Bug Summary

File:llvm/include/llvm/Bitstream/BitstreamReader.h
Warning:line 221, column 39
The result of the right shift is undefined due to shifting by '64', which is greater or equal to the width of type 'llvm::SimpleBitstreamCursor::word_t'

Annotated Source Code

Press '?' to see keyboard shortcuts

clang -cc1 -cc1 -triple x86_64-pc-linux-gnu -analyze -disable-free -disable-llvm-verifier -discard-value-names -main-file-name GlobalModuleIndex.cpp -analyzer-store=region -analyzer-opt-analyze-nested-blocks -analyzer-checker=core -analyzer-checker=apiModeling -analyzer-checker=unix -analyzer-checker=deadcode -analyzer-checker=cplusplus -analyzer-checker=security.insecureAPI.UncheckedReturn -analyzer-checker=security.insecureAPI.getpw -analyzer-checker=security.insecureAPI.gets -analyzer-checker=security.insecureAPI.mktemp -analyzer-checker=security.insecureAPI.mkstemp -analyzer-checker=security.insecureAPI.vfork -analyzer-checker=nullability.NullPassedToNonnull -analyzer-checker=nullability.NullReturnedFromNonnull -analyzer-output plist -w -setup-static-analyzer -analyzer-config-compatibility-mode=true -mrelocation-model pic -pic-level 2 -mframe-pointer=none -relaxed-aliasing -fmath-errno -fno-rounding-math -mconstructor-aliases -munwind-tables -target-cpu x86-64 -tune-cpu generic -debugger-tuning=gdb -ffunction-sections -fdata-sections -fcoverage-compilation-dir=/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/build-llvm/tools/clang/lib/Serialization -resource-dir /usr/lib/llvm-14/lib/clang/14.0.0 -D CLANG_ROUND_TRIP_CC1_ARGS=ON -D _DEBUG -D _GNU_SOURCE -D __STDC_CONSTANT_MACROS -D __STDC_FORMAT_MACROS -D __STDC_LIMIT_MACROS -I /build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/build-llvm/tools/clang/lib/Serialization -I /build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/clang/lib/Serialization -I /build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/clang/include -I /build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/build-llvm/tools/clang/include -I /build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/build-llvm/include -I /build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include -D NDEBUG -U NDEBUG -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/10/../../../../include/c++/10 -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/10/../../../../include/x86_64-linux-gnu/c++/10 -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/10/../../../../include/c++/10/backward -internal-isystem /usr/lib/llvm-14/lib/clang/14.0.0/include -internal-isystem /usr/local/include -internal-isystem /usr/lib/gcc/x86_64-linux-gnu/10/../../../../x86_64-linux-gnu/include -internal-externc-isystem /usr/include/x86_64-linux-gnu -internal-externc-isystem /include -internal-externc-isystem /usr/include -O2 -Wno-unused-parameter -Wwrite-strings -Wno-missing-field-initializers -Wno-long-long -Wno-maybe-uninitialized -Wno-class-memaccess -Wno-redundant-move -Wno-pessimizing-move -Wno-noexcept-type -Wno-comment -std=c++14 -fdeprecated-macro -fdebug-compilation-dir=/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/build-llvm/tools/clang/lib/Serialization -fdebug-prefix-map=/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0=. -ferror-limit 19 -fvisibility-inlines-hidden -stack-protector 2 -fgnuc-version=4.2.1 -vectorize-loops -vectorize-slp -analyzer-output=html -analyzer-config stable-report-filename=true -faddrsig -D__GCC_HAVE_DWARF2_CFI_ASM=1 -o /tmp/scan-build-2021-08-28-193554-24367-1 -x c++ /build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/clang/lib/Serialization/GlobalModuleIndex.cpp

/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/clang/lib/Serialization/GlobalModuleIndex.cpp

