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/StringRef.h" |
24 | #include "llvm/ADT/StringSwitch.h" |
25 | #include "llvm/Support/Allocator.h" |
26 | #include "llvm/Support/Capacity.h" |
27 | #include "llvm/Support/Compiler.h" |
28 | #include "llvm/Support/Endian.h" |
29 | #include "llvm/Support/ErrorHandling.h" |
30 | #include "llvm/Support/FileSystem.h" |
31 | #include "llvm/Support/MathExtras.h" |
32 | #include "llvm/Support/MemoryBuffer.h" |
33 | #include "llvm/Support/Path.h" |
34 | #include "llvm/Support/raw_ostream.h" |
35 | #include <algorithm> |
36 | #include <cassert> |
37 | #include <cstddef> |
38 | #include <cstdint> |
39 | #include <memory> |
40 | #include <optional> |
41 | #include <tuple> |
42 | #include <utility> |
43 | #include <vector> |
44 | |
45 | using namespace clang; |
46 | using namespace SrcMgr; |
47 | using llvm::MemoryBuffer; |
48 | |
49 | //===----------------------------------------------------------------------===// |
50 | // SourceManager Helper Classes |
51 | //===----------------------------------------------------------------------===// |
52 | |
53 | /// getSizeBytesMapped - Returns the number of bytes actually mapped for this |
54 | /// ContentCache. This can be 0 if the MemBuffer was not actually expanded. |
55 | unsigned ContentCache::getSizeBytesMapped() const { |
56 | return Buffer ? Buffer->getBufferSize() : 0; |
57 | } |
58 | |
59 | /// Returns the kind of memory used to back the memory buffer for |
60 | /// this content cache. This is used for performance analysis. |
61 | llvm::MemoryBuffer::BufferKind ContentCache::getMemoryBufferKind() const { |
62 | if (Buffer == nullptr) { |
63 | assert(0 && "Buffer should never be null" ); |
64 | return llvm::MemoryBuffer::MemoryBuffer_Malloc; |
65 | } |
66 | return Buffer->getBufferKind(); |
67 | } |
68 | |
69 | /// getSize - Returns the size of the content encapsulated by this ContentCache. |
70 | /// This can be the size of the source file or the size of an arbitrary |
71 | /// scratch buffer. If the ContentCache encapsulates a source file, that |
72 | /// file is not lazily brought in from disk to satisfy this query. |
73 | unsigned ContentCache::getSize() const { |
74 | return Buffer ? (unsigned)Buffer->getBufferSize() |
75 | : (unsigned)ContentsEntry->getSize(); |
76 | } |
77 | |
78 | const char *ContentCache::getInvalidBOM(StringRef BufStr) { |
79 | // If the buffer is valid, check to see if it has a UTF Byte Order Mark |
80 | // (BOM). We only support UTF-8 with and without a BOM right now. See |
81 | // http://en.wikipedia.org/wiki/Byte_order_mark for more information. |
82 | const char *InvalidBOM = |
83 | llvm::StringSwitch<const char *>(BufStr) |
84 | .StartsWith(S: llvm::StringLiteral::withInnerNUL(Str: "\x00\x00\xFE\xFF" ), |
85 | Value: "UTF-32 (BE)" ) |
86 | .StartsWith(S: llvm::StringLiteral::withInnerNUL(Str: "\xFF\xFE\x00\x00" ), |
87 | Value: "UTF-32 (LE)" ) |
88 | .StartsWith(S: "\xFE\xFF" , Value: "UTF-16 (BE)" ) |
89 | .StartsWith(S: "\xFF\xFE" , Value: "UTF-16 (LE)" ) |
90 | .StartsWith(S: "\x2B\x2F\x76" , Value: "UTF-7" ) |
91 | .StartsWith(S: "\xF7\x64\x4C" , Value: "UTF-1" ) |
92 | .StartsWith(S: "\xDD\x73\x66\x73" , Value: "UTF-EBCDIC" ) |
93 | .StartsWith(S: "\x0E\xFE\xFF" , Value: "SCSU" ) |
94 | .StartsWith(S: "\xFB\xEE\x28" , Value: "BOCU-1" ) |
95 | .StartsWith(S: "\x84\x31\x95\x33" , Value: "GB-18030" ) |
96 | .Default(Value: nullptr); |
97 | |
98 | return InvalidBOM; |
99 | } |
100 | |
101 | std::optional<llvm::MemoryBufferRef> |
102 | ContentCache::getBufferOrNone(DiagnosticsEngine &Diag, FileManager &FM, |
103 | SourceLocation Loc) const { |
104 | // Lazily create the Buffer for ContentCaches that wrap files. If we already |
105 | // computed it, just return what we have. |
106 | if (IsBufferInvalid) |
107 | return std::nullopt; |
108 | if (Buffer) |
109 | return Buffer->getMemBufferRef(); |
110 | if (!ContentsEntry) |
111 | return std::nullopt; |
112 | |
113 | // Start with the assumption that the buffer is invalid to simplify early |
114 | // return paths. |
115 | IsBufferInvalid = true; |
116 | |
117 | auto BufferOrError = FM.getBufferForFile(Entry: *ContentsEntry, isVolatile: IsFileVolatile); |
118 | |
119 | // If we were unable to open the file, then we are in an inconsistent |
120 | // situation where the content cache referenced a file which no longer |
121 | // exists. Most likely, we were using a stat cache with an invalid entry but |
122 | // the file could also have been removed during processing. Since we can't |
123 | // really deal with this situation, just create an empty buffer. |
124 | if (!BufferOrError) { |
125 | if (Diag.isDiagnosticInFlight()) |
126 | Diag.SetDelayedDiagnostic(diag::DiagID: err_cannot_open_file, |
127 | Arg1: ContentsEntry->getName(), |
128 | Arg2: BufferOrError.getError().message()); |
129 | else |
130 | Diag.Report(Loc, diag::err_cannot_open_file) |
131 | << ContentsEntry->getName() << BufferOrError.getError().message(); |
132 | |
133 | return std::nullopt; |
134 | } |
135 | |
136 | Buffer = std::move(*BufferOrError); |
137 | |
138 | // Check that the file's size fits in an 'unsigned' (with room for a |
139 | // past-the-end value). This is deeply regrettable, but various parts of |
140 | // Clang (including elsewhere in this file!) use 'unsigned' to represent file |
141 | // offsets, line numbers, string literal lengths, and so on, and fail |
142 | // miserably on large source files. |
143 | // |
144 | // Note: ContentsEntry could be a named pipe, in which case |
145 | // ContentsEntry::getSize() could have the wrong size. Use |
146 | // MemoryBuffer::getBufferSize() instead. |
147 | if (Buffer->getBufferSize() >= std::numeric_limits<unsigned>::max()) { |
148 | if (Diag.isDiagnosticInFlight()) |
149 | Diag.SetDelayedDiagnostic(diag::DiagID: err_file_too_large, |
150 | Arg1: ContentsEntry->getName()); |
151 | else |
152 | Diag.Report(Loc, diag::err_file_too_large) |
153 | << ContentsEntry->getName(); |
154 | |
155 | return std::nullopt; |
156 | } |
157 | |
158 | // Unless this is a named pipe (in which case we can handle a mismatch), |
159 | // check that the file's size is the same as in the file entry (which may |
160 | // have come from a stat cache). |
161 | if (!ContentsEntry->isNamedPipe() && |
162 | Buffer->getBufferSize() != (size_t)ContentsEntry->getSize()) { |
163 | if (Diag.isDiagnosticInFlight()) |
164 | Diag.SetDelayedDiagnostic(diag::DiagID: err_file_modified, |
165 | Arg1: ContentsEntry->getName()); |
166 | else |
167 | Diag.Report(Loc, diag::err_file_modified) |
168 | << ContentsEntry->getName(); |
169 | |
170 | return std::nullopt; |
171 | } |
172 | |
173 | // If the buffer is valid, check to see if it has a UTF Byte Order Mark |
174 | // (BOM). We only support UTF-8 with and without a BOM right now. See |
175 | // http://en.wikipedia.org/wiki/Byte_order_mark for more information. |
176 | StringRef BufStr = Buffer->getBuffer(); |
177 | const char *InvalidBOM = getInvalidBOM(BufStr); |
178 | |
179 | if (InvalidBOM) { |
180 | Diag.Report(Loc, diag::err_unsupported_bom) |
181 | << InvalidBOM << ContentsEntry->getName(); |
182 | return std::nullopt; |
183 | } |
184 | |
185 | // Buffer has been validated. |
186 | IsBufferInvalid = false; |
187 | return Buffer->getMemBufferRef(); |
188 | } |
189 | |
190 | unsigned LineTableInfo::getLineTableFilenameID(StringRef Name) { |
191 | auto IterBool = FilenameIDs.try_emplace(Key: Name, Args: FilenamesByID.size()); |
192 | if (IterBool.second) |
193 | FilenamesByID.push_back(x: &*IterBool.first); |
194 | return IterBool.first->second; |
195 | } |
196 | |
197 | /// Add a line note to the line table that indicates that there is a \#line or |
198 | /// GNU line marker at the specified FID/Offset location which changes the |
199 | /// presumed location to LineNo/FilenameID. If EntryExit is 0, then this doesn't |
200 | /// change the presumed \#include stack. If it is 1, this is a file entry, if |
201 | /// it is 2 then this is a file exit. FileKind specifies whether this is a |
202 | /// system header or extern C system header. |
203 | void LineTableInfo::AddLineNote(FileID FID, unsigned Offset, unsigned LineNo, |
204 | int FilenameID, unsigned EntryExit, |
205 | SrcMgr::CharacteristicKind FileKind) { |
206 | std::vector<LineEntry> &Entries = LineEntries[FID]; |
207 | |
208 | assert((Entries.empty() || Entries.back().FileOffset < Offset) && |
209 | "Adding line entries out of order!" ); |
210 | |
211 | unsigned IncludeOffset = 0; |
212 | if (EntryExit == 1) { |
213 | // Push #include |
214 | IncludeOffset = Offset-1; |
215 | } else { |
216 | const auto *PrevEntry = Entries.empty() ? nullptr : &Entries.back(); |
217 | if (EntryExit == 2) { |
218 | // Pop #include |
219 | assert(PrevEntry && PrevEntry->IncludeOffset && |
220 | "PPDirectives should have caught case when popping empty include " |
221 | "stack" ); |
222 | PrevEntry = FindNearestLineEntry(FID, Offset: PrevEntry->IncludeOffset); |
223 | } |
224 | if (PrevEntry) { |
225 | IncludeOffset = PrevEntry->IncludeOffset; |
226 | if (FilenameID == -1) { |
227 | // An unspecified FilenameID means use the previous (or containing) |
228 | // filename if available, or the main source file otherwise. |
229 | FilenameID = PrevEntry->FilenameID; |
230 | } |
231 | } |
232 | } |
233 | |
234 | Entries.push_back(x: LineEntry::get(Offs: Offset, Line: LineNo, Filename: FilenameID, FileKind, |
235 | IncludeOffset)); |
236 | } |
237 | |
238 | /// FindNearestLineEntry - Find the line entry nearest to FID that is before |
239 | /// it. If there is no line entry before Offset in FID, return null. |
240 | const LineEntry *LineTableInfo::FindNearestLineEntry(FileID FID, |
241 | unsigned Offset) { |
242 | const std::vector<LineEntry> &Entries = LineEntries[FID]; |
243 | assert(!Entries.empty() && "No #line entries for this FID after all!" ); |
244 | |
245 | // It is very common for the query to be after the last #line, check this |
246 | // first. |
247 | if (Entries.back().FileOffset <= Offset) |
248 | return &Entries.back(); |
249 | |
250 | // Do a binary search to find the maximal element that is still before Offset. |
251 | std::vector<LineEntry>::const_iterator I = llvm::upper_bound(Range: Entries, Value&: Offset); |
252 | if (I == Entries.begin()) |
253 | return nullptr; |
254 | return &*--I; |
255 | } |
256 | |
257 | /// Add a new line entry that has already been encoded into |
258 | /// the internal representation of the line table. |
259 | void LineTableInfo::AddEntry(FileID FID, |
260 | const std::vector<LineEntry> &Entries) { |
261 | LineEntries[FID] = Entries; |
262 | } |
263 | |
264 | /// getLineTableFilenameID - Return the uniqued ID for the specified filename. |
265 | unsigned SourceManager::getLineTableFilenameID(StringRef Name) { |
266 | return getLineTable().getLineTableFilenameID(Name); |
267 | } |
268 | |
269 | /// AddLineNote - Add a line note to the line table for the FileID and offset |
270 | /// specified by Loc. If FilenameID is -1, it is considered to be |
271 | /// unspecified. |
272 | void SourceManager::AddLineNote(SourceLocation Loc, unsigned LineNo, |
273 | int FilenameID, bool IsFileEntry, |
274 | bool IsFileExit, |
275 | SrcMgr::CharacteristicKind FileKind) { |
276 | std::pair<FileID, unsigned> LocInfo = getDecomposedExpansionLoc(Loc); |
277 | |
278 | bool Invalid = false; |
279 | const SLocEntry &Entry = getSLocEntry(FID: LocInfo.first, Invalid: &Invalid); |
280 | if (!Entry.isFile() || Invalid) |
281 | return; |
282 | |
283 | const SrcMgr::FileInfo &FileInfo = Entry.getFile(); |
284 | |
285 | // Remember that this file has #line directives now if it doesn't already. |
286 | const_cast<SrcMgr::FileInfo&>(FileInfo).setHasLineDirectives(); |
287 | |
288 | (void) getLineTable(); |
289 | |
290 | unsigned EntryExit = 0; |
291 | if (IsFileEntry) |
292 | EntryExit = 1; |
293 | else if (IsFileExit) |
294 | EntryExit = 2; |
295 | |
296 | LineTable->AddLineNote(FID: LocInfo.first, Offset: LocInfo.second, LineNo, FilenameID, |
297 | EntryExit, FileKind); |
298 | } |
299 | |
300 | LineTableInfo &SourceManager::getLineTable() { |
301 | if (!LineTable) |
302 | LineTable.reset(p: new LineTableInfo()); |
303 | return *LineTable; |
304 | } |
305 | |
306 | //===----------------------------------------------------------------------===// |
307 | // Private 'Create' methods. |
308 | //===----------------------------------------------------------------------===// |
309 | |
310 | SourceManager::SourceManager(DiagnosticsEngine &Diag, FileManager &FileMgr, |
311 | bool UserFilesAreVolatile) |
312 | : Diag(Diag), FileMgr(FileMgr), UserFilesAreVolatile(UserFilesAreVolatile) { |
313 | clearIDTables(); |
314 | Diag.setSourceManager(this); |
315 | } |
316 | |
317 | SourceManager::~SourceManager() { |
318 | // Delete FileEntry objects corresponding to content caches. Since the actual |
