsets.push_back(Elt: Buf - Start + 1);
1282 }
1283 ++Buf;
1284 }
1285
1286 return LineOffsetMapping(LineOffsets, Alloc);
1287}
1288
1289LineOffsetMapping::LineOffsetMapping(ArrayRef<unsigned> LineOffsets,
1290 llvm::BumpPtrAllocator &Alloc)
1291 : Storage(Alloc.Allocate<unsigned>(Num: LineOffsets.size() + 1)) {
1292 Storage[0] = LineOffsets.size();
1293 std::copy(LineOffsets.begin(), LineOffsets.end(), Storage + 1);
1294}
1295
1296/// getLineNumber - Given a SourceLocation, return the spelling line number
1297/// for the position indicated. This requires building and caching a table of
1298/// line offsets for the MemoryBuffer, so this is not cheap: use only when
1299/// about to emit a diagnostic.
1300unsigned SourceManager::getLineNumber(FileID FID, unsigned FilePos,
1301 bool *Invalid) const {
1302 if (FID.isInvalid()) {
1303 if (Invalid)
1304 *Invalid = true;
1305 return 1;
1306 }
1307
1308 const ContentCache *Content;
1309 if (LastLineNoFileIDQuery == FID)
1310 Content = LastLineNoContentCache;
1311 else {
1312 bool MyInvalid = false;
1313 const SLocEntry &Entry = getSLocEntry(FID, Invalid: &MyInvalid);
1314 if (MyInvalid || !Entry.isFile()) {
1315 if (Invalid)
1316 *Invalid = true;
1317 return 1;
1318 }
1319
1320 Content = &Entry.getFile().getContentCache();
1321 }
1322
1323 // If this is the first use of line information for this buffer, compute the
1324 // SourceLineCache for it on demand.
1325 if (!Content->SourceLineCache) {
1326 std::optional<llvm::MemoryBufferRef> Buffer =
1327 Content->getBufferOrNone(Diag, FM&: getFileManager(), Loc: SourceLocation());
1328 if (Invalid)
1329 *Invalid = !Buffer;
1330 if (!Buffer)
1331 return 1;
1332
1333 Content->SourceLineCache =
1334 LineOffsetMapping::get(Buffer: *Buffer, Alloc&: ContentCacheAlloc);
1335 } else if (Invalid)
1336 *Invalid = false;
1337
1338 // Okay, we know we have a line number table. Do a binary search to find the
1339 // line number that this character position lands on.
1340 const unsigned *SourceLineCache = Content->SourceLineCache.begin();
1341 const unsigned *SourceLineCacheStart = SourceLineCache;
1342 const unsigned *SourceLineCacheEnd = Content->SourceLineCache.end();
1343
1344 unsigned QueriedFilePos = FilePos+1;
1345
1346 // FIXME: I would like to be convinced that this code is worth being as
1347 // complicated as it is, binary search isn't that slow.
1348 //
1349 // If it is worth being optimized, then in my opinion it could be more
1350 // performant, simpler, and more obviously correct by just "galloping" outward
1351 // from the queried file position. In fact, this could be incorporated into a
1352 // generic algorithm such as lower_bound_with_hint.
1353 //
1354 // If someone gives me a test case where this matters, and I will do it! - DWD
1355
1356 // If the previous query was to the same file, we know both the file pos from
1357 // that query and the line number returned. This allows us to narrow the
1358 // search space from the entire file to something near the match.
1359 if (LastLineNoFileIDQuery == FID) {
1360 if (QueriedFilePos >= LastLineNoFilePos) {
1361 // FIXME: Potential overflow?
1362 SourceLineCache = SourceLineCache+LastLineNoResult-1;
1363
1364 // The query is likely to be nearby the previous one. Here we check to
1365 // see if it is within 5, 10 or 20 lines. It can be far away in cases
1366 // where big comment blocks and vertical whitespace eat up lines but
1367 // contribute no tokens.
1368 if (SourceLineCache+5 < SourceLineCacheEnd) {
1369 if (SourceLineCache[5] > QueriedFilePos)
1370 SourceLineCacheEnd = SourceLineCache+5;
1371 else if (SourceLineCache+10 < SourceLineCacheEnd) {
1372 if (SourceLineCache[10] > QueriedFilePos)
1373 SourceLineCacheEnd = SourceLineCache+10;
1374 else if (SourceLineCache+20 < SourceLineCacheEnd) {
1375 if (SourceLineCache[20] > QueriedFilePos)
1376 SourceLineCacheEnd = SourceLineCache+20;
1377 }
1378 }
1379 }
1380 } else {
1381 if (LastLineNoResult < Content->SourceLineCache.size())
1382 SourceLineCacheEnd = SourceLineCache+LastLineNoResult+1;
1383 }
1384 }
1385
1386 const unsigned *Pos =
1387 std::lower_bound(SourceLineCache, SourceLineCacheEnd, QueriedFilePos);
1388 unsigned LineNo = Pos-SourceLineCacheStart;
1389
1390 LastLineNoFileIDQuery = FID;
1391 LastLineNoContentCache = Content;
1392 LastLineNoFilePos = QueriedFilePos;
1393 LastLineNoResult = LineNo;
1394 return LineNo;
1395}
1396
1397unsigned SourceManager::getSpellingLineNumber(SourceLocation Loc,
1398 bool *Invalid) const {
1399 if (isInvalid(Loc, Invalid)) return 0;
1400 std::pair<FileID, unsigned> LocInfo = getDecomposedSpellingLoc(Loc);
1401 return getLineNumber(FID: LocInfo.first, FilePos: LocInfo.second);
1402}
1403unsigned SourceManager::getExpansionLineNumber(SourceLocation Loc,
1404 bool *Invalid) const {
1405 if (isInvalid(Loc, Invalid)) return 0;
1406 std::pair<FileID, unsigned> LocInfo = getDecomposedExpansionLoc(Loc);
1407 return getLineNumber(FID: LocInfo.first, FilePos: LocInfo.second);
1408}
1409unsigned SourceManager::getPresumedLineNumber(SourceLocation Loc,
1410 bool *Invalid) const {
1411 PresumedLoc PLoc = getPresumedLoc(Loc);
1412 if (isInvalid(Loc: PLoc, Invalid)) return 0;
1413 return PLoc.getLine();
1414}
1415
1416/// getFileCharacteristic - return the file characteristic of the specified
1417/// source location, indicating whether this is a normal file, a system
1418/// header, or an "implicit extern C" system header.
1419///
1420/// This state can be modified with flags on GNU linemarker directives like:
1421/// # 4 "foo.h" 3
1422/// which changes all source locations in the current file after that to be
1423/// considered to be from a system header.
