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