319 | // content cache objects are bump pointer allocated, we just have to run the |
320 | // dtors, but we call the deallocate method for completeness. |
321 | for (unsigned i = 0, e = MemBufferInfos.size(); i != e; ++i) { |
322 | if (MemBufferInfos[i]) { |
323 | MemBufferInfos[i]->~ContentCache(); |
324 | ContentCacheAlloc.Deallocate(Ptr: MemBufferInfos[i]); |
325 | } |
326 | } |
327 | for (auto I = FileInfos.begin(), E = FileInfos.end(); I != E; ++I) { |
328 | if (I->second) { |
329 | I->second->~ContentCache(); |
330 | ContentCacheAlloc.Deallocate(Ptr: I->second); |
331 | } |
332 | } |
333 | } |
334 | |
335 | void SourceManager::clearIDTables() { |
336 | MainFileID = FileID(); |
337 | LocalSLocEntryTable.clear(); |
338 | LoadedSLocEntryTable.clear(); |
339 | SLocEntryLoaded.clear(); |
340 | SLocEntryOffsetLoaded.clear(); |
341 | LastLineNoFileIDQuery = FileID(); |
342 | LastLineNoContentCache = nullptr; |
343 | LastFileIDLookup = FileID(); |
344 | |
345 | if (LineTable) |
346 | LineTable->clear(); |
347 | |
348 | // Use up FileID #0 as an invalid expansion. |
349 | NextLocalOffset = 0; |
350 | CurrentLoadedOffset = MaxLoadedOffset; |
351 | createExpansionLoc(SpellingLoc: SourceLocation(), ExpansionLocStart: SourceLocation(), ExpansionLocEnd: SourceLocation(), Length: 1); |
352 | } |
353 | |
354 | bool SourceManager::isMainFile(const FileEntry &SourceFile) { |
355 | assert(MainFileID.isValid() && "expected initialized SourceManager" ); |
356 | if (auto *FE = getFileEntryForID(FID: MainFileID)) |
357 | return FE->getUID() == SourceFile.getUID(); |
358 | return false; |
359 | } |
360 | |
361 | void SourceManager::initializeForReplay(const SourceManager &Old) { |
362 | assert(MainFileID.isInvalid() && "expected uninitialized SourceManager" ); |
363 | |
364 | auto CloneContentCache = [&](const ContentCache *Cache) -> ContentCache * { |
365 | auto *Clone = new (ContentCacheAlloc.Allocate<ContentCache>()) ContentCache; |
366 | Clone->OrigEntry = Cache->OrigEntry; |
367 | Clone->ContentsEntry = Cache->ContentsEntry; |
368 | Clone->BufferOverridden = Cache->BufferOverridden; |
369 | Clone->IsFileVolatile = Cache->IsFileVolatile; |
370 | Clone->IsTransient = Cache->IsTransient; |
371 | Clone->setUnownedBuffer(Cache->getBufferIfLoaded()); |
372 | return Clone; |
373 | }; |
374 | |
375 | // Ensure all SLocEntries are loaded from the external source. |
376 | for (unsigned I = 0, N = Old.LoadedSLocEntryTable.size(); I != N; ++I) |
377 | if (!Old.SLocEntryLoaded[I]) |
378 | Old.loadSLocEntry(Index: I, Invalid: nullptr); |
379 | |
380 | // Inherit any content cache data from the old source manager. |
381 | for (auto &FileInfo : Old.FileInfos) { |
382 | SrcMgr::ContentCache *&Slot = FileInfos[FileInfo.first]; |
383 | if (Slot) |
384 | continue; |
385 | Slot = CloneContentCache(FileInfo.second); |
386 | } |
387 | } |
388 | |
389 | ContentCache &SourceManager::getOrCreateContentCache(FileEntryRef FileEnt, |
390 | bool isSystemFile) { |
391 | // Do we already have information about this file? |
392 | ContentCache *&Entry = FileInfos[FileEnt]; |
393 | if (Entry) |
394 | return *Entry; |
395 | |
396 | // Nope, create a new Cache entry. |
397 | Entry = ContentCacheAlloc.Allocate<ContentCache>(); |
398 | |
399 | if (OverriddenFilesInfo) { |
400 | // If the file contents are overridden with contents from another file, |
401 | // pass that file to ContentCache. |
402 | auto overI = OverriddenFilesInfo->OverriddenFiles.find(Val: FileEnt); |
403 | if (overI == OverriddenFilesInfo->OverriddenFiles.end()) |
404 | new (Entry) ContentCache(FileEnt); |
405 | else |
406 | new (Entry) ContentCache(OverridenFilesKeepOriginalName ? FileEnt |
407 | : overI->second, |
408 | overI->second); |
409 | } else { |
410 | new (Entry) ContentCache(FileEnt); |
411 | } |
412 | |
413 | Entry->IsFileVolatile = UserFilesAreVolatile && !isSystemFile; |
414 | Entry->IsTransient = FilesAreTransient; |
415 | Entry->BufferOverridden |= FileEnt.isNamedPipe(); |
416 | |
417 | return *Entry; |
418 | } |
419 | |
420 | /// Create a new ContentCache for the specified memory buffer. |
421 | /// This does no caching. |
422 | ContentCache &SourceManager::createMemBufferContentCache( |
423 | std::unique_ptr<llvm::MemoryBuffer> Buffer) { |
424 | // Add a new ContentCache to the MemBufferInfos list and return it. |
425 | ContentCache *Entry = ContentCacheAlloc.Allocate<ContentCache>(); |
426 | new (Entry) ContentCache(); |
427 | MemBufferInfos.push_back(x: Entry); |
428 | Entry->setBuffer(std::move(Buffer)); |
429 | return *Entry; |
430 | } |
431 | |
432 | const SrcMgr::SLocEntry &SourceManager::loadSLocEntry(unsigned Index, |
433 | bool *Invalid) const { |
434 | assert(!SLocEntryLoaded[Index]); |
435 | if (ExternalSLocEntries->ReadSLocEntry(ID: -(static_cast<int>(Index) + 2))) { |
436 | if (Invalid) |
437 | *Invalid = true; |
438 | // If the file of the SLocEntry changed we could still have loaded it. |
439 | if (!SLocEntryLoaded[Index]) { |
440 | // Try to recover; create a SLocEntry so the rest of clang can handle it. |
441 | if (!FakeSLocEntryForRecovery) |
442 | FakeSLocEntryForRecovery = std::make_unique<SLocEntry>(args: SLocEntry::get( |
443 | Offset: 0, FI: FileInfo::get(IL: SourceLocation(), Con&: getFakeContentCacheForRecovery(), |
444 | FileCharacter: SrcMgr::C_User, Filename: "" ))); |
445 | return *FakeSLocEntryForRecovery; |
446 | } |
447 | } |
448 | |
449 | return LoadedSLocEntryTable[Index]; |
450 | } |
451 | |
452 | std::pair<int, SourceLocation::UIntTy> |
453 | SourceManager::AllocateLoadedSLocEntries(unsigned NumSLocEntries, |
454 | SourceLocation::UIntTy TotalSize) { |
455 | assert(ExternalSLocEntries && "Don't have an external sloc source" ); |
456 | // Make sure we're not about to run out of source locations. |
457 | if (CurrentLoadedOffset < TotalSize || |
458 | CurrentLoadedOffset - TotalSize < NextLocalOffset) { |
459 | return std::make_pair(x: 0, y: 0); |
460 | } |
461 | LoadedSLocEntryTable.resize(NewSize: LoadedSLocEntryTable.size() + NumSLocEntries); |
462 | SLocEntryLoaded.resize(N: LoadedSLocEntryTable.size()); |
463 | SLocEntryOffsetLoaded.resize(N: LoadedSLocEntryTable.size()); |
464 | CurrentLoadedOffset -= TotalSize; |
465 | int BaseID = -int(LoadedSLocEntryTable.size()) - 1; |
466 | LoadedSLocEntryAllocBegin.push_back(Elt: FileID::get(V: BaseID)); |
467 | return std::make_pair(x&: BaseID, y&: CurrentLoadedOffset); |
468 | } |
469 | |
470 | /// As part of recovering from missing or changed content, produce a |
471 | /// fake, non-empty buffer. |
472 | llvm::MemoryBufferRef SourceManager::getFakeBufferForRecovery() const { |
473 | if (!FakeBufferForRecovery) |
474 | FakeBufferForRecovery = |
475 | llvm::MemoryBuffer::getMemBuffer(InputData: "<<<INVALID BUFFER>>" ); |
476 | |
477 | return *FakeBufferForRecovery; |
478 | } |
479 | |
480 | /// As part of recovering from missing or changed content, produce a |
481 | /// fake content cache. |
482 | SrcMgr::ContentCache &SourceManager::getFakeContentCacheForRecovery() const { |
483 | if (!FakeContentCacheForRecovery) { |
484 | FakeContentCacheForRecovery = std::make_unique<SrcMgr::ContentCache>(); |
485 | FakeContentCacheForRecovery->setUnownedBuffer(getFakeBufferForRecovery()); |
486 | } |
487 | return *FakeContentCacheForRecovery; |
488 | } |
489 | |
490 | /// Returns the previous in-order FileID or an invalid FileID if there |
491 | /// is no previous one. |
492 | FileID SourceManager::getPreviousFileID(FileID FID) const { |
493 | if (FID.isInvalid()) |
494 | return FileID(); |
495 | |
496 | int ID = FID.ID; |
497 | if (ID == -1) |
498 | return FileID(); |
499 | |
500 | if (ID > 0) { |
501 | if (ID-1 == 0) |
502 | return FileID(); |
503 | } else if (unsigned(-(ID-1) - 2) >= LoadedSLocEntryTable.size()) { |
504 | return FileID(); |
505 | } |
506 | |
507 | return FileID::get(V: ID-1); |
508 | } |
509 | |
510 | /// Returns the next in-order FileID or an invalid FileID if there is |
511 | /// no next one. |
512 | FileID SourceManager::getNextFileID(FileID FID) const { |
513 | if (FID.isInvalid()) |
514 | return FileID(); |
515 | |
516 | int ID = FID.ID; |
517 | if (ID > 0) { |
518 | if (unsigned(ID+1) >= local_sloc_entry_size()) |
519 | return FileID(); |
520 | } else if (ID+1 >= -1) { |
521 | return FileID(); |
522 | } |
523 | |
524 | return FileID::get(V: ID+1); |
525 | } |
526 | |
527 | //===----------------------------------------------------------------------===// |
528 | // Methods to create new FileID's and macro expansions. |
529 | //===----------------------------------------------------------------------===// |
530 | |
531 | /// Create a new FileID that represents the specified file |
532 | /// being \#included from the specified IncludePosition. |
533 | FileID SourceManager::createFileID(FileEntryRef SourceFile, |
534 | SourceLocation IncludePos, |
535 | SrcMgr::CharacteristicKind FileCharacter, |
536 | int LoadedID, |
537 | SourceLocation::UIntTy LoadedOffset) { |
538 | SrcMgr::ContentCache &IR = getOrCreateContentCache(FileEnt: SourceFile, |
539 | isSystemFile: isSystem(CK: FileCharacter)); |
540 | |
541 | // If this is a named pipe, immediately load the buffer to ensure subsequent |
542 | // calls to ContentCache::getSize() are accurate. |
543 | if (IR.ContentsEntry->isNamedPipe()) |
544 | (void)IR.getBufferOrNone(Diag, FM&: getFileManager(), Loc: SourceLocation()); |
545 | |
546 | return createFileIDImpl(File&: IR, Filename: SourceFile.getName(), IncludePos, DirCharacter: FileCharacter, |
547 | LoadedID, LoadedOffset); |
548 | } |
549 | |
550 | /// Create a new FileID that represents the specified memory buffer. |
551 | /// |
552 | /// This does no caching of the buffer and takes ownership of the |
553 | /// MemoryBuffer, so only pass a MemoryBuffer to this once. |
554 | FileID SourceManager::createFileID(std::unique_ptr<llvm::MemoryBuffer> Buffer, |
555 | SrcMgr::CharacteristicKind FileCharacter, |
556 | int LoadedID, |
557 | SourceLocation::UIntTy LoadedOffset, |
558 | SourceLocation IncludeLoc) { |
559 | StringRef Name = Buffer->getBufferIdentifier(); |
560 | return createFileIDImpl(File&: createMemBufferContentCache(Buffer: std::move(Buffer)), Filename: Name, |
561 | IncludePos: IncludeLoc, DirCharacter: FileCharacter, LoadedID, LoadedOffset); |
562 | } |
563 | |
564 | /// Create a new FileID that represents the specified memory buffer. |
565 | /// |
566 | /// This does not take ownership of the MemoryBuffer. The memory buffer must |
567 | /// outlive the SourceManager. |
568 | FileID SourceManager::createFileID(const llvm::MemoryBufferRef &Buffer, |
569 | SrcMgr::CharacteristicKind FileCharacter, |
570 | int LoadedID, |
571 | SourceLocation::UIntTy LoadedOffset, |
572 | SourceLocation IncludeLoc) { |
573 | return createFileID(Buffer: llvm::MemoryBuffer::getMemBuffer(Ref: Buffer), FileCharacter, |
574 | LoadedID, LoadedOffset, IncludeLoc); |
575 | } |
576 | |
577 | /// Get the FileID for \p SourceFile if it exists. Otherwise, create a |
578 | /// new FileID for the \p SourceFile. |
579 | FileID |
580 | SourceManager::getOrCreateFileID(FileEntryRef SourceFile, |
581 | SrcMgr::CharacteristicKind FileCharacter) { |
582 | FileID ID = translateFile(SourceFile); |
583 | return ID.isValid() ? ID : createFileID(SourceFile, IncludePos: SourceLocation(), |
584 | FileCharacter); |
585 | } |
586 | |
587 | /// createFileID - Create a new FileID for the specified ContentCache and |
588 | /// include position. This works regardless of whether the ContentCache |
589 | /// corresponds to a file or some other input source. |
590 | FileID SourceManager::createFileIDImpl(ContentCache &File, StringRef Filename, |
591 | SourceLocation IncludePos, |
592 | SrcMgr::CharacteristicKind FileCharacter, |
593 | int LoadedID, |
594 | SourceLocation::UIntTy LoadedOffset) { |
595 | if (LoadedID < 0) { |
596 | assert(LoadedID != -1 && "Loading sentinel FileID" ); |
597 | unsigned Index = unsigned(-LoadedID) - 2; |
598 | assert(Index < LoadedSLocEntryTable.size() && "FileID out of range" ); |
599 | assert(!SLocEntryLoaded[Index] && "FileID already loaded" ); |
600 | LoadedSLocEntryTable[Index] = SLocEntry::get( |
601 | Offset: LoadedOffset, FI: FileInfo::get(IL: IncludePos, Con&: File, FileCharacter, Filename)); |
602 | SLocEntryLoaded[Index] = SLocEntryOffsetLoaded[Index] = true; |
603 | return FileID::get(V: LoadedID); |
604 | } |
605 | unsigned FileSize = File.getSize(); |
606 | if (!(NextLocalOffset + FileSize + 1 > NextLocalOffset && |
607 | NextLocalOffset + FileSize + 1 <= CurrentLoadedOffset)) { |
608 | Diag.Report(IncludePos, diag::err_sloc_space_too_large); |
609 | noteSLocAddressSpaceUsage(Diag); |
610 | return FileID(); |
611 | } |
612 | LocalSLocEntryTable.push_back( |
613 | Elt: SLocEntry::get(Offset: NextLocalOffset, |
614 | FI: FileInfo::get(IL: IncludePos, Con&: File, FileCharacter, Filename))); |