1424SrcMgr::CharacteristicKind
1425SourceManager::getFileCharacteristic(SourceLocation Loc) const {
1426 assert(Loc.isValid() && "Can't get file characteristic of invalid loc!");
1427 std::pair<FileID, unsigned> LocInfo = getDecomposedExpansionLoc(Loc);
1428 const SLocEntry *SEntry = getSLocEntryForFile(FID: LocInfo.first);
1429 if (!SEntry)
1430 return C_User;
1431
1432 const SrcMgr::FileInfo &FI = SEntry->getFile();
1433
1434 // If there are no #line directives in this file, just return the whole-file
1435 // state.
1436 if (!FI.hasLineDirectives())
1437 return FI.getFileCharacteristic();
1438
1439 assert(LineTable && "Can't have linetable entries without a LineTable!");
1440 // See if there is a #line directive before the location.
1441 const LineEntry *Entry =
1442 LineTable->FindNearestLineEntry(FID: LocInfo.first, Offset: LocInfo.second);
1443
1444 // If this is before the first line marker, use the file characteristic.
1445 if (!Entry)
1446 return FI.getFileCharacteristic();
1447
1448 return Entry->FileKind;
1449}
1450
1451/// Return the filename or buffer identifier of the buffer the location is in.
1452/// Note that this name does not respect \#line directives. Use getPresumedLoc
1453/// for normal clients.
1454StringRef SourceManager::getBufferName(SourceLocation Loc,
1455 bool *Invalid) const {
1456 if (isInvalid(Loc, Invalid)) return "<invalid loc>";
1457
1458 auto B = getBufferOrNone(FID: getFileID(SpellingLoc: Loc));
1459 if (Invalid)
1460 *Invalid = !B;
1461 return B ? B->getBufferIdentifier() : "<invalid buffer>";
1462}
1463
1464/// getPresumedLoc - This method returns the "presumed" location of a
1465/// SourceLocation specifies. A "presumed location" can be modified by \#line
1466/// or GNU line marker directives. This provides a view on the data that a
1467/// user should see in diagnostics, for example.
1468///
1469/// Note that a presumed location is always given as the expansion point of an
1470/// expansion location, not at the spelling location.
1471PresumedLoc SourceManager::getPresumedLoc(SourceLocation Loc,
1472 bool UseLineDirectives) const {
1473 if (Loc.isInvalid()) return PresumedLoc();
1474
1475 // Presumed locations are always for expansion points.
1476 std::pair<FileID, unsigned> LocInfo = getDecomposedExpansionLoc(Loc);
1477
1478 bool Invalid = false;
1479 const SLocEntry &Entry = getSLocEntry(FID: LocInfo.first, Invalid: &Invalid);
1480 if (Invalid || !Entry.isFile())
1481 return PresumedLoc();
1482
1483 const SrcMgr::FileInfo &FI = Entry.getFile();
1484 const SrcMgr::ContentCache *C = &FI.getContentCache();
1485
1486 // To get the source name, first consult the FileEntry (if one exists)
1487 // before the MemBuffer as this will avoid unnecessarily paging in the
1488 // MemBuffer.
1489 FileID FID = LocInfo.first;
1490 StringRef Filename;
1491 if (C->OrigEntry)
1492 Filename = C->OrigEntry->getName();
1493 else if (auto Buffer = C->getBufferOrNone(Diag, FM&: getFileManager()))
1494 Filename = Buffer->getBufferIdentifier();
1495
1496 unsigned LineNo = getLineNumber(FID: LocInfo.first, FilePos: LocInfo.second, Invalid: &Invalid);
1497 if (Invalid)
1498 return PresumedLoc();
1499 unsigned ColNo = getColumnNumber(FID: LocInfo.first, FilePos: LocInfo.second, Invalid: &Invalid);
1500 if (Invalid)
1501 return PresumedLoc();
1502
1503 SourceLocation IncludeLoc = FI.getIncludeLoc();
1504
1505 // If we have #line directives in this file, update and overwrite the physical
1506 // location info if appropriate.
1507 if (UseLineDirectives && FI.hasLineDirectives()) {
1508 assert(LineTable && "Can't have linetable entries without a LineTable!");
1509 // See if there is a #line directive before this. If so, get it.
1510 if (const LineEntry *Entry =
1511 LineTable->FindNearestLineEntry(FID: LocInfo.first, Offset: LocInfo.second)) {
1512 // If the LineEntry indicates a filename, use it.
1513 if (Entry->FilenameID != -1) {
1514 Filename = LineTable->getFilename(ID: Entry->FilenameID);
1515 // The contents of files referenced by #line are not in the
1516 // SourceManager
1517 FID = FileID::get(V: 0);
1518 }
1519
1520 // Use the line number specified by the LineEntry. This line number may
1521 // be multiple lines down from the line entry. Add the difference in
1522 // physical line numbers from the query point and the line marker to the
1523 // total.
1524 unsigned MarkerLineNo = getLineNumber(FID: LocInfo.first, FilePos: Entry->FileOffset);
1525 LineNo = Entry->LineNo + (LineNo-MarkerLineNo-1);
1526
1527 // Note that column numbers are not molested by line markers.
1528
1529 // Handle virtual #include manipulation.
1530 if (Entry->IncludeOffset) {
1531 IncludeLoc = getLocForStartOfFile(FID: LocInfo.first);
1532 IncludeLoc = IncludeLoc.getLocWithOffset(Offset: Entry->IncludeOffset);
1533 }
1534 }
1535 }
1536
1537 return PresumedLoc(Filename.data(), FID, LineNo, ColNo, IncludeLoc);
1538}
1539
1540/// Returns whether the PresumedLoc for a given SourceLocation is
1541/// in the main file.
1542///
1543/// This computes the "presumed" location for a SourceLocation, then checks
1544/// whether it came from a file other than the main file. This is different
1545/// from isWrittenInMainFile() because it takes line marker directives into
1546/// account.
1547bool SourceManager::isInMainFile(SourceLocation Loc) const {
1548 if (Loc.isInvalid()) return false;
1549
1550 // Presumed locations are always for expansion points.
1551 std::pair<FileID, unsigned> LocInfo = getDecomposedExpansionLoc(Loc);
1552
1553 const SLocEntry *Entry = getSLocEntryForFile(FID: LocInfo.first);
1554 if (!Entry)
1555 return false;
1556
1557 const SrcMgr::FileInfo &FI = Entry->getFile();
1558
1559 // Check if there is a line directive for this location.