615 | // We do a +1 here because we want a SourceLocation that means "the end of the |
616 | // file", e.g. for the "no newline at the end of the file" diagnostic. |
617 | NextLocalOffset += FileSize + 1; |
618 | |
619 | // Set LastFileIDLookup to the newly created file. The next getFileID call is |
620 | // almost guaranteed to be from that file. |
621 | FileID FID = FileID::get(V: LocalSLocEntryTable.size()-1); |
622 | return LastFileIDLookup = FID; |
623 | } |
624 | |
625 | SourceLocation SourceManager::createMacroArgExpansionLoc( |
626 | SourceLocation SpellingLoc, SourceLocation ExpansionLoc, unsigned Length) { |
627 | ExpansionInfo Info = ExpansionInfo::createForMacroArg(SpellingLoc, |
628 | ExpansionLoc); |
629 | return createExpansionLocImpl(Expansion: Info, Length); |
630 | } |
631 | |
632 | SourceLocation SourceManager::createExpansionLoc( |
633 | SourceLocation SpellingLoc, SourceLocation ExpansionLocStart, |
634 | SourceLocation ExpansionLocEnd, unsigned Length, |
635 | bool ExpansionIsTokenRange, int LoadedID, |
636 | SourceLocation::UIntTy LoadedOffset) { |
637 | ExpansionInfo Info = ExpansionInfo::create( |
638 | SpellingLoc, Start: ExpansionLocStart, End: ExpansionLocEnd, ExpansionIsTokenRange); |
639 | return createExpansionLocImpl(Expansion: Info, Length, LoadedID, LoadedOffset); |
640 | } |
641 | |
642 | SourceLocation SourceManager::createTokenSplitLoc(SourceLocation Spelling, |
643 | SourceLocation TokenStart, |
644 | SourceLocation TokenEnd) { |
645 | assert(getFileID(TokenStart) == getFileID(TokenEnd) && |
646 | "token spans multiple files" ); |
647 | return createExpansionLocImpl( |
648 | Expansion: ExpansionInfo::createForTokenSplit(SpellingLoc: Spelling, Start: TokenStart, End: TokenEnd), |
649 | Length: TokenEnd.getOffset() - TokenStart.getOffset()); |
650 | } |
651 | |
652 | SourceLocation |
653 | SourceManager::createExpansionLocImpl(const ExpansionInfo &Info, |
654 | unsigned Length, int LoadedID, |
655 | SourceLocation::UIntTy LoadedOffset) { |
656 | if (LoadedID < 0) { |
657 | assert(LoadedID != -1 && "Loading sentinel FileID" ); |
658 | unsigned Index = unsigned(-LoadedID) - 2; |
659 | assert(Index < LoadedSLocEntryTable.size() && "FileID out of range" ); |
660 | assert(!SLocEntryLoaded[Index] && "FileID already loaded" ); |
661 | LoadedSLocEntryTable[Index] = SLocEntry::get(Offset: LoadedOffset, Expansion: Info); |
662 | SLocEntryLoaded[Index] = SLocEntryOffsetLoaded[Index] = true; |
663 | return SourceLocation::getMacroLoc(ID: LoadedOffset); |
664 | } |
665 | LocalSLocEntryTable.push_back(Elt: SLocEntry::get(Offset: NextLocalOffset, Expansion: Info)); |
666 | if (NextLocalOffset + Length + 1 <= NextLocalOffset || |
667 | NextLocalOffset + Length + 1 > CurrentLoadedOffset) { |
668 | Diag.Report(SourceLocation(), diag::err_sloc_space_too_large); |
669 | // FIXME: call `noteSLocAddressSpaceUsage` to report details to users and |
670 | // use a source location from `Info` to point at an error. |
671 | // Currently, both cause Clang to run indefinitely, this needs to be fixed. |
672 | // FIXME: return an error instead of crashing. Returning invalid source |
673 | // locations causes compiler to run indefinitely. |
674 | llvm::report_fatal_error(reason: "ran out of source locations" ); |
675 | } |
676 | // See createFileID for that +1. |
677 | NextLocalOffset += Length + 1; |
678 | return SourceLocation::getMacroLoc(ID: NextLocalOffset - (Length + 1)); |
679 | } |
680 | |
681 | std::optional<llvm::MemoryBufferRef> |
682 | SourceManager::getMemoryBufferForFileOrNone(FileEntryRef File) { |
683 | SrcMgr::ContentCache &IR = getOrCreateContentCache(FileEnt: File); |
684 | return IR.getBufferOrNone(Diag, FM&: getFileManager(), Loc: SourceLocation()); |
685 | } |
686 | |
687 | void SourceManager::overrideFileContents( |
688 | FileEntryRef SourceFile, std::unique_ptr<llvm::MemoryBuffer> Buffer) { |
689 | SrcMgr::ContentCache &IR = getOrCreateContentCache(FileEnt: SourceFile); |
690 | |
691 | IR.setBuffer(std::move(Buffer)); |
692 | IR.BufferOverridden = true; |
693 | |
694 | getOverriddenFilesInfo().OverriddenFilesWithBuffer.insert(V: SourceFile); |
695 | } |
696 | |
697 | void SourceManager::overrideFileContents(const FileEntry *SourceFile, |
698 | FileEntryRef NewFile) { |
699 | assert(SourceFile->getSize() == NewFile.getSize() && |
700 | "Different sizes, use the FileManager to create a virtual file with " |
701 | "the correct size" ); |
702 | assert(FileInfos.find_as(SourceFile) == FileInfos.end() && |
703 | "This function should be called at the initialization stage, before " |
704 | "any parsing occurs." ); |
705 | // FileEntryRef is not default-constructible. |
706 | auto Pair = getOverriddenFilesInfo().OverriddenFiles.insert( |
707 | KV: std::make_pair(x&: SourceFile, y&: NewFile)); |
708 | if (!Pair.second) |
709 | Pair.first->second = NewFile; |
710 | } |
711 | |
712 | OptionalFileEntryRef |
713 | SourceManager::bypassFileContentsOverride(FileEntryRef File) { |
714 | assert(isFileOverridden(&File.getFileEntry())); |
715 | OptionalFileEntryRef BypassFile = FileMgr.getBypassFile(VFE: File); |
716 | |
717 | // If the file can't be found in the FS, give up. |
718 | if (!BypassFile) |
719 | return std::nullopt; |
720 | |
721 | (void)getOrCreateContentCache(FileEnt: *BypassFile); |
722 | return BypassFile; |
723 | } |
724 | |
725 | void SourceManager::setFileIsTransient(FileEntryRef File) { |
726 | getOrCreateContentCache(FileEnt: File).IsTransient = true; |
727 | } |
728 | |
729 | std::optional<StringRef> |
730 | SourceManager::getNonBuiltinFilenameForID(FileID FID) const { |
731 | if (const SrcMgr::SLocEntry *Entry = getSLocEntryForFile(FID)) |
732 | if (Entry->getFile().getContentCache().OrigEntry) |
733 | return Entry->getFile().getName(); |
734 | return std::nullopt; |
735 | } |
736 | |
737 | StringRef SourceManager::getBufferData(FileID FID, bool *Invalid) const { |
738 | auto B = getBufferDataOrNone(FID); |
739 | if (Invalid) |
740 | *Invalid = !B; |
741 | return B ? *B : "<<<<<INVALID SOURCE LOCATION>>>>>" ; |
742 | } |
743 | |
744 | std::optional<StringRef> |
745 | SourceManager::getBufferDataIfLoaded(FileID FID) const { |
746 | if (const SrcMgr::SLocEntry *Entry = getSLocEntryForFile(FID)) |
747 | return Entry->getFile().getContentCache().getBufferDataIfLoaded(); |
748 | return std::nullopt; |
749 | } |
750 | |
751 | std::optional<StringRef> SourceManager::getBufferDataOrNone(FileID FID) const { |
752 | if (const SrcMgr::SLocEntry *Entry = getSLocEntryForFile(FID)) |
753 | if (auto B = Entry->getFile().getContentCache().getBufferOrNone( |
754 | Diag, FM&: getFileManager(), Loc: SourceLocation())) |
755 | return B->getBuffer(); |
756 | return std::nullopt; |
757 | } |
758 | |
759 | //===----------------------------------------------------------------------===// |
760 | // SourceLocation manipulation methods. |
761 | //===----------------------------------------------------------------------===// |
762 | |
763 | /// Return the FileID for a SourceLocation. |
764 | /// |
765 | /// This is the cache-miss path of getFileID. Not as hot as that function, but |
766 | /// still very important. It is responsible for finding the entry in the |
767 | /// SLocEntry tables that contains the specified location. |
768 | FileID SourceManager::getFileIDSlow(SourceLocation::UIntTy SLocOffset) const { |
769 | if (!SLocOffset) |
770 | return FileID::get(V: 0); |
771 | |
772 | // Now it is time to search for the correct file. See where the SLocOffset |
773 | // sits in the global view and consult local or loaded buffers for it. |
774 | if (SLocOffset < NextLocalOffset) |
775 | return getFileIDLocal(SLocOffset); |
776 | return getFileIDLoaded(SLocOffset); |
777 | } |
778 | |
779 | /// Return the FileID for a SourceLocation with a low offset. |
780 | /// |
781 | /// This function knows that the SourceLocation is in a local buffer, not a |
782 | /// loaded one. |
783 | FileID SourceManager::getFileIDLocal(SourceLocation::UIntTy SLocOffset) const { |
784 | assert(SLocOffset < NextLocalOffset && "Bad function choice" ); |
785 | |
786 | // After the first and second level caches, I see two common sorts of |
787 | // behavior: 1) a lot of searched FileID's are "near" the cached file |
788 | // location or are "near" the cached expansion location. 2) others are just |
789 | // completely random and may be a very long way away. |
790 | // |
791 | // To handle this, we do a linear search for up to 8 steps to catch #1 quickly |
792 | // then we fall back to a less cache efficient, but more scalable, binary |
793 | // search to find the location. |
794 | |
795 | // See if this is near the file point - worst case we start scanning from the |
796 | // most newly created FileID. |
797 | |
798 | // LessIndex - This is the lower bound of the range that we're searching. |
799 | // We know that the offset corresponding to the FileID is less than |
800 | // SLocOffset. |
801 | unsigned LessIndex = 0; |
802 | // upper bound of the search range. |
803 | unsigned GreaterIndex = LocalSLocEntryTable.size(); |
804 | if (LastFileIDLookup.ID >= 0) { |
805 | // Use the LastFileIDLookup to prune the search space. |
806 | if (LocalSLocEntryTable[LastFileIDLookup.ID].getOffset() < SLocOffset) |
807 | LessIndex = LastFileIDLookup.ID; |
808 | else |
809 | GreaterIndex = LastFileIDLookup.ID; |
810 | } |
811 | |
812 | // Find the FileID that contains this. |
813 | unsigned NumProbes = 0; |
814 | while (true) { |
815 | --GreaterIndex; |
816 | assert(GreaterIndex < LocalSLocEntryTable.size()); |
817 | if (LocalSLocEntryTable[GreaterIndex].getOffset() <= SLocOffset) { |
818 | FileID Res = FileID::get(V: int(GreaterIndex)); |
819 | // Remember it. We have good locality across FileID lookups. |
820 | LastFileIDLookup = Res; |
821 | NumLinearScans += NumProbes+1; |
822 | return Res; |
823 | } |
824 | if (++NumProbes == 8) |
825 | break; |
826 | } |
827 | |
828 | NumProbes = 0; |
829 | while (true) { |
830 | unsigned MiddleIndex = (GreaterIndex-LessIndex)/2+LessIndex; |
831 | SourceLocation::UIntTy MidOffset = |
832 | getLocalSLocEntry(Index: MiddleIndex).getOffset(); |
833 | |
834 | ++NumProbes; |
835 | |
836 | // If the offset of the midpoint is too large, chop the high side of the |
837 | // range to the midpoint. |
838 | if (MidOffset > SLocOffset) { |
839 | GreaterIndex = MiddleIndex; |
840 | continue; |
841 | } |
842 | |
843 | // If the middle index contains the value, succeed and return. |
844 | if (MiddleIndex + 1 == LocalSLocEntryTable.size() || |
845 | SLocOffset < getLocalSLocEntry(Index: MiddleIndex + 1).getOffset()) { |
846 | FileID Res = FileID::get(V: MiddleIndex); |
847 | |
848 | // Remember it. We have good locality across FileID lookups. |
849 | LastFileIDLookup = Res; |
850 | NumBinaryProbes += NumProbes; |
851 | return Res; |
852 | } |
853 | |
854 | // Otherwise, move the low-side up to the middle index. |
855 | LessIndex = MiddleIndex; |
856 | } |
857 | } |
858 | |
859 | /// Return the FileID for a SourceLocation with a high offset. |
860 | /// |
861 | /// This function knows that the SourceLocation is in a loaded buffer, not a |
862 | /// local one. |
863 | FileID SourceManager::getFileIDLoaded(SourceLocation::UIntTy SLocOffset) const { |
864 | if (SLocOffset < CurrentLoadedOffset) { |
865 | assert(0 && "Invalid SLocOffset or bad function choice" ); |
866 | return FileID(); |
867 | } |
868 | |
869 | return FileID::get(V: ExternalSLocEntries->getSLocEntryID(SLocOffset)); |
870 | } |
871 | |
872 | SourceLocation SourceManager:: |
873 | getExpansionLocSlowCase(SourceLocation Loc) const { |
874 | do { |
875 | // Note: If Loc indicates an offset into a token that came from a macro |
876 | // expansion (e.g. the 5th character of the token) we do not want to add |
877 | // this offset when going to the expansion location. The expansion |
878 | // location is the macro invocation, which the offset has nothing to do |
879 | // with. This is unlike when we get the spelling loc, because the offset |
880 | // directly correspond to the token whose spelling we're inspecting. |
881 | Loc = getSLocEntry(FID: getFileID(SpellingLoc: Loc)).getExpansion().getExpansionLocStart(); |
882 | } while (!Loc.isFileID()); |
883 | |
884 | return Loc; |
885 | } |
886 | |
887 | SourceLocation SourceManager::getSpellingLocSlowCase(SourceLocation Loc) const { |
888 | do { |
889 | std::pair<FileID, unsigned> LocInfo = getDecomposedLoc(Loc); |
890 | Loc = getSLocEntry(FID: LocInfo.first).getExpansion().getSpellingLoc(); |
891 | Loc = Loc.getLocWithOffset(Offset: LocInfo.second); |
892 | } while (!Loc.isFileID()); |
893 | return Loc; |
894 | } |
895 | |
896 | SourceLocation SourceManager::getFileLocSlowCase(SourceLocation Loc) const { |
897 | do { |
898 | if (isMacroArgExpansion(Loc)) |
899 | Loc = getImmediateSpellingLoc(Loc); |
900 | else |
901 | Loc = getImmediateExpansionRange(Loc).getBegin(); |
902 | } while (!Loc.isFileID()); |
903 | return Loc; |
904 | } |
905 | |
906 | |
907 | std::pair<FileID, unsigned> |
908 | SourceManager::getDecomposedExpansionLocSlowCase( |
909 | const SrcMgr::SLocEntry *E) const { |
910 | // If this is an expansion record, walk through all the expansion points. |
911 | FileID FID; |
912 | SourceLocation Loc; |
913 | unsigned Offset; |
914 | do { |
915 | Loc = E->getExpansion().getExpansionLocStart(); |
916 | |
917 | FID = getFileID(SpellingLoc: Loc); |
918 | E = &getSLocEntry(FID); |
919 | Offset = Loc.getOffset()-E->getOffset(); |
920 | } while (!Loc.isFileID()); |
921 | |
922 | return std::make_pair(x&: FID, y&: Offset); |
923 | } |
924 | |
925 | std::pair<FileID, unsigned> |
926 | SourceManager::getDecomposedSpellingLocSlowCase(const SrcMgr::SLocEntry *E, |
927 | unsigned Offset) const { |
928 | // If this is an expansion record, walk through all the expansion points. |
929 | FileID FID; |
930 | SourceLocation Loc; |
931 | do { |
932 | Loc = E->getExpansion().getSpellingLoc(); |
933 | Loc = Loc.getLocWithOffset(Offset); |
934 | |
935 | FID = getFileID(SpellingLoc: Loc); |
936 | E = &getSLocEntry(FID); |
937 | Offset = Loc.getOffset()-E->getOffset(); |
938 | } while (!Loc.isFileID()); |
939 | |
940 | return std::make_pair(x&: FID, y&: Offset); |
941 | } |
942 | |
943 | /// getImmediateSpellingLoc - Given a SourceLocation object, return the |
944 | /// spelling location referenced by the ID. This is the first level down |
945 | /// towards the place where the characters that make up the lexed token can be |
946 | /// found. This should not generally be used by clients. |
947 | SourceLocation SourceManager::getImmediateSpellingLoc(SourceLocation Loc) const{ |
948 | if (Loc.isFileID()) return Loc; |
949 | std::pair<FileID, unsigned> LocInfo = getDecomposedLoc(Loc); |
950 | Loc = getSLocEntry(FID: LocInfo.first).getExpansion().getSpellingLoc(); |
951 | return Loc.getLocWithOffset(Offset: LocInfo.second); |
952 | } |
953 | |
954 | /// Return the filename of the file containing a SourceLocation. |
955 | StringRef SourceManager::getFilename(SourceLocation SpellingLoc) const { |
956 | if (OptionalFileEntryRef F = getFileEntryRefForID(FID: getFileID(SpellingLoc))) |
957 | return F->getName(); |
958 | return StringRef(); |
959 | } |
960 | |
961 | /// getImmediateExpansionRange - Loc is required to be an expansion location. |
962 | /// Return the start/end of the expansion information. |
963 | CharSourceRange |
964 | SourceManager::getImmediateExpansionRange(SourceLocation Loc) const { |
965 | assert(Loc.isMacroID() && "Not a macro expansion loc!" ); |
966 | const ExpansionInfo &Expansion = getSLocEntry(FID: getFileID(SpellingLoc: Loc)).getExpansion(); |
967 | return Expansion.getExpansionLocRange(); |
968 | } |
969 | |
970 | SourceLocation SourceManager::getTopMacroCallerLoc(SourceLocation Loc) const { |
971 | while (isMacroArgExpansion(Loc)) |
972 | Loc = getImmediateSpellingLoc(Loc); |
973 | return Loc; |
974 | } |
975 | |
976 | /// getExpansionRange - Given a SourceLocation object, return the range of |
977 | /// tokens covered by the expansion in the ultimate file. |
978 | CharSourceRange SourceManager::getExpansionRange(SourceLocation Loc) const { |
979 | if (Loc.isFileID()) |
980 | return CharSourceRange(SourceRange(Loc, Loc), true); |
981 | |
982 | CharSourceRange Res = getImmediateExpansionRange(Loc); |
983 | |
984 | // Fully resolve the start and end locations to their ultimate expansion |
985 | // points. |
986 | while (!Res.getBegin().isFileID()) |
987 | Res.setBegin(getImmediateExpansionRange(Loc: Res.getBegin()).getBegin()); |
988 | while (!Res.getEnd().isFileID()) { |
989 | CharSourceRange EndRange = getImmediateExpansionRange(Loc: Res.getEnd()); |
990 | Res.setEnd(EndRange.getEnd()); |
991 | Res.setTokenRange(EndRange.isTokenRange()); |
992 | } |
993 | return Res; |
994 | } |
995 | |
996 | bool SourceManager::isMacroArgExpansion(SourceLocation Loc, |
997 | SourceLocation *StartLoc) const { |
998 | if (!Loc.isMacroID()) return false; |
999 | |
1000 | FileID FID = getFileID(SpellingLoc: Loc); |
1001 | const SrcMgr::ExpansionInfo &Expansion = getSLocEntry(FID).getExpansion(); |
1002 | if (!Expansion.isMacroArgExpansion()) return false; |
1003 | |
1004 | if (StartLoc) |
1005 | *StartLoc = Expansion.getExpansionLocStart(); |
1006 | return true; |
1007 | } |
1008 | |
1009 | bool SourceManager::isMacroBodyExpansion(SourceLocation Loc) const { |
1010 | if (!Loc.isMacroID()) return false; |
1011 | |
1012 | FileID FID = getFileID(SpellingLoc: Loc); |
1013 | const SrcMgr::ExpansionInfo &Expansion = getSLocEntry(FID).getExpansion(); |
1014 | return Expansion.isMacroBodyExpansion(); |
1015 | } |
1016 | |
1017 | bool SourceManager::isAtStartOfImmediateMacroExpansion(SourceLocation Loc, |
1018 | SourceLocation *MacroBegin) const { |
1019 | assert(Loc.isValid() && Loc.isMacroID() && "Expected a valid macro loc" ); |
1020 | |
1021 | std::pair<FileID, unsigned> DecompLoc = getDecomposedLoc(Loc); |
1022 | if (DecompLoc.second > 0) |
1023 | return false; // Does not point at the start of expansion range. |
1024 | |
1025 | bool Invalid = false; |
1026 | const SrcMgr::ExpansionInfo &ExpInfo = |
1027 | getSLocEntry(FID: DecompLoc.first, Invalid: &Invalid).getExpansion(); |
1028 | if (Invalid) |
1029 | return false; |
1030 | SourceLocation ExpLoc = ExpInfo.getExpansionLocStart(); |
1031 | |
1032 | if (ExpInfo.isMacroArgExpansion()) { |
1033 | // For macro argument expansions, check if the previous FileID is part of |
1034 | // the same argument expansion, in which case this Loc is not at the |
1035 | // beginning of the expansion. |
1036 | FileID PrevFID = getPreviousFileID(FID: DecompLoc.first); |
1037 | if (!PrevFID.isInvalid()) { |
1038 | const SrcMgr::SLocEntry &PrevEntry = getSLocEntry(FID: PrevFID, Invalid: &Invalid); |
1039 | if (Invalid) |
1040 | return false; |
1041 | if (PrevEntry.isExpansion() && |
1042 | PrevEntry.getExpansion().getExpansionLocStart() == ExpLoc) |
1043 | return false; |
1044 | } |
1045 | } |
1046 | |
1047 | if (MacroBegin) |
1048 | *MacroBegin = ExpLoc; |
1049 | return true; |
1050 | } |
1051 | |
1052 | bool SourceManager::isAtEndOfImmediateMacroExpansion(SourceLocation Loc, |
1053 | SourceLocation *MacroEnd) const { |
1054 | assert(Loc.isValid() && Loc.isMacroID() && "Expected a valid macro loc" ); |
1055 | |
1056 | FileID FID = getFileID(SpellingLoc: Loc); |
1057 | SourceLocation NextLoc = Loc.getLocWithOffset(Offset: 1); |
1058 | if (isInFileID(Loc: NextLoc, FID)) |
1059 | return false; // Does not point at the end of expansion range. |
1060 | |
1061 | bool Invalid = false; |
1062 | const SrcMgr::ExpansionInfo &ExpInfo = |
1063 | getSLocEntry(FID, Invalid: &Invalid).getExpansion(); |
1064 | if (Invalid) |
1065 | return false; |
1066 | |
1067 | if (ExpInfo.isMacroArgExpansion()) { |
1068 | // For macro argument expansions, check if the next FileID is part of the |
1069 | // same argument expansion, in which case this Loc is not at the end of the |
1070 | // expansion. |
1071 | FileID NextFID = getNextFileID(FID); |
1072 | if (!NextFID.isInvalid()) { |
1073 | const SrcMgr::SLocEntry &NextEntry = getSLocEntry(FID: NextFID, Invalid: &Invalid); |
1074 | if (Invalid) |
1075 | return false; |
1076 | if (NextEntry.isExpansion() && |
1077 | NextEntry.getExpansion().getExpansionLocStart() == |
1078 | ExpInfo.getExpansionLocStart()) |
1079 | return false; |
1080 | } |
1081 | } |
1082 | |
1083 | if (MacroEnd) |
1084 | *MacroEnd = ExpInfo.getExpansionLocEnd(); |
1085 | return true; |
1086 | } |
1087 | |
1088 | //===----------------------------------------------------------------------===// |
1089 | // Queries about the code at a SourceLocation. |
1090 | //===----------------------------------------------------------------------===// |
1091 | |
1092 | /// getCharacterData - Return a pointer to the start of the specified location |
1093 | /// in the appropriate MemoryBuffer. |
1094 | const char *SourceManager::getCharacterData(SourceLocation SL, |
1095 | bool *Invalid) const { |
1096 | // Note that this is a hot function in the getSpelling() path, which is |
1097 | // heavily used by -E mode. |
1098 | std::pair<FileID, unsigned> LocInfo = getDecomposedSpellingLoc(Loc: SL); |
1099 | |
1100 | // Note that calling 'getBuffer()' may lazily page in a source file. |
1101 | bool CharDataInvalid = false; |
1102 | const SLocEntry &Entry = getSLocEntry(FID: LocInfo.first, Invalid: &CharDataInvalid); |
1103 | if (CharDataInvalid || !Entry.isFile()) { |
1104 | if (Invalid) |
1105 | *Invalid = true; |
1106 | |
1107 | return "<<<<INVALID BUFFER>>>>" ; |
1108 | } |
1109 | std::optional<llvm::MemoryBufferRef> Buffer = |
1110 | Entry.getFile().getContentCache().getBufferOrNone(Diag, FM&: getFileManager(), |
1111 | Loc: SourceLocation()); |
1112 | if (Invalid) |
1113 | *Invalid = !Buffer; |
1114 | return Buffer ? Buffer->getBufferStart() + LocInfo.second |
1115 | : "<<<<INVALID BUFFER>>>>" ; |
1116 | } |
1117 | |
1118 | /// getColumnNumber - Return the column # for the specified file position. |
1119 | /// this is significantly cheaper to compute than the line number. |
1120 | unsigned SourceManager::getColumnNumber(FileID FID, unsigned FilePos, |
1121 | bool *Invalid) const { |
1122 | std::optional<llvm::MemoryBufferRef> MemBuf = getBufferOrNone(FID); |
1123 | if (Invalid) |
1124 | *Invalid = !MemBuf; |
1125 | |
1126 | if (!MemBuf) |
1127 | return 1; |
1128 | |
1129 | // It is okay to request a position just past the end of the buffer. |
1130 | if (FilePos > MemBuf->getBufferSize()) { |
1131 | if (Invalid) |
1132 | *Invalid = true; |
1133 | return 1; |
1134 | } |
1135 | |
1136 | const char *Buf = MemBuf->getBufferStart(); |
1137 | // See if we just calculated the line number for this FilePos and can use |
1138 | // that to lookup the start of the line instead of searching for it. |
1139 | if (LastLineNoFileIDQuery == FID && LastLineNoContentCache->SourceLineCache && |
1140 | LastLineNoResult < LastLineNoContentCache->SourceLineCache.size()) { |
1141 | const unsigned *SourceLineCache = |
1142 | LastLineNoContentCache->SourceLineCache.begin(); |
1143 | unsigned LineStart = SourceLineCache[LastLineNoResult - 1]; |
1144 | unsigned LineEnd = SourceLineCache[LastLineNoResult]; |
1145 | if (FilePos >= LineStart && FilePos < LineEnd) { |
1146 | // LineEnd is the LineStart of the next line. |
1147 | // A line ends with separator LF or CR+LF on Windows. |
1148 | // FilePos might point to the last separator, |
1149 | // but we need a column number at most 1 + the last column. |
1150 | if (FilePos + 1 == LineEnd && FilePos > LineStart) { |
1151 | if (Buf[FilePos - 1] == '\r' || Buf[FilePos - 1] == '\n') |
1152 | --FilePos; |
1153 | } |
1154 | return FilePos - LineStart + 1; |
1155 | } |
1156 | } |
1157 | |
1158 | unsigned LineStart = FilePos; |
1159 | while (LineStart && Buf[LineStart-1] != '\n' && Buf[LineStart-1] != '\r') |
1160 | --LineStart; |
1161 | return FilePos-LineStart+1; |
1162 | } |
1163 | |
1164 | // isInvalid - Return the result of calling loc.isInvalid(), and |
1165 | // if Invalid is not null, set its value to same. |
1166 | template<typename LocType> |
1167 | static bool isInvalid(LocType Loc, bool *Invalid) { |
1168 | bool MyInvalid = Loc.isInvalid(); |
1169 | if (Invalid) |
1170 | *Invalid = MyInvalid; |
1171 | return MyInvalid; |
1172 | } |
1173 | |
1174 | unsigned SourceManager::getSpellingColumnNumber(SourceLocation Loc, |
1175 | bool *Invalid) const { |
1176 | if (isInvalid(Loc, Invalid)) return 0; |
1177 | std::pair<FileID, unsigned> LocInfo = getDecomposedSpellingLoc(Loc); |
1178 | return getColumnNumber(FID: LocInfo.first, FilePos: LocInfo.second, Invalid); |
1179 | } |
1180 | |
1181 | unsigned SourceManager::getExpansionColumnNumber(SourceLocation Loc, |
1182 | bool *Invalid) const { |
1183 | if (isInvalid(Loc, Invalid)) return 0; |
1184 | std::pair<FileID, unsigned> LocInfo = getDecomposedExpansionLoc(Loc); |
1185 | return getColumnNumber(FID: LocInfo.first, FilePos: LocInfo.second, Invalid); |
1186 | } |
1187 | |
1188 | unsigned SourceManager::getPresumedColumnNumber(SourceLocation Loc, |
1189 | bool *Invalid) const { |
1190 | PresumedLoc PLoc = getPresumedLoc(Loc); |
1191 | if (isInvalid(Loc: PLoc, Invalid)) return 0; |
1192 | return PLoc.getColumn(); |
1193 | } |
1194 | |
1195 | // Check if mutli-byte word x has bytes between m and n, included. This may also |
1196 | // catch bytes equal to n + 1. |
1197 | // The returned value holds a 0x80 at each byte position that holds a match. |
1198 | // see http://graphics.stanford.edu/~seander/bithacks.html#HasBetweenInWord |
1199 | template <class T> |
1200 | static constexpr inline T likelyhasbetween(T x, unsigned char m, |