1560 if (FI.hasLineDirectives())
1561 if (const LineEntry *Entry =
1562 LineTable->FindNearestLineEntry(FID: LocInfo.first, Offset: LocInfo.second))
1563 if (Entry->IncludeOffset)
1564 return false;
1565
1566 return FI.getIncludeLoc().isInvalid();
1567}
1568
1569/// The size of the SLocEntry that \p FID represents.
1570unsigned SourceManager::getFileIDSize(FileID FID) const {
1571 bool Invalid = false;
1572 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, Invalid: &Invalid);
1573 if (Invalid)
1574 return 0;
1575
1576 int ID = FID.ID;
1577 SourceLocation::UIntTy NextOffset;
1578 if ((ID > 0 && unsigned(ID+1) == local_sloc_entry_size()))
1579 NextOffset = getNextLocalOffset();
1580 else if (ID+1 == -1)
1581 NextOffset = MaxLoadedOffset;
1582 else
1583 NextOffset = getSLocEntry(FID: FileID::get(V: ID+1)).getOffset();
1584
1585 return NextOffset - Entry.getOffset() - 1;
1586}
1587
1588//===----------------------------------------------------------------------===//
1589// Other miscellaneous methods.
1590//===----------------------------------------------------------------------===//
1591
1592/// Get the source location for the given file:line:col triplet.
1593///
1594/// If the source file is included multiple times, the source location will
1595/// be based upon an arbitrary inclusion.
1596SourceLocation SourceManager::translateFileLineCol(const FileEntry *SourceFile,
1597 unsigned Line,
1598 unsigned Col) const {
1599 assert(SourceFile && "Null source file!");
1600 assert(Line && Col && "Line and column should start from 1!");
1601
1602 FileID FirstFID = translateFile(SourceFile);
1603 return translateLineCol(FID: FirstFID, Line, Col);
1604}
1605
1606/// Get the FileID for the given file.
1607///
1608/// If the source file is included multiple times, the FileID will be the
1609/// first inclusion.
1610FileID SourceManager::translateFile(const FileEntry *SourceFile) const {
1611 assert(SourceFile && "Null source file!");
1612
1613 // First, check the main file ID, since it is common to look for a
1614 // location in the main file.
1615 if (MainFileID.isValid()) {
1616 bool Invalid = false;
1617 const SLocEntry &MainSLoc = getSLocEntry(FID: MainFileID, Invalid: &Invalid);
1618 if (Invalid)
1619 return FileID();
1620
1621 if (MainSLoc.isFile()) {
1622 if (MainSLoc.getFile().getContentCache().OrigEntry == SourceFile)
1623 return MainFileID;
1624 }
1625 }
1626
1627 // The location we're looking for isn't in the main file; look
1628 // through all of the local source locations.
1629 for (unsigned I = 0, N = local_sloc_entry_size(); I != N; ++I) {
1630 const SLocEntry &SLoc = getLocalSLocEntry(Index: I);
1631 if (SLoc.isFile() &&
1632 SLoc.getFile().getContentCache().OrigEntry == SourceFile)
1633 return FileID::get(V: I);
1634 }
1635
1636 // If that still didn't help, try the modules.
1637 for (unsigned I = 0, N = loaded_sloc_entry_size(); I != N; ++I) {
1638 const SLocEntry &SLoc = getLoadedSLocEntry(Index: I);
1639 if (SLoc.isFile() &&
1640 SLoc.getFile().getContentCache().OrigEntry == SourceFile)
1641 return FileID::get(V: -int(I) - 2);
1642 }
1643
1644 return FileID();
1645}
1646
1647/// Get the source location in \arg FID for the given line:col.
1648/// Returns null location if \arg FID is not a file SLocEntry.
1649SourceLocation SourceManager::translateLineCol(FileID FID,
1650 unsigned Line,
1651 unsigned Col) const {
1652 // Lines are used as a one-based index into a zero-based array. This assert
1653 // checks for possible buffer underruns.
1654 assert(Line && Col && "Line and column should start from 1!");
1655
1656 if (FID.isInvalid())
1657 return SourceLocation();
1658
1659 bool Invalid = false;
1660 const SLocEntry &Entry = getSLocEntry(FID, Invalid: &Invalid);
1661 if (Invalid)
1662 return SourceLocation();
1663
1664 if (!Entry.isFile())
1665 return SourceLocation();
1666
1667 SourceLocation FileLoc = SourceLocation::getFileLoc(ID: Entry.getOffset());
1668
1669 if (Line == 1 && Col == 1)
1670 return FileLoc;
1671
1672 const ContentCache *Content = &Entry.getFile().getContentCache();
1673
1674 // If this is the first use of line information for this buffer, compute the
1675 // SourceLineCache for it on demand.
1676 std::optional<llvm::MemoryBufferRef> Buffer =
1677 Content->getBufferOrNone(Diag, FM&: getFileManager());
1678 if (!Buffer)
1679 return SourceLocation();
1680 if (!Content->SourceLineCache)
1681 Content->SourceLineCache =
1682 LineOffsetMapping::get(Buffer: *Buffer, Alloc&: ContentCacheAlloc);
1683
1684 if (Line > Content->SourceLineCache.size()) {
1685 unsigned Size = Buffer->getBufferSize();
1686 if (Size > 0)
1687 --Size;
1688 return FileLoc.getLocWithOffset(Offset: Size);
1689 }
1690
1691 unsigned FilePos = Content->SourceLineCache[Line - 1];
1692 const char *Buf = Buffer->getBufferStart() + FilePos;
1693 unsigned BufLength = Buffer->getBufferSize() - FilePos;
1694 if (BufLength == 0)
1695 return FileLoc.getLocWithOffset(Offset: FilePos);
1696
1697 unsigned i = 0;
1698
1699 // Check that the given column is valid.
1700 while (i < BufLength-1 && i < Col-1 && Buf[i] != '\n' && Buf[i] != '\r')
1701 ++i;
1702 return FileLoc.getLocWithOffset(Offset: FilePos + i);
1703}
1704
1705/// Compute a map of macro argument chunks to their expanded source
1706/// location. Chunks that are not part of a macro argument will map to an
1707/// invalid source location. e.g. if a file contains one macro argument at
1708/// offset 100 with length 10, this is how the map will be formed:
1709/// 0 -> SourceLocation()
1710/// 100 -> Expanded macro arg location
1711/// 110 -> SourceLocation()
1712void SourceManager::computeMacroArgsCache(MacroArgsMap &MacroArgsCache,
1713 FileID FID) const {
1714 assert(FID.isValid());
1715
1716 // Initially no macro argument chunk is present.