1201 | unsigned char n) { |
1202 | return ((x - ~static_cast<T>(0) / 255 * (n + 1)) & ~x & |
1203 | ((x & ~static_cast<T>(0) / 255 * 127) + |
1204 | (~static_cast<T>(0) / 255 * (127 - (m - 1))))) & |
1205 | ~static_cast<T>(0) / 255 * 128; |
1206 | } |
1207 | |
1208 | LineOffsetMapping LineOffsetMapping::get(llvm::MemoryBufferRef Buffer, |
1209 | llvm::BumpPtrAllocator &Alloc) { |
1210 | |
1211 | // Find the file offsets of all of the *physical* source lines. This does |
1212 | // not look at trigraphs, escaped newlines, or anything else tricky. |
1213 | SmallVector<unsigned, 256> LineOffsets; |
1214 | |
1215 | // Line #1 starts at char 0. |
1216 | LineOffsets.push_back(Elt: 0); |
1217 | |
1218 | const unsigned char *Start = (const unsigned char *)Buffer.getBufferStart(); |
1219 | const unsigned char *End = (const unsigned char *)Buffer.getBufferEnd(); |
1220 | const unsigned char *Buf = Start; |
1221 | |
1222 | uint64_t Word; |
1223 | |
1224 | // scan sizeof(Word) bytes at a time for new lines. |
1225 | // This is much faster than scanning each byte independently. |
1226 | if ((unsigned long)(End - Start) > sizeof(Word)) { |
1227 | do { |
1228 | Word = llvm::support::endian::read64(P: Buf, E: llvm::endianness::little); |
1229 | // no new line => jump over sizeof(Word) bytes. |
1230 | auto Mask = likelyhasbetween(x: Word, m: '\n', n: '\r'); |
1231 | if (!Mask) { |
1232 | Buf += sizeof(Word); |
1233 | continue; |
1234 | } |
1235 | |
1236 | // At that point, Mask contains 0x80 set at each byte that holds a value |
1237 | // in [\n, \r + 1 [ |
1238 | |
1239 | // Scan for the next newline - it's very likely there's one. |
1240 | unsigned N = llvm::countr_zero(Val: Mask) - 7; // -7 because 0x80 is the marker |
1241 | Word >>= N; |
1242 | Buf += N / 8 + 1; |
1243 | unsigned char Byte = Word; |
1244 | switch (Byte) { |
1245 | case '\r': |
1246 | // If this is \r\n, skip both characters. |
1247 | if (*Buf == '\n') { |
1248 | ++Buf; |
1249 | } |
1250 | [[fallthrough]]; |
1251 | case '\n': |
1252 | LineOffsets.push_back(Elt: Buf - Start); |
1253 | }; |
1254 | } while (Buf < End - sizeof(Word) - 1); |
1255 | } |
1256 | |
1257 | // Handle tail using a regular check. |
1258 | while (Buf < End) { |
1259 | if (*Buf == '\n') { |
1260 | LineOffsets.push_back(Elt: Buf - Start + 1); |
1261 | } else if (*Buf == '\r') { |
1262 | // If this is \r\n, skip both characters. |
1263 | if (Buf + 1 < End && Buf[1] == '\n') { |
1264 | ++Buf; |
1265 | } |
1266 | LineOffsets.push_back(Elt: Buf - Start + 1); |
1267 | } |
1268 | ++Buf; |
1269 | } |
1270 | |
1271 | return LineOffsetMapping(LineOffsets, Alloc); |
1272 | } |
1273 | |
1274 | LineOffsetMapping::LineOffsetMapping(ArrayRef<unsigned> LineOffsets, |
1275 | llvm::BumpPtrAllocator &Alloc) |
1276 | : Storage(Alloc.Allocate<unsigned>(Num: LineOffsets.size() + 1)) { |
1277 | Storage[0] = LineOffsets.size(); |
1278 | std::copy(first: LineOffsets.begin(), last: LineOffsets.end(), result: Storage + 1); |
1279 | } |
1280 | |
1281 | /// getLineNumber - Given a SourceLocation, return the spelling line number |
1282 | /// for the position indicated. This requires building and caching a table of |
1283 | /// line offsets for the MemoryBuffer, so this is not cheap: use only when |
1284 | /// about to emit a diagnostic. |
1285 | unsigned SourceManager::getLineNumber(FileID FID, unsigned FilePos, |
1286 | bool *Invalid) const { |
1287 | if (FID.isInvalid()) { |
1288 | if (Invalid) |
1289 | *Invalid = true; |
1290 | return 1; |
1291 | } |
1292 | |
1293 | const ContentCache *Content; |
1294 | if (LastLineNoFileIDQuery == FID) |
1295 | Content = LastLineNoContentCache; |
1296 | else { |
1297 | bool MyInvalid = false; |
1298 | const SLocEntry &Entry = getSLocEntry(FID, Invalid: &MyInvalid); |
1299 | if (MyInvalid || !Entry.isFile()) { |
1300 | if (Invalid) |
1301 | *Invalid = true; |
1302 | return 1; |
1303 | } |
1304 | |
1305 | Content = &Entry.getFile().getContentCache(); |
1306 | } |
1307 | |
1308 | // If this is the first use of line information for this buffer, compute the |
1309 | // SourceLineCache for it on demand. |
1310 | if (!Content->SourceLineCache) { |
1311 | std::optional<llvm::MemoryBufferRef> Buffer = |
1312 | Content->getBufferOrNone(Diag, FM&: getFileManager(), Loc: SourceLocation()); |
1313 | if (Invalid) |
1314 | *Invalid = !Buffer; |
1315 | if (!Buffer) |
1316 | return 1; |
1317 | |
1318 | Content->SourceLineCache = |
1319 | LineOffsetMapping::get(Buffer: *Buffer, Alloc&: ContentCacheAlloc); |
1320 | } else if (Invalid) |
1321 | *Invalid = false; |
1322 | |
1323 | // Okay, we know we have a line number table. Do a binary search to find the |
1324 | // line number that this character position lands on. |
1325 | const unsigned *SourceLineCache = Content->SourceLineCache.begin(); |
1326 | const unsigned *SourceLineCacheStart = SourceLineCache; |
1327 | const unsigned *SourceLineCacheEnd = Content->SourceLineCache.end(); |
1328 | |
1329 | unsigned QueriedFilePos = FilePos+1; |
1330 | |
1331 | // FIXME: I would like to be convinced that this code is worth being as |
1332 | // complicated as it is, binary search isn't that slow. |
1333 | // |
1334 | // If it is worth being optimized, then in my opinion it could be more |
1335 | // performant, simpler, and more obviously correct by just "galloping" outward |
1336 | // from the queried file position. In fact, this could be incorporated into a |
1337 | // generic algorithm such as lower_bound_with_hint. |
1338 | // |
1339 | // If someone gives me a test case where this matters, and I will do it! - DWD |
1340 | |
1341 | // If the previous query was to the same file, we know both the file pos from |
1342 | // that query and the line number returned. This allows us to narrow the |
1343 | // search space from the entire file to something near the match. |
1344 | if (LastLineNoFileIDQuery == FID) { |
1345 | if (QueriedFilePos >= LastLineNoFilePos) { |
1346 | // FIXME: Potential overflow? |
1347 | SourceLineCache = SourceLineCache+LastLineNoResult-1; |
1348 | |
1349 | // The query is likely to be nearby the previous one. Here we check to |
1350 | // see if it is within 5, 10 or 20 lines. It can be far away in cases |
1351 | // where big comment blocks and vertical whitespace eat up lines but |
1352 | // contribute no tokens. |
1353 | if (SourceLineCache+5 < SourceLineCacheEnd) { |
1354 | if (SourceLineCache[5] > QueriedFilePos) |
1355 | SourceLineCacheEnd = SourceLineCache+5; |
1356 | else if (SourceLineCache+10 < SourceLineCacheEnd) { |
1357 | if (SourceLineCache[10] > QueriedFilePos) |
1358 | SourceLineCacheEnd = SourceLineCache+10; |
1359 | else if (SourceLineCache+20 < SourceLineCacheEnd) { |
1360 | if (SourceLineCache[20] > QueriedFilePos) |
1361 | SourceLineCacheEnd = SourceLineCache+20; |
1362 | } |
1363 | } |
1364 | } |
1365 | } else { |
1366 | if (LastLineNoResult < Content->SourceLineCache.size()) |
1367 | SourceLineCacheEnd = SourceLineCache+LastLineNoResult+1; |
1368 | } |
1369 | } |
1370 | |
1371 | const unsigned *Pos = |
1372 | std::lower_bound(first: SourceLineCache, last: SourceLineCacheEnd, val: QueriedFilePos); |
1373 | unsigned LineNo = Pos-SourceLineCacheStart; |
1374 | |
1375 | LastLineNoFileIDQuery = FID; |
1376 | LastLineNoContentCache = Content; |
1377 | LastLineNoFilePos = QueriedFilePos; |
1378 | LastLineNoResult = LineNo; |
1379 | return LineNo; |
1380 | } |
1381 | |
1382 | unsigned SourceManager::getSpellingLineNumber(SourceLocation Loc, |
1383 | bool *Invalid) const { |
1384 | if (isInvalid(Loc, Invalid)) return 0; |
1385 | std::pair<FileID, unsigned> LocInfo = getDecomposedSpellingLoc(Loc); |
1386 | return getLineNumber(FID: LocInfo.first, FilePos: LocInfo.second); |
1387 | } |
1388 | unsigned SourceManager::getExpansionLineNumber(SourceLocation Loc, |
1389 | bool *Invalid) const { |
1390 | if (isInvalid(Loc, Invalid)) return 0; |
1391 | std::pair<FileID, unsigned> LocInfo = getDecomposedExpansionLoc(Loc); |
1392 | return getLineNumber(FID: LocInfo.first, FilePos: LocInfo.second); |
1393 | } |
1394 | unsigned SourceManager::getPresumedLineNumber(SourceLocation Loc, |
1395 | bool *Invalid) const { |
1396 | PresumedLoc PLoc = getPresumedLoc(Loc); |
1397 | if (isInvalid(Loc: PLoc, Invalid)) return 0; |
1398 | return PLoc.getLine(); |
1399 | } |
1400 | |
1401 | /// getFileCharacteristic - return the file characteristic of the specified |
1402 | /// source location, indicating whether this is a normal file, a system |
1403 | /// header, or an "implicit extern C" system header. |
1404 | /// |
1405 | /// This state can be modified with flags on GNU linemarker directives like: |
1406 | /// # 4 "foo.h" 3 |
1407 | /// which changes all source locations in the current file after that to be |
1408 | /// considered to be from a system header. |
1409 | SrcMgr::CharacteristicKind |
1410 | SourceManager::getFileCharacteristic(SourceLocation Loc) const { |
1411 | assert(Loc.isValid() && "Can't get file characteristic of invalid loc!" ); |
1412 | std::pair<FileID, unsigned> LocInfo = getDecomposedExpansionLoc(Loc); |
1413 | const SLocEntry *SEntry = getSLocEntryForFile(FID: LocInfo.first); |
1414 | if (!SEntry) |
1415 | return C_User; |
1416 | |
1417 | const SrcMgr::FileInfo &FI = SEntry->getFile(); |
1418 | |
1419 | // If there are no #line directives in this file, just return the whole-file |
1420 | // state. |
1421 | if (!FI.hasLineDirectives()) |
1422 | return FI.getFileCharacteristic(); |
1423 | |
1424 | assert(LineTable && "Can't have linetable entries without a LineTable!" ); |
1425 | // See if there is a #line directive before the location. |
1426 | const LineEntry *Entry = |
1427 | LineTable->FindNearestLineEntry(FID: LocInfo.first, Offset: LocInfo.second); |
1428 | |
1429 | // If this is before the first line marker, use the file characteristic. |
1430 | if (!Entry) |
1431 | return FI.getFileCharacteristic(); |
1432 | |
1433 | return Entry->FileKind; |
1434 | } |
1435 | |
1436 | /// Return the filename or buffer identifier of the buffer the location is in. |
1437 | /// Note that this name does not respect \#line directives. Use getPresumedLoc |
1438 | /// for normal clients. |
1439 | StringRef SourceManager::getBufferName(SourceLocation Loc, |
1440 | bool *Invalid) const { |
1441 | if (isInvalid(Loc, Invalid)) return "<invalid loc>" ; |
1442 | |
1443 | auto B = getBufferOrNone(FID: getFileID(SpellingLoc: Loc)); |
1444 | if (Invalid) |
1445 | *Invalid = !B; |
1446 | return B ? B->getBufferIdentifier() : "<invalid buffer>" ; |
1447 | } |
1448 | |
1449 | /// getPresumedLoc - This method returns the "presumed" location of a |
1450 | /// SourceLocation specifies. A "presumed location" can be modified by \#line |
1451 | /// or GNU line marker directives. This provides a view on the data that a |
1452 | /// user should see in diagnostics, for example. |
1453 | /// |
1454 | /// Note that a presumed location is always given as the expansion point of an |
1455 | /// expansion location, not at the spelling location. |
1456 | PresumedLoc SourceManager::getPresumedLoc(SourceLocation Loc, |
1457 | bool UseLineDirectives) const { |
1458 | if (Loc.isInvalid()) return PresumedLoc(); |
1459 | |
1460 | // Presumed locations are always for expansion points. |
1461 | std::pair<FileID, unsigned> LocInfo = getDecomposedExpansionLoc(Loc); |
1462 | |
1463 | bool Invalid = false; |
1464 | const SLocEntry &Entry = getSLocEntry(FID: LocInfo.first, Invalid: &Invalid); |
1465 | if (Invalid || !Entry.isFile()) |
1466 | return PresumedLoc(); |
1467 | |
1468 | const SrcMgr::FileInfo &FI = Entry.getFile(); |
1469 | const SrcMgr::ContentCache *C = &FI.getContentCache(); |
1470 | |
1471 | // To get the source name, first consult the FileEntry (if one exists) |
1472 | // before the MemBuffer as this will avoid unnecessarily paging in the |
1473 | // MemBuffer. |
1474 | FileID FID = LocInfo.first; |
1475 | StringRef Filename; |
1476 | if (C->OrigEntry) |
1477 | Filename = C->OrigEntry->getName(); |
1478 | else if (auto Buffer = C->getBufferOrNone(Diag, FM&: getFileManager())) |
1479 | Filename = Buffer->getBufferIdentifier(); |
1480 | |
1481 | unsigned LineNo = getLineNumber(FID: LocInfo.first, FilePos: LocInfo.second, Invalid: &Invalid); |
1482 | if (Invalid) |
1483 | return PresumedLoc(); |
1484 | unsigned ColNo = getColumnNumber(FID: LocInfo.first, FilePos: LocInfo.second, Invalid: &Invalid); |
1485 | if (Invalid) |
1486 | return PresumedLoc(); |
1487 | |
1488 | SourceLocation IncludeLoc = FI.getIncludeLoc(); |
1489 | |
1490 | // If we have #line directives in this file, update and overwrite the physical |
1491 | // location info if appropriate. |
1492 | if (UseLineDirectives && FI.hasLineDirectives()) { |
1493 | assert(LineTable && "Can't have linetable entries without a LineTable!" ); |
1494 | // See if there is a #line directive before this. If so, get it. |
1495 | if (const LineEntry *Entry = |
1496 | LineTable->FindNearestLineEntry(FID: LocInfo.first, Offset: LocInfo.second)) { |
1497 | // If the LineEntry indicates a filename, use it. |
1498 | if (Entry->FilenameID != -1) { |
1499 | Filename = LineTable->getFilename(ID: Entry->FilenameID); |
1500 | // The contents of files referenced by #line are not in the |
1501 | // SourceManager |
1502 | FID = FileID::get(V: 0); |
1503 | } |
1504 | |
1505 | // Use the line number specified by the LineEntry. This line number may |
1506 | // be multiple lines down from the line entry. Add the difference in |
1507 | // physical line numbers from the query point and the line marker to the |
1508 | // total. |
1509 | unsigned MarkerLineNo = getLineNumber(FID: LocInfo.first, FilePos: Entry->FileOffset); |
1510 | LineNo = Entry->LineNo + (LineNo-MarkerLineNo-1); |
1511 | |
1512 | // Note that column numbers are not molested by line markers. |
1513 | |
1514 | // Handle virtual #include manipulation. |
1515 | if (Entry->IncludeOffset) { |
1516 | IncludeLoc = getLocForStartOfFile(FID: LocInfo.first); |
1517 | IncludeLoc = IncludeLoc.getLocWithOffset(Offset: Entry->IncludeOffset); |
1518 | } |
1519 | } |
1520 | } |
1521 | |
1522 | return PresumedLoc(Filename.data(), FID, LineNo, ColNo, IncludeLoc); |
1523 | } |
1524 | |
1525 | /// Returns whether the PresumedLoc for a given SourceLocation is |
1526 | /// in the main file. |
1527 | /// |
1528 | /// This computes the "presumed" location for a SourceLocation, then checks |
1529 | /// whether it came from a file other than the main file. This is different |
1530 | /// from isWrittenInMainFile() because it takes line marker directives into |
1531 | /// account. |
1532 | bool SourceManager::isInMainFile(SourceLocation Loc) const { |
1533 | if (Loc.isInvalid()) return false; |
1534 | |
1535 | // Presumed locations are always for expansion points. |
1536 | std::pair<FileID, unsigned> LocInfo = getDecomposedExpansionLoc(Loc); |
1537 | |
1538 | const SLocEntry *Entry = getSLocEntryForFile(FID: LocInfo.first); |
1539 | if (!Entry) |
1540 | return false; |
1541 | |
1542 | const SrcMgr::FileInfo &FI = Entry->getFile(); |
1543 | |
1544 | // Check if there is a line directive for this location. |
1545 | if (FI.hasLineDirectives()) |
1546 | if (const LineEntry *Entry = |
1547 | LineTable->FindNearestLineEntry(FID: LocInfo.first, Offset: LocInfo.second)) |
1548 | if (Entry->IncludeOffset) |
1549 | return false; |
1550 | |
1551 | return FI.getIncludeLoc().isInvalid(); |
1552 | } |
1553 | |
1554 | /// The size of the SLocEntry that \p FID represents. |
1555 | unsigned SourceManager::getFileIDSize(FileID FID) const { |
1556 | bool Invalid = false; |
1557 | const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, Invalid: &Invalid); |
1558 | if (Invalid) |
1559 | return 0; |
1560 | |
1561 | int ID = FID.ID; |
1562 | SourceLocation::UIntTy NextOffset; |
1563 | if ((ID > 0 && unsigned(ID+1) == local_sloc_entry_size())) |
1564 | NextOffset = getNextLocalOffset(); |
1565 | else if (ID+1 == -1) |
1566 | NextOffset = MaxLoadedOffset; |
1567 | else |
1568 | NextOffset = getSLocEntry(FID: FileID::get(V: ID+1)).getOffset(); |
1569 | |
1570 | return NextOffset - Entry.getOffset() - 1; |
1571 | } |
1572 | |
1573 | //===----------------------------------------------------------------------===// |
1574 | // Other miscellaneous methods. |
1575 | //===----------------------------------------------------------------------===// |
1576 | |
1577 | /// Get the source location for the given file:line:col triplet. |
1578 | /// |
1579 | /// If the source file is included multiple times, the source location will |
1580 | /// be based upon an arbitrary inclusion. |
1581 | SourceLocation SourceManager::translateFileLineCol(const FileEntry *SourceFile, |
1582 | unsigned Line, |
1583 | unsigned Col) const { |
1584 | assert(SourceFile && "Null source file!" ); |
1585 | assert(Line && Col && "Line and column should start from 1!" ); |
1586 | |
1587 | FileID FirstFID = translateFile(SourceFile); |
1588 | return translateLineCol(FID: FirstFID, Line, Col); |
1589 | } |
1590 | |
1591 | /// Get the FileID for the given file. |
1592 | /// |
1593 | /// If the source file is included multiple times, the FileID will be the |
1594 | /// first inclusion. |
1595 | FileID SourceManager::translateFile(const FileEntry *SourceFile) const { |
1596 | assert(SourceFile && "Null source file!" ); |
1597 | |
1598 | // First, check the main file ID, since it is common to look for a |
1599 | // location in the main file. |
1600 | if (MainFileID.isValid()) { |
1601 | bool Invalid = false; |
1602 | const SLocEntry &MainSLoc = getSLocEntry(FID: MainFileID, Invalid: &Invalid); |
1603 | if (Invalid) |
1604 | return FileID(); |
1605 | |
1606 | if (MainSLoc.isFile()) { |
1607 | if (MainSLoc.getFile().getContentCache().OrigEntry == SourceFile) |
1608 | return MainFileID; |
1609 | } |
1610 | } |
1611 | |
1612 | // The location we're looking for isn't in the main file; look |
1613 | // through all of the local source locations. |
1614 | for (unsigned I = 0, N = local_sloc_entry_size(); I != N; ++I) { |
1615 | const SLocEntry &SLoc = getLocalSLocEntry(Index: I); |
1616 | if (SLoc.isFile() && |
1617 | SLoc.getFile().getContentCache().OrigEntry == SourceFile) |
1618 | return FileID::get(V: I); |
1619 | } |
1620 | |
1621 | // If that still didn't help, try the modules. |
1622 | for (unsigned I = 0, N = loaded_sloc_entry_size(); I != N; ++I) { |
1623 | const SLocEntry &SLoc = getLoadedSLocEntry(Index: I); |
1624 | if (SLoc.isFile() && |
1625 | SLoc.getFile().getContentCache().OrigEntry == SourceFile) |
1626 | return FileID::get(V: -int(I) - 2); |
1627 | } |
1628 | |
1629 | return FileID(); |
1630 | } |
1631 | |
1632 | /// Get the source location in \arg FID for the given line:col. |
1633 | /// Returns null location if \arg FID is not a file SLocEntry. |
1634 | SourceLocation SourceManager::translateLineCol(FileID FID, |
1635 | unsigned Line, |
1636 | unsigned Col) const { |
1637 | // Lines are used as a one-based index into a zero-based array. This assert |
1638 | // checks for possible buffer underruns. |
1639 | assert(Line && Col && "Line and column should start from 1!" ); |
1640 | |
1641 | if (FID.isInvalid()) |
1642 | return SourceLocation(); |
1643 | |
1644 | bool Invalid = false; |
1645 | const SLocEntry &Entry = getSLocEntry(FID, Invalid: &Invalid); |
1646 | if (Invalid) |
1647 | return SourceLocation(); |
1648 | |
1649 | if (!Entry.isFile()) |
1650 | return SourceLocation(); |
1651 | |
1652 | SourceLocation FileLoc = SourceLocation::getFileLoc(ID: Entry.getOffset()); |
1653 | |
1654 | if (Line == 1 && Col == 1) |
1655 | return FileLoc; |
1656 | |
1657 | const ContentCache *Content = &Entry.getFile().getContentCache(); |
1658 | |
1659 | // If this is the first use of line information for this buffer, compute the |
1660 | // SourceLineCache for it on demand. |
1661 | std::optional<llvm::MemoryBufferRef> Buffer = |
1662 | Content->getBufferOrNone(Diag, FM&: getFileManager()); |
1663 | if (!Buffer) |
1664 | return SourceLocation(); |
1665 | if (!Content->SourceLineCache) |
1666 | Content->SourceLineCache = |
1667 | LineOffsetMapping::get(Buffer: *Buffer, Alloc&: ContentCacheAlloc); |
1668 | |
1669 | if (Line > Content->SourceLineCache.size()) { |
1670 | unsigned Size = Buffer->getBufferSize(); |
1671 | if (Size > 0) |
1672 | --Size; |
1673 | return FileLoc.getLocWithOffset(Offset: Size); |
1674 | } |
1675 | |
1676 | unsigned FilePos = Content->SourceLineCache[Line - 1]; |
1677 | const char *Buf = Buffer->getBufferStart() + FilePos; |
1678 | unsigned BufLength = Buffer->getBufferSize() - FilePos; |
1679 | if (BufLength == 0) |
1680 | return FileLoc.getLocWithOffset(Offset: FilePos); |
1681 | |
1682 | unsigned i = 0; |
1683 | |
1684 | // Check that the given column is valid. |
1685 | while (i < BufLength-1 && i < Col-1 && Buf[i] != '\n' && Buf[i] != '\r') |
1686 | ++i; |
1687 | return FileLoc.getLocWithOffset(Offset: FilePos + i); |
1688 | } |
1689 | |
1690 | /// Compute a map of macro argument chunks to their expanded source |
1691 | /// location. Chunks that are not part of a macro argument will map to an |
1692 | /// invalid source location. e.g. if a file contains one macro argument at |
1693 | /// offset 100 with length 10, this is how the map will be formed: |
1694 | /// 0 -> SourceLocation() |
1695 | /// 100 -> Expanded macro arg location |
1696 | /// 110 -> SourceLocation() |
1697 | void SourceManager::computeMacroArgsCache(MacroArgsMap &MacroArgsCache, |
1698 | FileID FID) const { |
1699 | assert(FID.isValid()); |
1700 | |
1701 | // Initially no macro argument chunk is present. |
1702 | MacroArgsCache.insert(x: std::make_pair(x: 0, y: SourceLocation())); |
1703 | |
1704 | int ID = FID.ID; |
1705 | while (true) { |
1706 | ++ID; |
1707 | // Stop if there are no more FileIDs to check. |
1708 | if (ID > 0) { |
1709 | if (unsigned(ID) >= local_sloc_entry_size()) |
1710 | return; |
1711 | } else if (ID == -1) { |
1712 | return; |
1713 | } |
1714 | |
1715 | bool Invalid = false; |
1716 | const SrcMgr::SLocEntry &Entry = getSLocEntryByID(ID, Invalid: &Invalid); |
1717 | if (Invalid) |
1718 | return; |
1719 | if (Entry.isFile()) { |
1720 | auto& File = Entry.getFile(); |
1721 | if (File.getFileCharacteristic() == C_User_ModuleMap || |
1722 | File.getFileCharacteristic() == C_System_ModuleMap) |
1723 | continue; |
1724 | |
1725 | SourceLocation IncludeLoc = File.getIncludeLoc(); |
1726 | bool IncludedInFID = |
1727 | (IncludeLoc.isValid() && isInFileID(Loc: IncludeLoc, FID)) || |
1728 | // Predefined header doesn't have a valid include location in main |
1729 | // file, but any files created by it should still be skipped when |
1730 | // computing macro args expanded in the main file. |
1731 | (FID == MainFileID && Entry.getFile().getName() == "<built-in>" ); |
1732 | if (IncludedInFID) { |
1733 | // Skip the files/macros of the #include'd file, we only care about |
1734 | // macros that lexed macro arguments from our file. |
1735 | if (Entry.getFile().NumCreatedFIDs) |
1736 | ID += Entry.getFile().NumCreatedFIDs - 1 /*because of next ++ID*/; |
1737 | continue; |
1738 | } |
1739 | // If file was included but not from FID, there is no more files/macros |
1740 | // that may be "contained" in this file. |
1741 | if (IncludeLoc.isValid()) |
1742 | return; |
1743 | continue; |
1744 | } |
1745 | |
1746 | const ExpansionInfo &ExpInfo = Entry.getExpansion(); |
1747 | |
1748 | if (ExpInfo.getExpansionLocStart().isFileID()) { |
1749 | if (!isInFileID(Loc: ExpInfo.getExpansionLocStart(), FID)) |
1750 | return; // No more files/macros that may be "contained" in this file. |
1751 | } |
1752 | |
1753 | if (!ExpInfo.isMacroArgExpansion()) |
1754 | continue; |
1755 | |
1756 | associateFileChunkWithMacroArgExp(MacroArgsCache, FID, |
1757 | SpellLoc: ExpInfo.getSpellingLoc(), |
1758 | ExpansionLoc: SourceLocation::getMacroLoc(ID: Entry.getOffset()), |
1759 | ExpansionLength: getFileIDSize(FID: FileID::get(V: ID))); |
1760 | } |
1761 | } |
1762 | |
1763 | void SourceManager::associateFileChunkWithMacroArgExp( |
1764 | MacroArgsMap &MacroArgsCache, |
1765 | FileID FID, |
1766 | SourceLocation SpellLoc, |
1767 | SourceLocation ExpansionLoc, |
1768 | unsigned ExpansionLength) const { |
1769 | if (!SpellLoc.isFileID()) { |
1770 | SourceLocation::UIntTy SpellBeginOffs = SpellLoc.getOffset(); |
1771 | SourceLocation::UIntTy SpellEndOffs = SpellBeginOffs + ExpansionLength; |
1772 | |
1773 | // The spelling range for this macro argument expansion can span multiple |
1774 | // consecutive FileID entries. Go through each entry contained in the |
1775 | // spelling range and if one is itself a macro argument expansion, recurse |
1776 | // and associate the file chunk that it represents. |
1777 | |
1778 | FileID SpellFID; // Current FileID in the spelling range. |
1779 | unsigned SpellRelativeOffs; |
1780 | std::tie(args&: SpellFID, args&: SpellRelativeOffs) = getDecomposedLoc(Loc: SpellLoc); |
1781 | while (true) { |
1782 | const SLocEntry &Entry = getSLocEntry(FID: SpellFID); |
1783 | SourceLocation::UIntTy SpellFIDBeginOffs = Entry.getOffset(); |
1784 | unsigned SpellFIDSize = getFileIDSize(FID: SpellFID); |
1785 | SourceLocation::UIntTy SpellFIDEndOffs = SpellFIDBeginOffs + SpellFIDSize; |
1786 | const ExpansionInfo &Info = Entry.getExpansion(); |
1787 | if (Info.isMacroArgExpansion()) { |
1788 | unsigned CurrSpellLength; |
1789 | if (SpellFIDEndOffs < SpellEndOffs) |
1790 | CurrSpellLength = SpellFIDSize - SpellRelativeOffs; |
1791 | else |
1792 | CurrSpellLength = ExpansionLength; |
1793 | associateFileChunkWithMacroArgExp(MacroArgsCache, FID, |
1794 | SpellLoc: Info.getSpellingLoc().getLocWithOffset(Offset: SpellRelativeOffs), |
1795 | ExpansionLoc, ExpansionLength: CurrSpellLength); |