1717 MacroArgsCache.insert(x: std::make_pair(x: 0, y: SourceLocation()));
1718
1719 int ID = FID.ID;
1720 while (true) {
1721 ++ID;
1722 // Stop if there are no more FileIDs to check.
1723 if (ID > 0) {
1724 if (unsigned(ID) >= local_sloc_entry_size())
1725 return;
1726 } else if (ID == -1) {
1727 return;
1728 }
1729
1730 bool Invalid = false;
1731 const SrcMgr::SLocEntry &Entry = getSLocEntryByID(ID, Invalid: &Invalid);
1732 if (Invalid)
1733 return;
1734 if (Entry.isFile()) {
1735 auto& File = Entry.getFile();
1736 if (File.getFileCharacteristic() == C_User_ModuleMap ||
1737 File.getFileCharacteristic() == C_System_ModuleMap)
1738 continue;
1739
1740 SourceLocation IncludeLoc = File.getIncludeLoc();
1741 bool IncludedInFID =
1742 (IncludeLoc.isValid() && isInFileID(Loc: IncludeLoc, FID)) ||
1743 // Predefined header doesn't have a valid include location in main
1744 // file, but any files created by it should still be skipped when
1745 // computing macro args expanded in the main file.
1746 (FID == MainFileID && Entry.getFile().getName() == "<built-in>");
1747 if (IncludedInFID) {
1748 // Skip the files/macros of the #include'd file, we only care about
1749 // macros that lexed macro arguments from our file.
1750 if (Entry.getFile().NumCreatedFIDs)
1751 ID += Entry.getFile().NumCreatedFIDs - 1 /*because of next ++ID*/;
1752 continue;
1753 }
1754 // If file was included but not from FID, there is no more files/macros
1755 // that may be "contained" in this file.
1756 if (IncludeLoc.isValid())
1757 return;
1758 continue;
1759 }
1760
1761 const ExpansionInfo &ExpInfo = Entry.getExpansion();
1762
1763 if (ExpInfo.getExpansionLocStart().isFileID()) {
1764 if (!isInFileID(Loc: ExpInfo.getExpansionLocStart(), FID))
1765 return; // No more files/macros that may be "contained" in this file.
1766 }
1767
1768 if (!ExpInfo.isMacroArgExpansion())
1769 continue;
1770
1771 associateFileChunkWithMacroArgExp(MacroArgsCache, FID,
1772 SpellLoc: ExpInfo.getSpellingLoc(),
1773 ExpansionLoc: SourceLocation::getMacroLoc(ID: Entry.getOffset()),
1774 ExpansionLength: getFileIDSize(FID: FileID::get(V: ID)));
1775 }
1776}
1777
1778void SourceManager::associateFileChunkWithMacroArgExp(
1779 MacroArgsMap &MacroArgsCache,
1780 FileID FID,
1781 SourceLocation SpellLoc,
1782 SourceLocation ExpansionLoc,
1783 unsigned ExpansionLength) const {
1784 if (!SpellLoc.isFileID()) {
1785 SourceLocation::UIntTy SpellBeginOffs = SpellLoc.getOffset();
1786 SourceLocation::UIntTy SpellEndOffs = SpellBeginOffs + ExpansionLength;
1787
1788 // The spelling range for this macro argument expansion can span multiple
1789 // consecutive FileID entries. Go through each entry contained in the
1790 // spelling range and if one is itself a macro argument expansion, recurse
1791 // and associate the file chunk that it represents.
1792
1793 FileID SpellFID; // Current FileID in the spelling range.
1794 unsigned SpellRelativeOffs;
1795 std::tie(args&: SpellFID, args&: SpellRelativeOffs) = getDecomposedLoc(Loc: SpellLoc);
1796 while (true) {
1797 const SLocEntry &Entry = getSLocEntry(FID: SpellFID);
1798 SourceLocation::UIntTy SpellFIDBeginOffs = Entry.getOffset();
1799 unsigned SpellFIDSize = getFileIDSize(FID: SpellFID);
1800 SourceLocation::UIntTy SpellFIDEndOffs = SpellFIDBeginOffs + SpellFIDSize;
1801 const ExpansionInfo &Info = Entry.getExpansion();
1802 if (Info.isMacroArgExpansion()) {
1803 unsigned CurrSpellLength;
1804 if (SpellFIDEndOffs < SpellEndOffs)
1805 CurrSpellLength = SpellFIDSize - SpellRelativeOffs;
1806 else
1807 CurrSpellLength = ExpansionLength;
1808 associateFileChunkWithMacroArgExp(MacroArgsCache, FID,
1809 SpellLoc: Info.getSpellingLoc().getLocWithOffset(Offset: SpellRelativeOffs),
1810 ExpansionLoc, ExpansionLength: CurrSpellLength);
1811 }
1812
1813 if (SpellFIDEndOffs >= SpellEndOffs)
1814 return; // we covered all FileID entries in the spelling range.
1815
1816 // Move to the next FileID entry in the spelling range.
1817 unsigned advance = SpellFIDSize - SpellRelativeOffs + 1;
1818 ExpansionLoc = ExpansionLoc.getLocWithOffset(Offset: advance);
1819 ExpansionLength -= advance;
1820 ++SpellFID.ID;
1821 SpellRelativeOffs = 0;
1822 }
1823 }
1824
1825 assert(SpellLoc.isFileID());
1826
1827 unsigned BeginOffs;
1828 if (!isInFileID(Loc: SpellLoc, FID, RelativeOffset: &BeginOffs))
1829 return;
1830
1831 unsigned EndOffs = BeginOffs + ExpansionLength;
1832
1833 // Add a new chunk for this macro argument. A previous macro argument chunk
1834 // may have been lexed again, so e.g. if the map is
1835 // 0 -> SourceLocation()
1836 // 100 -> Expanded loc #1
1837 // 110 -> SourceLocation()
1838 // and we found a new macro FileID that lexed from offset 105 with length 3,
1839 // the new map will be:
1840 // 0 -> SourceLocation()
1841 // 100 -> Expanded loc #1
1842 // 105 -> Expanded loc #2
1843 // 108 -> Expanded loc #1
1844 // 110 -> SourceLocation()
1845 //
1846 // Since re-lexed macro chunks will always be the same size or less of
1847 // previous chunks, we only need to find where the ending of the new macro
1848 // chunk is mapped to and update the map with new begin/end mappings.