1796 | } |
1797 | |
1798 | if (SpellFIDEndOffs >= SpellEndOffs) |
1799 | return; // we covered all FileID entries in the spelling range. |
1800 | |
1801 | // Move to the next FileID entry in the spelling range. |
1802 | unsigned advance = SpellFIDSize - SpellRelativeOffs + 1; |
1803 | ExpansionLoc = ExpansionLoc.getLocWithOffset(Offset: advance); |
1804 | ExpansionLength -= advance; |
1805 | ++SpellFID.ID; |
1806 | SpellRelativeOffs = 0; |
1807 | } |
1808 | } |
1809 | |
1810 | assert(SpellLoc.isFileID()); |
1811 | |
1812 | unsigned BeginOffs; |
1813 | if (!isInFileID(Loc: SpellLoc, FID, RelativeOffset: &BeginOffs)) |
1814 | return; |
1815 | |
1816 | unsigned EndOffs = BeginOffs + ExpansionLength; |
1817 | |
1818 | // Add a new chunk for this macro argument. A previous macro argument chunk |
1819 | // may have been lexed again, so e.g. if the map is |
1820 | // 0 -> SourceLocation() |
1821 | // 100 -> Expanded loc #1 |
1822 | // 110 -> SourceLocation() |
1823 | // and we found a new macro FileID that lexed from offset 105 with length 3, |
1824 | // the new map will be: |
1825 | // 0 -> SourceLocation() |
1826 | // 100 -> Expanded loc #1 |
1827 | // 105 -> Expanded loc #2 |
1828 | // 108 -> Expanded loc #1 |
1829 | // 110 -> SourceLocation() |
1830 | // |
1831 | // Since re-lexed macro chunks will always be the same size or less of |
1832 | // previous chunks, we only need to find where the ending of the new macro |
1833 | // chunk is mapped to and update the map with new begin/end mappings. |
1834 | |
1835 | MacroArgsMap::iterator I = MacroArgsCache.upper_bound(x: EndOffs); |
1836 | --I; |
1837 | SourceLocation EndOffsMappedLoc = I->second; |
1838 | MacroArgsCache[BeginOffs] = ExpansionLoc; |
1839 | MacroArgsCache[EndOffs] = EndOffsMappedLoc; |
1840 | } |
1841 | |
1842 | /// If \arg Loc points inside a function macro argument, the returned |
1843 | /// location will be the macro location in which the argument was expanded. |
1844 | /// If a macro argument is used multiple times, the expanded location will |
1845 | /// be at the first expansion of the argument. |
1846 | /// e.g. |
1847 | /// MY_MACRO(foo); |
1848 | /// ^ |
1849 | /// Passing a file location pointing at 'foo', will yield a macro location |
1850 | /// where 'foo' was expanded into. |
1851 | SourceLocation |
1852 | SourceManager::getMacroArgExpandedLocation(SourceLocation Loc) const { |
1853 | if (Loc.isInvalid() || !Loc.isFileID()) |
1854 | return Loc; |
1855 | |
1856 | FileID FID; |
1857 | unsigned Offset; |
1858 | std::tie(args&: FID, args&: Offset) = getDecomposedLoc(Loc); |
1859 | if (FID.isInvalid()) |
1860 | return Loc; |
1861 | |
1862 | std::unique_ptr<MacroArgsMap> &MacroArgsCache = MacroArgsCacheMap[FID]; |
1863 | if (!MacroArgsCache) { |
1864 | MacroArgsCache = std::make_unique<MacroArgsMap>(); |
1865 | computeMacroArgsCache(MacroArgsCache&: *MacroArgsCache, FID); |
1866 | } |
1867 | |
1868 | assert(!MacroArgsCache->empty()); |
1869 | MacroArgsMap::iterator I = MacroArgsCache->upper_bound(x: Offset); |
1870 | // In case every element in MacroArgsCache is greater than Offset we can't |
1871 | // decrement the iterator. |
1872 | if (I == MacroArgsCache->begin()) |
1873 | return Loc; |
1874 | |
1875 | --I; |
1876 | |
1877 | SourceLocation::UIntTy MacroArgBeginOffs = I->first; |
1878 | SourceLocation MacroArgExpandedLoc = I->second; |
1879 | if (MacroArgExpandedLoc.isValid()) |
1880 | return MacroArgExpandedLoc.getLocWithOffset(Offset: Offset - MacroArgBeginOffs); |
1881 | |
1882 | return Loc; |
1883 | } |
1884 | |
1885 | std::pair<FileID, unsigned> |
1886 | SourceManager::getDecomposedIncludedLoc(FileID FID) const { |
1887 | if (FID.isInvalid()) |
1888 | return std::make_pair(x: FileID(), y: 0); |
1889 | |
1890 | // Uses IncludedLocMap to retrieve/cache the decomposed loc. |
1891 | |
1892 | using DecompTy = std::pair<FileID, unsigned>; |
1893 | auto InsertOp = IncludedLocMap.try_emplace(Key: FID); |
1894 | DecompTy &DecompLoc = InsertOp.first->second; |
1895 | if (!InsertOp.second) |
1896 | return DecompLoc; // already in map. |
1897 | |
1898 | SourceLocation UpperLoc; |
1899 | bool Invalid = false; |
1900 | const SrcMgr::SLocEntry &Entry = getSLocEntry(FID, Invalid: &Invalid); |
1901 | if (!Invalid) { |
1902 | if (Entry.isExpansion()) |
1903 | UpperLoc = Entry.getExpansion().getExpansionLocStart(); |
1904 | else |
1905 | UpperLoc = Entry.getFile().getIncludeLoc(); |
1906 | } |
1907 | |
1908 | if (UpperLoc.isValid()) |
1909 | DecompLoc = getDecomposedLoc(Loc: UpperLoc); |
1910 | |
1911 | return DecompLoc; |
1912 | } |
1913 | |
1914 | bool SourceManager::isInTheSameTranslationUnitImpl( |
1915 | const std::pair<FileID, unsigned> &LOffs, |
1916 | const std::pair<FileID, unsigned> &ROffs) const { |
1917 | // If one is local while the other is loaded. |
1918 | if (isLoadedFileID(FID: LOffs.first) != isLoadedFileID(FID: ROffs.first)) |
1919 | return false; |
1920 | |
1921 | if (isLoadedFileID(FID: LOffs.first) && isLoadedFileID(FID: ROffs.first)) { |
1922 | auto FindSLocEntryAlloc = [this](FileID FID) { |
1923 | // Loaded FileIDs are negative, we store the lowest FileID from each |
1924 | // allocation, later allocations have lower FileIDs. |
1925 | return llvm::lower_bound(Range: LoadedSLocEntryAllocBegin, Value&: FID, |
1926 | C: std::greater<FileID>{}); |
1927 | }; |
1928 | |
1929 | // If both are loaded from different AST files. |
1930 | if (FindSLocEntryAlloc(LOffs.first) != FindSLocEntryAlloc(ROffs.first)) |
1931 | return false; |
1932 | } |
1933 | |
1934 | return true; |
1935 | } |
1936 | |
1937 | /// Given a decomposed source location, move it up the include/expansion stack |
1938 | /// to the parent source location within the same translation unit. If this is |
1939 | /// possible, return the decomposed version of the parent in Loc and return |
1940 | /// false. If Loc is a top-level entry, return true and don't modify it. |
1941 | static bool |
1942 | MoveUpTranslationUnitIncludeHierarchy(std::pair<FileID, unsigned> &Loc, |
1943 | const SourceManager &SM) { |
1944 | std::pair<FileID, unsigned> UpperLoc = SM.getDecomposedIncludedLoc(FID: Loc.first); |
1945 | if (UpperLoc.first.isInvalid() || |
1946 | !SM.isInTheSameTranslationUnitImpl(LOffs: UpperLoc, ROffs: Loc)) |
1947 | return true; // We reached the top. |
1948 | |
1949 | Loc = UpperLoc; |
1950 | return false; |
1951 | } |
1952 | |
1953 | /// Return the cache entry for comparing the given file IDs |
1954 | /// for isBeforeInTranslationUnit. |
1955 | InBeforeInTUCacheEntry &SourceManager::getInBeforeInTUCache(FileID LFID, |
1956 | FileID RFID) const { |
1957 | // This is a magic number for limiting the cache size. It was experimentally |
1958 | // derived from a small Objective-C project (where the cache filled |
1959 | // out to ~250 items). We can make it larger if necessary. |
1960 | // FIXME: this is almost certainly full these days. Use an LRU cache? |
1961 | enum { MagicCacheSize = 300 }; |
1962 | IsBeforeInTUCacheKey Key(LFID, RFID); |
1963 | |
1964 | // If the cache size isn't too large, do a lookup and if necessary default |
1965 | // construct an entry. We can then return it to the caller for direct |
1966 | // use. When they update the value, the cache will get automatically |
1967 | // updated as well. |
1968 | if (IBTUCache.size() < MagicCacheSize) |
1969 | return IBTUCache.try_emplace(Key, Args&: LFID, Args&: RFID).first->second; |
1970 | |
1971 | // Otherwise, do a lookup that will not construct a new value. |
1972 | InBeforeInTUCache::iterator I = IBTUCache.find(Val: Key); |
1973 | if (I != IBTUCache.end()) |
1974 | return I->second; |
1975 | |
1976 | // Fall back to the overflow value. |
1977 | IBTUCacheOverflow.setQueryFIDs(LHS: LFID, RHS: RFID); |
1978 | return IBTUCacheOverflow; |
1979 | } |
1980 | |
1981 | /// Determines the order of 2 source locations in the translation unit. |
1982 | /// |
1983 | /// \returns true if LHS source location comes before RHS, false otherwise. |
1984 | bool SourceManager::isBeforeInTranslationUnit(SourceLocation LHS, |
1985 | SourceLocation RHS) const { |
1986 | assert(LHS.isValid() && RHS.isValid() && "Passed invalid source location!" ); |
1987 | if (LHS == RHS) |
1988 | return false; |
1989 | |
1990 | std::pair<FileID, unsigned> LOffs = getDecomposedLoc(Loc: LHS); |
1991 | std::pair<FileID, unsigned> ROffs = getDecomposedLoc(Loc: RHS); |
1992 | |
1993 | // getDecomposedLoc may have failed to return a valid FileID because, e.g. it |
1994 | // is a serialized one referring to a file that was removed after we loaded |
1995 | // the PCH. |
1996 | if (LOffs.first.isInvalid() || ROffs.first.isInvalid()) |
1997 | return LOffs.first.isInvalid() && !ROffs.first.isInvalid(); |
1998 | |
1999 | std::pair<bool, bool> InSameTU = isInTheSameTranslationUnit(LOffs, ROffs); |
2000 | if (InSameTU.first) |
2001 | return InSameTU.second; |
2002 | // TODO: This should be unreachable, but some clients are calling this |
2003 | // function before making sure LHS and RHS are in the same TU. |
2004 | return LOffs.first < ROffs.first; |
2005 | } |
2006 | |
2007 | std::pair<bool, bool> SourceManager::isInTheSameTranslationUnit( |
2008 | std::pair<FileID, unsigned> &LOffs, |
2009 | std::pair<FileID, unsigned> &ROffs) const { |
2010 | // If the source locations are not in the same TU, return early. |
2011 | if (!isInTheSameTranslationUnitImpl(LOffs, ROffs)) |
2012 | return std::make_pair(x: false, y: false); |
2013 | |
2014 | // If the source locations are in the same file, just compare offsets. |
2015 | if (LOffs.first == ROffs.first) |
2016 | return std::make_pair(x: true, y: LOffs.second < ROffs.second); |
2017 | |
2018 | // If we are comparing a source location with multiple locations in the same |
2019 | // file, we get a big win by caching the result. |
2020 | InBeforeInTUCacheEntry &IsBeforeInTUCache = |
2021 | getInBeforeInTUCache(LFID: LOffs.first, RFID: ROffs.first); |
2022 | |
2023 | // If we are comparing a source location with multiple locations in the same |
2024 | // file, we get a big win by caching the result. |
2025 | if (IsBeforeInTUCache.isCacheValid()) |
2026 | return std::make_pair( |
2027 | x: true, y: IsBeforeInTUCache.getCachedResult(LOffset: LOffs.second, ROffset: ROffs.second)); |
2028 | |
2029 | // Okay, we missed in the cache, we'll compute the answer and populate it. |
2030 | // We need to find the common ancestor. The only way of doing this is to |
2031 | // build the complete include chain for one and then walking up the chain |
2032 | // of the other looking for a match. |
2033 | |
2034 | // A location within a FileID on the path up from LOffs to the main file. |
2035 | struct Entry { |
2036 | std::pair<FileID, unsigned> DecomposedLoc; // FileID redundant, but clearer. |
2037 | FileID ChildFID; // Used for breaking ties. Invalid for the initial loc. |
2038 | }; |
2039 | llvm::SmallDenseMap<FileID, Entry, 16> LChain; |
2040 | |
2041 | FileID LChild; |
2042 | do { |
2043 | LChain.try_emplace(Key: LOffs.first, Args: Entry{.DecomposedLoc: LOffs, .ChildFID: LChild}); |
2044 | // We catch the case where LOffs is in a file included by ROffs and |
2045 | // quit early. The other way round unfortunately remains suboptimal. |
2046 | if (LOffs.first == ROffs.first) |
2047 | break; |
2048 | LChild = LOffs.first; |
2049 | } while (!MoveUpTranslationUnitIncludeHierarchy(Loc&: LOffs, SM: *this)); |
2050 | |
2051 | FileID RChild; |
2052 | do { |
2053 | auto LIt = LChain.find(Val: ROffs.first); |
2054 | if (LIt != LChain.end()) { |
2055 | // Compare the locations within the common file and cache them. |
2056 | LOffs = LIt->second.DecomposedLoc; |
2057 | LChild = LIt->second.ChildFID; |
2058 | // The relative order of LChild and RChild is a tiebreaker when |
2059 | // - locs expand to the same location (occurs in macro arg expansion) |
2060 | // - one loc is a parent of the other (we consider the parent as "first") |
2061 | // For the parent entry to be first, its invalid child file ID must |
2062 | // compare smaller to the valid child file ID of the other entry. |
2063 | // However loaded FileIDs are <0, so we perform *unsigned* comparison! |
2064 | // This changes the relative order of local vs loaded FileIDs, but it |
2065 | // doesn't matter as these are never mixed in macro expansion. |
2066 | unsigned LChildID = LChild.ID; |
2067 | unsigned RChildID = RChild.ID; |
2068 | assert(((LOffs.second != ROffs.second) || |
2069 | (LChildID == 0 || RChildID == 0) || |
2070 | isInSameSLocAddrSpace(getComposedLoc(LChild, 0), |
2071 | getComposedLoc(RChild, 0), nullptr)) && |
2072 | "Mixed local/loaded FileIDs with same include location?" ); |
2073 | IsBeforeInTUCache.setCommonLoc(commonFID: LOffs.first, lCommonOffset: LOffs.second, rCommonOffset: ROffs.second, |
2074 | LParentBeforeRParent: LChildID < RChildID); |
2075 | return std::make_pair( |
2076 | x: true, y: IsBeforeInTUCache.getCachedResult(LOffset: LOffs.second, ROffset: ROffs.second)); |
2077 | } |
2078 | RChild = ROffs.first; |
2079 | } while (!MoveUpTranslationUnitIncludeHierarchy(Loc&: ROffs, SM: *this)); |
2080 | |
2081 | // If we found no match, the location is either in a built-ins buffer or |
2082 | // associated with global inline asm. PR5662 and PR22576 are examples. |
2083 | |
2084 | StringRef LB = getBufferOrFake(FID: LOffs.first).getBufferIdentifier(); |
2085 | StringRef RB = getBufferOrFake(FID: ROffs.first).getBufferIdentifier(); |
2086 | |
2087 | bool LIsBuiltins = LB == "<built-in>" ; |
2088 | bool RIsBuiltins = RB == "<built-in>" ; |
2089 | // Sort built-in before non-built-in. |
2090 | if (LIsBuiltins || RIsBuiltins) { |
2091 | if (LIsBuiltins != RIsBuiltins) |
2092 | return std::make_pair(x: true, y&: LIsBuiltins); |
2093 | // Both are in built-in buffers, but from different files. We just claim |
2094 | // that lower IDs come first. |
2095 | return std::make_pair(x: true, y: LOffs.first < ROffs.first); |
2096 | } |
2097 | |
2098 | bool LIsAsm = LB == "<inline asm>" ; |
2099 | bool RIsAsm = RB == "<inline asm>" ; |
2100 | // Sort assembler after built-ins, but before the rest. |
2101 | if (LIsAsm || RIsAsm) { |
2102 | if (LIsAsm != RIsAsm) |
2103 | return std::make_pair(x: true, y&: RIsAsm); |
2104 | assert(LOffs.first == ROffs.first); |
2105 | return std::make_pair(x: true, y: false); |
2106 | } |
2107 | |
2108 | bool LIsScratch = LB == "<scratch space>" ; |
2109 | bool RIsScratch = RB == "<scratch space>" ; |
2110 | // Sort scratch after inline asm, but before the rest. |
2111 | if (LIsScratch || RIsScratch) { |
2112 | if (LIsScratch != RIsScratch) |
2113 | return std::make_pair(x: true, y&: LIsScratch); |
2114 | return std::make_pair(x: true, y: LOffs.second < ROffs.second); |
2115 | } |
2116 | |
2117 | llvm_unreachable("Unsortable locations found" ); |
2118 | } |
2119 | |
2120 | void SourceManager::PrintStats() const { |
2121 | llvm::errs() << "\n*** Source Manager Stats:\n" ; |
2122 | llvm::errs() << FileInfos.size() << " files mapped, " << MemBufferInfos.size() |
2123 | << " mem buffers mapped.\n" ; |
2124 | llvm::errs() << LocalSLocEntryTable.size() << " local SLocEntries allocated (" |
2125 | << llvm::capacity_in_bytes(X: LocalSLocEntryTable) |
2126 | << " bytes of capacity), " << NextLocalOffset |
2127 | << "B of SLoc address space used.\n" ; |
2128 | llvm::errs() << LoadedSLocEntryTable.size() |
2129 | << " loaded SLocEntries allocated (" |
2130 | << llvm::capacity_in_bytes(x: LoadedSLocEntryTable) |
2131 | << " bytes of capacity), " |
2132 | << MaxLoadedOffset - CurrentLoadedOffset |
2133 | << "B of SLoc address space used.\n" ; |
2134 | |
2135 | unsigned NumLineNumsComputed = 0; |
2136 | unsigned NumFileBytesMapped = 0; |
2137 | for (fileinfo_iterator I = fileinfo_begin(), E = fileinfo_end(); I != E; ++I){ |
2138 | NumLineNumsComputed += bool(I->second->SourceLineCache); |
2139 | NumFileBytesMapped += I->second->getSizeBytesMapped(); |
2140 | } |
2141 | unsigned NumMacroArgsComputed = MacroArgsCacheMap.size(); |
2142 | |
2143 | llvm::errs() << NumFileBytesMapped << " bytes of files mapped, " |
2144 | << NumLineNumsComputed << " files with line #'s computed, " |
2145 | << NumMacroArgsComputed << " files with macro args computed.\n" ; |
2146 | llvm::errs() << "FileID scans: " << NumLinearScans << " linear, " |
2147 | << NumBinaryProbes << " binary.\n" ; |
2148 | } |
2149 | |
2150 | LLVM_DUMP_METHOD void SourceManager::dump() const { |
2151 | llvm::raw_ostream &out = llvm::errs(); |
2152 | |
2153 | auto DumpSLocEntry = [&](int ID, const SrcMgr::SLocEntry &Entry, |
2154 | std::optional<SourceLocation::UIntTy> NextStart) { |
2155 | out << "SLocEntry <FileID " << ID << "> " << (Entry.isFile() ? "file" : "expansion" ) |
2156 | << " <SourceLocation " << Entry.getOffset() << ":" ; |
2157 | if (NextStart) |
2158 | out << *NextStart << ">\n" ; |
2159 | else |
2160 | out << "???\?>\n" ; |
2161 | if (Entry.isFile()) { |
2162 | auto &FI = Entry.getFile(); |
2163 | if (FI.NumCreatedFIDs) |
2164 | out << " covers <FileID " << ID << ":" << int(ID + FI.NumCreatedFIDs) |
2165 | << ">\n" ; |
2166 | if (FI.getIncludeLoc().isValid()) |
2167 | out << " included from " << FI.getIncludeLoc().getOffset() << "\n" ; |
2168 | auto &CC = FI.getContentCache(); |
2169 | out << " for " << (CC.OrigEntry ? CC.OrigEntry->getName() : "<none>" ) |
2170 | << "\n" ; |
2171 | if (CC.BufferOverridden) |
2172 | out << " contents overridden\n" ; |
2173 | if (CC.ContentsEntry != CC.OrigEntry) { |
2174 | out << " contents from " |
2175 | << (CC.ContentsEntry ? CC.ContentsEntry->getName() : "<none>" ) |
2176 | << "\n" ; |
2177 | } |
2178 | } else { |
2179 | auto &EI = Entry.getExpansion(); |
2180 | out << " spelling from " << EI.getSpellingLoc().getOffset() << "\n" ; |
2181 | out << " macro " << (EI.isMacroArgExpansion() ? "arg" : "body" ) |
2182 | << " range <" << EI.getExpansionLocStart().getOffset() << ":" |
2183 | << EI.getExpansionLocEnd().getOffset() << ">\n" ; |
2184 | } |
2185 | }; |
2186 | |
2187 | // Dump local SLocEntries. |
2188 | for (unsigned ID = 0, NumIDs = LocalSLocEntryTable.size(); ID != NumIDs; ++ID) { |
2189 | DumpSLocEntry(ID, LocalSLocEntryTable[ID], |
2190 | ID == NumIDs - 1 ? NextLocalOffset |
2191 | : LocalSLocEntryTable[ID + 1].getOffset()); |
2192 | } |
2193 | // Dump loaded SLocEntries. |
2194 | std::optional<SourceLocation::UIntTy> NextStart; |
2195 | for (unsigned Index = 0; Index != LoadedSLocEntryTable.size(); ++Index) { |
2196 | int ID = -(int)Index - 2; |
2197 | if (SLocEntryLoaded[Index]) { |
2198 | DumpSLocEntry(ID, LoadedSLocEntryTable[Index], NextStart); |
2199 | NextStart = LoadedSLocEntryTable[Index].getOffset(); |
2200 | } else { |
2201 | NextStart = std::nullopt; |
2202 | } |
2203 | } |
2204 | } |
2205 | |
2206 | void SourceManager::noteSLocAddressSpaceUsage( |
2207 | DiagnosticsEngine &Diag, std::optional<unsigned> MaxNotes) const { |
2208 | struct Info { |
2209 | // A location where this file was entered. |
2210 | SourceLocation Loc; |
2211 | // Number of times this FileEntry was entered. |
2212 | unsigned Inclusions = 0; |
2213 | // Size usage from the file itself. |
2214 | uint64_t DirectSize = 0; |
2215 | // Total size usage from the file and its macro expansions. |
2216 | uint64_t TotalSize = 0; |
2217 | }; |
2218 | using UsageMap = llvm::MapVector<const FileEntry*, Info>; |
2219 | |
2220 | UsageMap Usage; |
2221 | uint64_t CountedSize = 0; |
2222 | |
2223 | auto AddUsageForFileID = [&](FileID ID) { |
2224 | // The +1 here is because getFileIDSize doesn't include the extra byte for |
2225 | // the one-past-the-end location. |
2226 | unsigned Size = getFileIDSize(FID: ID) + 1; |
2227 | |
2228 | // Find the file that used this address space, either directly or by |
2229 | // macro expansion. |
2230 | SourceLocation FileStart = getFileLoc(Loc: getComposedLoc(FID: ID, Offset: 0)); |
2231 | FileID FileLocID = getFileID(SpellingLoc: FileStart); |
2232 | const FileEntry *Entry = getFileEntryForID(FID: FileLocID); |
2233 | |
2234 | Info &EntryInfo = Usage[Entry]; |
2235 | if (EntryInfo.Loc.isInvalid()) |
2236 | EntryInfo.Loc = FileStart; |
2237 | if (ID == FileLocID) { |
2238 | ++EntryInfo.Inclusions; |
2239 | EntryInfo.DirectSize += Size; |
2240 | } |
2241 | EntryInfo.TotalSize += Size; |
2242 | CountedSize += Size; |
2243 | }; |
2244 | |
2245 | // Loaded SLocEntries have indexes counting downwards from -2. |
2246 | for (size_t Index = 0; Index != LoadedSLocEntryTable.size(); ++Index) { |
2247 | AddUsageForFileID(FileID::get(V: -2 - Index)); |
2248 | } |
2249 | // Local SLocEntries have indexes counting upwards from 0. |
2250 | for (size_t Index = 0; Index != LocalSLocEntryTable.size(); ++Index) { |
2251 | AddUsageForFileID(FileID::get(V: Index)); |
2252 | } |
2253 | |
2254 | // Sort the usage by size from largest to smallest. Break ties by raw source |
2255 | // location. |
2256 | auto SortedUsage = Usage.takeVector(); |
2257 | auto Cmp = [](const UsageMap::value_type &A, const UsageMap::value_type &B) { |
2258 | return A.second.TotalSize > B.second.TotalSize || |
2259 | (A.second.TotalSize == B.second.TotalSize && |
2260 | A.second.Loc < B.second.Loc); |
2261 | }; |
2262 | auto SortedEnd = SortedUsage.end(); |
2263 | if (MaxNotes && SortedUsage.size() > *MaxNotes) { |
2264 | SortedEnd = SortedUsage.begin() + *MaxNotes; |
2265 | std::nth_element(first: SortedUsage.begin(), nth: SortedEnd, last: SortedUsage.end(), comp: Cmp); |
2266 | } |
2267 | std::sort(first: SortedUsage.begin(), last: SortedEnd, comp: Cmp); |
2268 | |
2269 | // Produce note on sloc address space usage total. |
2270 | uint64_t LocalUsage = NextLocalOffset; |
2271 | uint64_t LoadedUsage = MaxLoadedOffset - CurrentLoadedOffset; |
2272 | int UsagePercent = static_cast<int>(100.0 * double(LocalUsage + LoadedUsage) / |
2273 | MaxLoadedOffset); |
2274 | Diag.Report(SourceLocation(), diag::note_total_sloc_usage) |
2275 | << LocalUsage << LoadedUsage << (LocalUsage + LoadedUsage) << UsagePercent; |
2276 | |
2277 | // Produce notes on sloc address space usage for each file with a high usage. |
2278 | uint64_t ReportedSize = 0; |
2279 | for (auto &[Entry, FileInfo] : |
2280 | llvm::make_range(x: SortedUsage.begin(), y: SortedEnd)) { |
2281 | Diag.Report(FileInfo.Loc, diag::note_file_sloc_usage) |
2282 | << FileInfo.Inclusions << FileInfo.DirectSize |
2283 | << (FileInfo.TotalSize - FileInfo.DirectSize); |
2284 | ReportedSize += FileInfo.TotalSize; |
2285 | } |
2286 | |
2287 | // Describe any remaining usage not reported in the per-file usage. |
2288 | if (ReportedSize != CountedSize) { |
2289 | Diag.Report(SourceLocation(), diag::note_file_misc_sloc_usage) |
2290 | << (SortedUsage.end() - SortedEnd) << CountedSize - ReportedSize; |
2291 | } |
2292 | } |
2293 | |
2294 | ExternalSLocEntrySource::~ExternalSLocEntrySource() = default; |
2295 | |
2296 | /// Return the amount of memory used by memory buffers, breaking down |
2297 | /// by heap-backed versus mmap'ed memory. |
2298 | SourceManager::MemoryBufferSizes SourceManager::getMemoryBufferSizes() const { |
2299 | size_t malloc_bytes = 0; |
2300 | size_t mmap_bytes = 0; |
2301 | |
2302 | for (unsigned i = 0, e = MemBufferInfos.size(); i != e; ++i) |
2303 | if (size_t sized_mapped = MemBufferInfos[i]->getSizeBytesMapped()) |
2304 | switch (MemBufferInfos[i]->getMemoryBufferKind()) { |
2305 | case llvm::MemoryBuffer::MemoryBuffer_MMap: |
2306 | mmap_bytes += sized_mapped; |
2307 | break; |
2308 | case llvm::MemoryBuffer::MemoryBuffer_Malloc: |
2309 | malloc_bytes += sized_mapped; |
2310 | break; |
2311 | } |
2312 | |
2313 | return MemoryBufferSizes(malloc_bytes, mmap_bytes); |
2314 | } |
2315 | |
2316 | size_t SourceManager::getDataStructureSizes() const { |
2317 | size_t size = llvm::capacity_in_bytes(x: MemBufferInfos) + |
2318 | llvm::capacity_in_bytes(X: LocalSLocEntryTable) + |
2319 | llvm::capacity_in_bytes(x: LoadedSLocEntryTable) + |
2320 | llvm::capacity_in_bytes(X: SLocEntryLoaded) + |
2321 | llvm::capacity_in_bytes(X: FileInfos); |
2322 | |
2323 | if (OverriddenFilesInfo) |
2324 | size += llvm::capacity_in_bytes(X: OverriddenFilesInfo->OverriddenFiles); |
2325 | |
2326 | return size; |
2327 | } |
2328 | |
2329 | SourceManagerForFile::SourceManagerForFile(StringRef FileName, |
2330 | StringRef Content) { |
2331 | // This is referenced by `FileMgr` and will be released by `FileMgr` when it |
2332 | // is deleted. |
2333 | IntrusiveRefCntPtr<llvm::vfs::InMemoryFileSystem> InMemoryFileSystem( |
2334 | new llvm::vfs::InMemoryFileSystem); |
2335 | InMemoryFileSystem->addFile( |
2336 | Path: FileName, ModificationTime: 0, |
2337 | Buffer: llvm::MemoryBuffer::getMemBuffer(InputData: Content, BufferName: FileName, |
2338 | /*RequiresNullTerminator=*/false)); |
2339 | // This is passed to `SM` as reference, so the pointer has to be referenced |
2340 | // in `Environment` so that `FileMgr` can out-live this function scope. |
2341 | FileMgr = |
2342 | std::make_unique<FileManager>(args: FileSystemOptions(), args&: InMemoryFileSystem); |
2343 | // This is passed to `SM` as reference, so the pointer has to be referenced |
2344 | // by `Environment` due to the same reason above. |
2345 | Diagnostics = std::make_unique<DiagnosticsEngine>( |
2346 | args: IntrusiveRefCntPtr<DiagnosticIDs>(new DiagnosticIDs), |
2347 | args: new DiagnosticOptions); |
2348 | SourceMgr = std::make_unique<SourceManager>(args&: *Diagnostics, args&: *FileMgr); |
2349 | FileEntryRef FE = llvm::cantFail(ValOrErr: FileMgr->getFileRef(Filename: FileName)); |
2350 | FileID ID = |
2351 | SourceMgr->createFileID(SourceFile: FE, IncludePos: SourceLocation(), FileCharacter: clang::SrcMgr::C_User); |
2352 | assert(ID.isValid()); |
2353 | SourceMgr->setMainFileID(ID); |
2354 | } |
2355 | |