1849
1850 MacroArgsMap::iterator I = MacroArgsCache.upper_bound(x: EndOffs);
1851 --I;
1852 SourceLocation EndOffsMappedLoc = I->second;
1853 MacroArgsCache[BeginOffs] = ExpansionLoc;
1854 MacroArgsCache[EndOffs] = EndOffsMappedLoc;
1855}
1856
1857void SourceManager::updateSlocUsageStats() const {
1858 SourceLocation::UIntTy UsedBytes =
1859 NextLocalOffset + (MaxLoadedOffset - CurrentLoadedOffset);
1860 MaxUsedSLocBytes.updateMax(V: UsedBytes);
1861}
1862
1863/// If \arg Loc points inside a function macro argument, the returned
1864/// location will be the macro location in which the argument was expanded.
1865/// If a macro argument is used multiple times, the expanded location will
1866/// be at the first expansion of the argument.
1867/// e.g.
1868/// MY_MACRO(foo);
1869/// ^
1870/// Passing a file location pointing at 'foo', will yield a macro location
1871/// where 'foo' was expanded into.
1872SourceLocation
1873SourceManager::getMacroArgExpandedLocation(SourceLocation Loc) const {
1874 if (Loc.isInvalid() || !Loc.isFileID())
1875 return Loc;
1876
1877 FileID FID;
1878 unsigned Offset;
1879 std::tie(args&: FID, args&: Offset) = getDecomposedLoc(Loc);
1880 if (FID.isInvalid())
1881 return Loc;
1882
1883 std::unique_ptr<MacroArgsMap> &MacroArgsCache = MacroArgsCacheMap[FID];
1884 if (!MacroArgsCache) {
1885 MacroArgsCache = std::make_unique<MacroArgsMap>();
1886 computeMacroArgsCache(MacroArgsCache&: *MacroArgsCache, FID);
1887 }
1888
1889 assert(!MacroArgsCache->empty());
1890 MacroArgsMap::iterator I = MacroArgsCache->upper_bound(x: Offset);
1891 // In case every element in MacroArgsCache is greater than Offset we can't
1892 // decrement the iterator.
1893 if (I == MacroArgsCache->begin())
1894 return Loc;
1895
1896 --I;
1897
1898 SourceLocation::UIntTy MacroArgBeginOffs = I->first;
1899 SourceLocation MacroArgExpandedLoc = I->second;
1900 if (MacroArgExpandedLoc.isValid())
1901 return MacroArgExpandedLoc.getLocWithOffset(Offset: Offset - MacroArgBeginOffs);
1902
1903 return Loc;
1904}
1905
1906std::pair<FileID, unsigned>
1907SourceManager::getDecomposedIncludedLoc(FileID FID) const {
1908 if (FID.isInvalid())
1909 return std::make_pair(x: FileID(), y: 0);
1910
1911 // Uses IncludedLocMap to retrieve/cache the decomposed loc.
1912
1913 using DecompTy = std::pair<FileID, unsigned>;
1914 auto InsertOp = IncludedLocMap.try_emplace(Key: FID);
1915 DecompTy &DecompLoc = InsertOp.first->second;
1916 if (!InsertOp.second)
1917 return DecompLoc; // already in map.
1918
1919 SourceLocation UpperLoc;
1920 bool Invalid = false;
1921 const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, Invalid: &Invalid);
1922 if (!Invalid) {
1923 if (Entry.isExpansion())
1924 UpperLoc = Entry.getExpansion().getExpansionLocStart();
1925 else
1926 UpperLoc = Entry.getFile().getIncludeLoc();
1927 }
1928
1929 if (UpperLoc.isValid())
1930 DecompLoc = getDecomposedLoc(Loc: UpperLoc);
1931
1932 return DecompLoc;
1933}
1934
1935FileID SourceManager::getUniqueLoadedASTFileID(SourceLocation Loc) const {
1936 assert(isLoadedSourceLocation(Loc) &&
1937 "Must be a source location in a loaded PCH/Module file");
1938
1939 auto [FID, Ignore] = getDecomposedLoc(Loc);
1940 // `LoadedSLocEntryAllocBegin` stores the sorted lowest FID of each loaded
1941 // allocation. Later allocations have lower FileIDs. The call below is to find
1942 // the lowest FID of a loaded allocation from any FID in the same allocation.
1943 // The lowest FID is used to identify a loaded allocation.
1944 const FileID *FirstFID =
1945 llvm::lower_bound(Range: LoadedSLocEntryAllocBegin, Value&: FID, C: std::greater<FileID>{});
1946
1947 assert(FirstFID &&
1948 "The failure to find the first FileID of a "
1949 "loaded AST from a loaded source location was unexpected.");
1950 return *FirstFID;
1951}
1952
1953bool SourceManager::isInTheSameTranslationUnitImpl(
1954 const std::pair<FileID, unsigned> &LOffs,
1955 const std::pair<FileID, unsigned> &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
1981MoveUpTranslationUnitIncludeHierarchy(std::pair<FileID, unsigned> &Loc,
1982 const SourceManager &SM) {
1983 std::pair<FileID, unsigned> UpperLoc = SM.getDecomposedIncludedLoc(FID: Loc.first);
1984 if (UpperLoc.first.isInvalid() ||
1985 !SM.isInTheSameTranslationUnitImpl(LOffs: UpperLoc, ROffs: Loc))
1986 return true; // We reached the top.
1987
1988 Loc = UpperLoc;
1989 return false;
1990}
1991
1992/// Return the cache entry for comparing the given file IDs
1993/// for isBeforeInTranslationUnit.
1994InBeforeInTUCacheEntry &SourceManager::getInBeforeInTUCache(FileID LFID,
1995 FileID RFID) const {
1996 // This is a magic number for limiting the cache size. It was experimentally
1997 // derived from a small Objective-C project (where the cache filled
1998 // out to ~250 items). We can make it larger if necessary.
1999 // FIXME: this is almost certainly full these days. Use an LRU cache?
2000 enum { MagicCacheSize = 300 };
2001 IsBeforeInTUCacheKey Key(LFID, RFID);
2002
2003 // If the cache size isn't too large, do a lookup and if necessary default
2004 // construct an entry. We can then return it to the caller for direct
2005 // use. When they update the value, the cache will get automatically
2006 // updated as well.
2007 if (IBTUCache.size() < MagicCacheSize)
2008 return IBTUCache.try_emplace(Key, Args&: LFID, Args&: RFID).first->second;
2009
2010 // Otherwise, do a lookup that will not construct a new value.
2011 InBeforeInTUCache::iterator I = IBTUCache.find(Val: Key);
2012 if (I != IBTUCache.end())
2013 return I->second;
2014
2015 // Fall back to the overflow value.
2016 IBTUCacheOverflow.setQueryFIDs(LHS: LFID, RHS: RFID);
2017 return IBTUCacheOverflow;
2018}
2019
2020/// Determines the order of 2 source locations in the translation unit.
2021///
2022/// \returns true if LHS source location comes before RHS, false otherwise.
2023bool SourceManager::isBeforeInTranslationUnit(SourceLocation LHS,
2024 SourceLocation RHS) const {
2025 assert(LHS.isValid() && RHS.isValid() && "Passed invalid source location!");
2026 if (LHS == RHS)
2027 return false;
2028
2029 std::pair<FileID, unsigned> LOffs = getDecomposedLoc(Loc: LHS);
2030 std::pair<FileID, unsigned> ROffs = getDecomposedLoc(Loc: RHS);
2031
2032 // getDecomposedLoc may have failed to return a valid FileID because, e.g. it
2033 // is a serialized one referring to a file that was removed after we loaded
2034 // the PCH.
2035 if (LOffs.first.isInvalid() || ROffs.first.isInvalid())
2036 return LOffs.first.isInvalid() && !ROffs.first.isInvalid();
2037
2038 std::pair<bool, bool> InSameTU = isInTheSameTranslationUnit(LOffs, ROffs);
2039 if (InSameTU.first)
2040 return InSameTU.second;
2041 // TODO: This should be unreachable, but some clients are calling this
2042 // function before making sure LHS and RHS are in the same TU.
2043 return LOffs.first < ROffs.first;
2044}
2045
2046std::pair<bool, bool> SourceManager::isInTheSameTranslationUnit(
2047 std::pair<FileID, unsigned> &LOffs,
2048 std::pair<FileID, unsigned> &ROffs) const {
2049 // If the source locations are not in the same TU, return early.
2050 if (!isInTheSameTranslationUnitImpl(LOffs, ROffs))
2051 return std::make_pair(x: false, y: false);
2052
2053 // If the source locations are in the same file, just compare offsets.
2054 if (LOffs.first == ROffs.first)
2055 return std::make_pair(x: true, y: LOffs.second < ROffs.second);
2056
2057 // If we are comparing a source location with multiple locations in the same
2058 // file, we get a big win by caching the result.
2059 InBeforeInTUCacheEntry &IsBeforeInTUCache =
2060 getInBeforeInTUCache(LFID: LOffs.first, RFID: ROffs.first);
2061
2062 // If we are comparing a source location with multiple locations in the same
2063 // file, we get a big win by caching the result.
2064 if (IsBeforeInTUCache.isCacheValid())
2065 return std::make_pair(
2066 x: true, y: IsBeforeInTUCache.getCachedResult(LOffset: LOffs.second, ROffset: ROffs.second));
2067
2068 // Okay, we missed in the cache, we'll compute the answer and populate it.
2069 // We need to find the common ancestor. The only way of doing this is to
2070 // build the complete include chain for one and then walking up the chain
2071 // of the other looking for a match.
2072
2073 // A location within a FileID on the path up from LOffs to the main file.
2074 struct Entry {
2075 std::pair<FileID, unsigned> DecomposedLoc; // FileID redundant, but clearer.
2076 FileID ChildFID; // Used for breaking ties. Invalid for the initial loc.
2077 };
2078 llvm::SmallDenseMap<FileID, Entry, 16> LChain;
2079
2080 FileID LChild;
2081 do {
2082 LChain.try_emplace(Key: LOffs.first, Args: Entry{.DecomposedLoc: LOffs, .ChildFID: LChild});
2083 // We catch the case where LOffs is in a file included by ROffs and
2084 // quit early. The other way round unfortunately remains suboptimal.
2085 if (LOffs.first == ROffs.first)
2086 break;
2087 LChild = LOffs.first;
2088 } while (!MoveUpTranslationUnitIncludeHierarchy(Loc&: LOffs, SM: *this));
2089
2090 FileID RChild;
2091 do {
2092 auto LIt = LChain.find(Val: ROffs.first);
2093 if (LIt != LChain.end()) {
2094 // Compare the locations within the common file and cache them.
2095 LOffs = LIt->second.DecomposedLoc;
2096 LChild = LIt->second.ChildFID;
2097 // The relative order of LChild and RChild is a tiebreaker when
2098 // - locs expand to the same location (occurs in macro arg expansion)
2099 // - one loc is a parent of the other (we consider the parent as "first")
2100 // For the parent entry to be first, its invalid child file ID must
2101 // compare smaller to the valid child file ID of the other entry.
2102 // However loaded FileIDs are <0, so we perform *unsigned* comparison!
2103 // This changes the relative order of local vs loaded FileIDs, but it
2104 // doesn't matter as these are never mixed in macro expansion.
2105 unsigned LChildID = LChild.ID;
2106 unsigned RChildID = RChild.ID;
2107 assert(((LOffs.second != ROffs.second) ||
2108 (LChildID == 0 || RChildID == 0) ||
2109 isInSameSLocAddrSpace(getComposedLoc(LChild, 0),
2110 getComposedLoc(RChild, 0), nullptr)) &&
2111 "Mixed local/loaded FileIDs with same include location?");
2112 IsBeforeInTUCache.setCommonLoc(commonFID: LOffs.first, lCommonOffset: LOffs.second, rCommonOffset: ROffs.second,
2113 LParentBeforeRParent: LChildID < RChildID);
2114 return std::make_pair(
2115 x: true, y: IsBeforeInTUCache.getCachedResult(LOffset: LOffs.second, ROffset: ROffs.second));
2116 }
2117 RChild = ROffs.first;
2118 } while (!MoveUpTranslationUnitIncludeHierarchy(Loc&: ROffs, SM: *this));
2119
2120 // If we found no match, the location is either in a built-ins buffer or
2121 // associated with global inline asm. PR5662 and PR22576 are examples.
2122
2123 StringRef LB = getBufferOrFake(FID: LOffs.first).getBufferIdentifier();
2124 StringRef RB = getBufferOrFake(FID: ROffs.first).getBufferIdentifier();
2125
2126 bool LIsBuiltins = LB == "<built-in>";
2127 bool RIsBuiltins = RB == "<built-in>";
2128 // Sort built-in before non-built-in.
2129 if (LIsBuiltins || RIsBuiltins) {
2130 if (LIsBuiltins != RIsBuiltins)
2131 return std::make_pair(x: true, y&: LIsBuiltins);
2132 // Both are in built-in buffers, but from different files. We just claim
2133 // that lower IDs come first.
2134 return std::make_pair(x: true, y: LOffs.first < ROffs.first);
2135 }
2136
2137 bool LIsAsm = LB == "<inline asm>";
2138 bool RIsAsm = RB == "<inline asm>";
2139 // Sort assembler after built-ins, but before the rest.
2140 if (LIsAsm || RIsAsm) {
2141 if (LIsAsm != RIsAsm)
2142 return std::make_pair(x: true, y&: RIsAsm);
2143 assert(LOffs.first == ROffs.first);
2144 return std::make_pair(x: true, y: false);
2145 }
2146
2147 bool LIsScratch = LB == "<scratch space>";
2148 bool RIsScratch = RB == "<scratch space>";
2149 // Sort scratch after inline asm, but before the rest.
2150 if (LIsScratch || RIsScratch) {
2151 if (LIsScratch != RIsScratch)
2152 return std::make_pair(x: true, y&: LIsScratch);
2153 return std::make_pair(x: true, y: LOffs.second < ROffs.second);
2154 }
2155
2156 llvm_unreachable("Unsortable locations found");
2157}
2158
2159void SourceManager::PrintStats() const {
2160 llvm::errs() << "\n*** Source Manager Stats:\n";
2161 llvm::errs() << FileInfos.size() << " files mapped, " << MemBufferInfos.size()
2162 << " mem buffers mapped.\n";
2163 llvm::errs() << LocalSLocEntryTable.size() << " local SLocEntries allocated ("
2164 << llvm::capacity_in_bytes(X: LocalSLocEntryTable)
2165 << " bytes of capacity), " << NextLocalOffset
2166 << "B of SLoc address space used.\n";
2167 llvm::errs() << LoadedSLocEntryTable.size()
2168 << " loaded SLocEntries allocated ("
2169 << llvm::capacity_in_bytes(x: LoadedSLocEntryTable)
2170 << " bytes of capacity), "
2171 << MaxLoadedOffset - CurrentLoadedOffset
2172 << "B of SLoc address space used.\n";
2173
2174 unsigned NumLineNumsComputed = 0;
2175 unsigned NumFileBytesMapped = 0;
2176 for (fileinfo_iterator I = fileinfo_begin(), E = fileinfo_end(); I != E; ++I){
2177 NumLineNumsComputed += bool(I->second->SourceLineCache);
2178 NumFileBytesMapped += I->second->getSizeBytesMapped();
2179 }
2180 unsigned NumMacroArgsComputed = MacroArgsCacheMap.size();
2181
2182 llvm::errs() << NumFileBytesMapped << " bytes of files mapped, "
2183 << NumLineNumsComputed << " files with line #'s computed, "
2184 << NumMacroArgsComputed << " files with macro args computed.\n";
2185 llvm::errs() << "FileID scans: " << NumLinearScans << " linear, "
2186 << NumBinaryProbes << " binary.\n";
2187}
2188
2189LLVM_DUMP_METHOD void SourceManager::dump() const {
2190 llvm::raw_ostream &out = llvm::errs();
2191
2192 auto DumpSLocEntry = [&](int ID, const SrcMgr::SLocEntry &Entry,
2193 std::optional<SourceLocation::UIntTy> NextStart) {
2194 out << "SLocEntry <FileID " << ID << "> " << (Entry.isFile() ? "file" : "expansion")
2195 << " <SourceLocation " << Entry.getOffset() << ":";
2196 if (NextStart)
2197 out << *NextStart << ">\n";
2198 else
2199 out << "???\?>\n";
2200 if (Entry.isFile()) {
2201 auto &FI = Entry.getFile();
2202 if (FI.NumCreatedFIDs)
2203 out << " covers <FileID " << ID << ":" << int(ID + FI.NumCreatedFIDs)
2204 << ">\n";
2205 if (FI.getIncludeLoc().isValid())
2206 out << " included from " << FI.getIncludeLoc().getOffset() << "\n";
2207 auto &CC = FI.getContentCache();
2208 out << " for " << (CC.OrigEntry ? CC.OrigEntry->getName() : "<none>")
2209 << "\n";
2210 if (CC.BufferOverridden)
2211 out << " contents overridden\n";
2212 if (CC.ContentsEntry != CC.OrigEntry) {
2213 out << " contents from "
2214 << (CC.ContentsEntry ? CC.ContentsEntry->getName() : "<none>")
2215 << "\n";
2216 }
2217 } else {
2218 auto &EI = Entry.getExpansion();
2219 out << " spelling from " << EI.getSpellingLoc().getOffset() << "\n";
2220 out << " macro " << (EI.isMacroArgExpansion() ? "arg" : "body")
2221 << " range <" << EI.getExpansionLocStart().getOffset() << ":"
2222 << EI.getExpansionLocEnd().getOffset() << ">\n";
2223 }
2224 };
2225
2226 // Dump local SLocEntries.
2227 for (unsigned ID = 0, NumIDs = LocalSLocEntryTable.size(); ID != NumIDs; ++ID) {
2228 DumpSLocEntry(ID, LocalSLocEntryTable[ID],
2229 ID == NumIDs - 1 ? NextLocalOffset
2230 : LocalSLocEntryTable[ID + 1].getOffset());
2231 }
2232 // Dump loaded SLocEntries.
2233 std::optional<SourceLocation::UIntTy> NextStart;
2234 for (unsigned Index = 0; Index != LoadedSLocEntryTable.size(); ++Index) {
2235 int ID = -(int)Index - 2;
2236 if (SLocEntryLoaded[Index]) {
2237 DumpSLocEntry(ID, LoadedSLocEntryTable[Index], NextStart);
2238 NextStart = LoadedSLocEntryTable[Index].getOffset();
2239 } else {
2240 NextStart = std::nullopt;
2241 }
2242 }
2243}
2244
2245void SourceManager::noteSLocAddressSpaceUsage(
2246 DiagnosticsEngine &Diag, std::optional<unsigned> MaxNotes) const {
2247 struct Info {
2248 // A location where this file was entered.
2249 SourceLocation Loc;
2250 // Number of times this FileEntry was entered.
2251 unsigned Inclusions = 0;
2252 // Size usage from the file itself.
2253 uint64_t DirectSize = 0;
2254 // Total size usage from the file and its macro expansions.
2255 uint64_t TotalSize = 0;
2256 };
2257 using UsageMap = llvm::MapVector<const FileEntry*, Info>;
2258
2259 UsageMap Usage;
2260 uint64_t CountedSize = 0;
2261
2262 auto AddUsageForFileID = [&](FileID ID) {
2263 // The +1 here is because getFileIDSize doesn't include the extra byte for
2264 // the one-past-the-end location.
2265 unsigned Size = getFileIDSize(FID: ID) + 1;
2266
2267 // Find the file that used this address space, either directly or by
2268 // macro expansion.
2269 SourceLocation FileStart = getFileLoc(Loc: getComposedLoc(FID: ID, Offset: 0));
2270 FileID FileLocID = getFileID(SpellingLoc: FileStart);
2271 const FileEntry *Entry = getFileEntryForID(FID: FileLocID);
2272
2273 Info &EntryInfo = Usage[Entry];
2274 if (EntryInfo.Loc.isInvalid())
2275 EntryInfo.Loc = FileStart;
2276 if (ID == FileLocID) {
2277 ++EntryInfo.Inclusions;
2278 EntryInfo.DirectSize += Size;
2279 }
2280 EntryInfo.TotalSize += Size;
2281 CountedSize += Size;
2282 };
2283
2284 // Loaded SLocEntries have indexes counting downwards from -2.
2285 for (size_t Index = 0; Index != LoadedSLocEntryTable.size(); ++Index) {
2286 AddUsageForFileID(FileID::get(V: -2 - Index));
2287 }
2288 // Local SLocEntries have indexes counting upwards from 0.
2289 for (size_t Index = 0; Index != LocalSLocEntryTable.size(); ++Index) {
2290 AddUsageForFileID(FileID::get(V: Index));
2291 }
2292
2293 // Sort the usage by size from largest to smallest. Break ties by raw source
2294 // location.
2295 auto SortedUsage = Usage.takeVector();
2296 auto Cmp = [](const UsageMap::value_type &A, const UsageMap::value_type &B) {
2297 return A.second.TotalSize > B.second.TotalSize ||
2298 (A.second.TotalSize == B.second.TotalSize &&
2299 A.second.Loc < B.second.Loc);
2300 };
2301 auto SortedEnd = SortedUsage.end();
2302 if (MaxNotes && SortedUsage.size() > *MaxNotes) {
2303 SortedEnd = SortedUsage.begin() + *MaxNotes;
2304 std::nth_element(first: SortedUsage.begin(), nth: SortedEnd, last: SortedUsage.end(), comp: Cmp);
2305 }
2306 std::sort(first: SortedUsage.begin(), last: SortedEnd, comp: Cmp);
2307
2308 // Produce note on sloc address space usage total.
2309 uint64_t LocalUsage = NextLocalOffset;
2310 uint64_t LoadedUsage = MaxLoadedOffset - CurrentLoadedOffset;
2311 int UsagePercent = static_cast<int>(100.0 * double(LocalUsage + LoadedUsage) /
2312 MaxLoadedOffset);
2313 Diag.Report(Loc: SourceLocation(), DiagID: diag::note_total_sloc_usage)
2314 << LocalUsage << LoadedUsage << (LocalUsage + LoadedUsage) << UsagePercent;
2315
2316 // Produce notes on sloc address space usage for each file with a high usage.
2317 uint64_t ReportedSize = 0;
2318 for (auto &[Entry, FileInfo] :
2319 llvm::make_range(x: SortedUsage.begin(), y: SortedEnd)) {
2320 Diag.Report(Loc: FileInfo.Loc, DiagID: diag::note_file_sloc_usage)
2321 << FileInfo.Inclusions << FileInfo.DirectSize
2322 << (FileInfo.TotalSize - FileInfo.DirectSize);
2323 ReportedSize += FileInfo.TotalSize;
2324 }
2325
2326 // Describe any remaining usage not reported in the per-file usage.
2327 if (ReportedSize != CountedSize) {
2328 Diag.Report(Loc: SourceLocation(), DiagID: diag::note_file_misc_sloc_usage)
2329 << (SortedUsage.end() - SortedEnd) << CountedSize - ReportedSize;
2330 }
2331}
2332
2333ExternalSLocEntrySource::~ExternalSLocEntrySource() = default;
2334
2335/// Return the amount of memory used by memory buffers, breaking down
2336/// by heap-backed versus mmap'ed memory.
2337SourceManager::MemoryBufferSizes SourceManager::getMemoryBufferSizes() const {
2338 size_t malloc_bytes = 0;
2339 size_t mmap_bytes = 0;
2340
2341 for (unsigned i = 0, e = MemBufferInfos.size(); i != e; ++i)
2342 if (size_t sized_mapped = MemBufferInfos[i]->getSizeBytesMapped())
2343 switch (MemBufferInfos[i]->getMemoryBufferKind()) {
2344 case llvm::MemoryBuffer::MemoryBuffer_MMap:
2345 mmap_bytes += sized_mapped;
2346 break;
2347 case llvm::MemoryBuffer::MemoryBuffer_Malloc:
2348 malloc_bytes += sized_mapped;
2349 break;
2350 }
2351
2352 return MemoryBufferSizes(malloc_bytes, mmap_bytes);
2353}
2354
2355size_t SourceManager::getDataStructureSizes() const {
2356 size_t size = llvm::capacity_in_bytes(x: MemBufferInfos) +
2357 llvm::capacity_in_bytes(X: LocalSLocEntryTable) +
2358 llvm::capacity_in_bytes(x: LoadedSLocEntryTable) +
2359 llvm::capacity_in_bytes(X: SLocEntryLoaded) +
2360 llvm::capacity_in_bytes(X: FileInfos);
2361
2362 if (OverriddenFilesInfo)
2363 size += llvm::capacity_in_bytes(X: OverriddenFilesInfo->OverriddenFiles);
2364
2365 return size;
2366}
2367
2368SourceManagerForFile::SourceManagerForFile(StringRef FileName,
2369 StringRef Content) {
2370 // This is referenced by `FileMgr` and will be released by `FileMgr` when it
2371 // is deleted.
2372 IntrusiveRefCntPtr<llvm::vfs::InMemoryFileSystem> InMemoryFileSystem(
2373 new llvm::vfs::InMemoryFileSystem);
2374 InMemoryFileSystem->addFile(
2375 Path: FileName, ModificationTime: 0,
2376 Buffer: llvm::MemoryBuffer::getMemBuffer(InputData: Content, BufferName: FileName,
2377 /*RequiresNullTerminator=*/false));
2378 // This is passed to `SM` as reference, so the pointer has to be referenced
2379 // in `Environment` so that `FileMgr` can out-live this function scope.
2380 FileMgr =
2381 std::make_unique<FileManager>(args: FileSystemOptions(), args&: InMemoryFileSystem);
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),
2386 args: new DiagnosticOptions);
2387 SourceMgr = std::make_unique<SourceManager>(args&: *Diagnostics, args&: *FileMgr);
2388 FileEntryRef FE = llvm::cantFail(ValOrErr: FileMgr->getFileRef(Filename: FileName));
2389 FileID ID =
2390 SourceMgr->createFileID(SourceFile: FE, IncludePos: SourceLocation(), FileCharacter: clang::SrcMgr::C_User);
2391 assert(ID.isValid());
2392 SourceMgr->setMainFileID(ID);
2393}
2394