→
1//===--- GlobalModuleIndex.cpp - Global Module Index ------------*- C++ -*-===//
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 GlobalModuleIndex class.
10//
11//===----------------------------------------------------------------------===//
12
13#include "clang/Serialization/GlobalModuleIndex.h"
14#include "ASTReaderInternals.h"
15#include "clang/Basic/FileManager.h"
16#include "clang/Lex/HeaderSearch.h"
17#include "clang/Serialization/ASTBitCodes.h"
18#include "clang/Serialization/ModuleFile.h"
19#include "clang/Serialization/PCHContainerOperations.h"
20#include "llvm/ADT/DenseMap.h"
21#include "llvm/ADT/MapVector.h"
22#include "llvm/ADT/SmallString.h"
23#include "llvm/ADT/StringRef.h"
24#include "llvm/Bitstream/BitstreamReader.h"
25#include "llvm/Bitstream/BitstreamWriter.h"
26#include "llvm/Support/DJB.h"
27#include "llvm/Support/FileSystem.h"
28#include "llvm/Support/FileUtilities.h"
29#include "llvm/Support/LockFileManager.h"
30#include "llvm/Support/MemoryBuffer.h"
31#include "llvm/Support/OnDiskHashTable.h"
32#include "llvm/Support/Path.h"
33#include "llvm/Support/TimeProfiler.h"
34#include <cstdio>
35using namespace clang;
36using namespace serialization;
37
38//----------------------------------------------------------------------------//
39// Shared constants
40//----------------------------------------------------------------------------//
41namespace {
42 enum {
43 /// The block containing the index.
44 GLOBAL_INDEX_BLOCK_ID = llvm::bitc::FIRST_APPLICATION_BLOCKID
45 };
46
47 /// Describes the record types in the index.
48 enum IndexRecordTypes {
49 /// Contains version information and potentially other metadata,
50 /// used to determine if we can read this global index file.
51 INDEX_METADATA,
52 /// Describes a module, including its file name and dependencies.
53 MODULE,
54 /// The index for identifiers.
55 IDENTIFIER_INDEX
56 };
57}
58
59/// The name of the global index file.
60static const char * const IndexFileName = "modules.idx";
61
62/// The global index file version.
63static const unsigned CurrentVersion = 1;
64
65//----------------------------------------------------------------------------//
66// Global module index reader.
67//----------------------------------------------------------------------------//
68
69namespace {
70
71/// Trait used to read the identifier index from the on-disk hash
72/// table.
73class IdentifierIndexReaderTrait {
74public:
75 typedef StringRef external_key_type;
76 typedef StringRef internal_key_type;
77 typedef SmallVector<unsigned, 2> data_type;
78 typedef unsigned hash_value_type;
79 typedef unsigned offset_type;
80
81 static bool EqualKey(const internal_key_type& a, const internal_key_type& b) {
82 return a == b;
83 }
84
85 static hash_value_type ComputeHash(const internal_key_type& a) {
86 return llvm::djbHash(a);
87 }
88
89 static std::pair<unsigned, unsigned>
90 ReadKeyDataLength(const unsigned char*& d) {
91 using namespace llvm::support;
92 unsigned KeyLen = endian::readNext<uint16_t, little, unaligned>(d);
93 unsigned DataLen = endian::readNext<uint16_t, little, unaligned>(d);
94 return std::make_pair(KeyLen, DataLen);
95 }
96
97 static const internal_key_type&
98 GetInternalKey(const external_key_type& x) { return x; }
99
100 static const external_key_type&
101 GetExternalKey(const internal_key_type& x) { return x; }
102
103 static internal_key_type ReadKey(const unsigned char* d, unsigned n) {
104 return StringRef((const char *)d, n);
105 }
106
107 static data_type ReadData(const internal_key_type& k,
108 const unsigned char* d,
109 unsigned DataLen) {
110 using namespace llvm::support;
111
112 data_type Result;
113 while (DataLen > 0) {
114 unsigned ID = endian::readNext<uint32_t, little, unaligned>(d);
115 Result.push_back(ID);
116 DataLen -= 4;
117 }
118
119 return Result;
120 }
121};
122
123typedef llvm::OnDiskIterableChainedHashTable<IdentifierIndexReaderTrait>
124 IdentifierIndexTable;
125
126}
127
128GlobalModuleIndex::GlobalModuleIndex(
129 std::unique_ptr<llvm::MemoryBuffer> IndexBuffer,
130 llvm::BitstreamCursor Cursor)
131 : Buffer(std::move(IndexBuffer)), IdentifierIndex(), NumIdentifierLookups(),
132 NumIdentifierLookupHits() {
133 auto Fail = [&](llvm::Error &&Err) {
134 report_fatal_error("Module index '" + Buffer->getBufferIdentifier() +
135 "' failed: " + toString(std::move(Err)));
136 };
137
138 llvm::TimeTraceScope TimeScope("Module LoadIndex");
139 // Read the global index.
140 bool InGlobalIndexBlock = false;
141 bool Done = false;
142 while (!Done) {
143 llvm::BitstreamEntry Entry;
144 if (Expected<llvm::BitstreamEntry> Res = Cursor.advance())
145 Entry = Res.get();
146 else
147 Fail(Res.takeError());
148
149 switch (Entry.Kind) {
150 case llvm::BitstreamEntry::Error:
151 return;
152
153 case llvm::BitstreamEntry::EndBlock:
154 if (InGlobalIndexBlock) {
155 InGlobalIndexBlock = false;
156 Done = true;
157 continue;
158 }
159 return;
160
161
162 case llvm::BitstreamEntry::Record:
163 // Entries in the global index block are handled below.
164 if (InGlobalIndexBlock)
165 break;
166
167 return;
168
169 case llvm::BitstreamEntry::SubBlock:
170 if (!InGlobalIndexBlock && Entry.ID == GLOBAL_INDEX_BLOCK_ID) {
171 if (llvm::Error Err = Cursor.EnterSubBlock(GLOBAL_INDEX_BLOCK_ID))
172 Fail(std::move(Err));
173 InGlobalIndexBlock = true;
174 } else if (llvm::Error Err = Cursor.SkipBlock())
175 Fail(std::move(Err));
176 continue;
177 }
178
179 SmallVector<uint64_t, 64> Record;
180 StringRef Blob;
181 Expected<unsigned> MaybeIndexRecord =
182 Cursor.readRecord(Entry.ID, Record, &Blob);
183 if (!MaybeIndexRecord)
184 Fail(MaybeIndexRecord.takeError());
185 IndexRecordTypes IndexRecord =
186 static_cast<IndexRecordTypes>(MaybeIndexRecord.get());
187 switch (IndexRecord) {
188 case INDEX_METADATA:
189 // Make sure that the version matches.
190 if (Record.size() < 1 || Record[0] != CurrentVersion)
191 return;
192 break;
193
194 case MODULE: {
195 unsigned Idx = 0;
196 unsigned ID = Record[Idx++];
197
198 // Make room for this module's information.
199 if (ID == Modules.size())
200 Modules.push_back(ModuleInfo());
201 else
202 Modules.resize(ID + 1);
203
204 // Size/modification time for this module file at the time the
205 // global index was built.
206 Modules[ID].Size = Record[Idx++];
207 Modules[ID].ModTime = Record[Idx++];
208
209 // File name.
210 unsigned NameLen = Record[Idx++];
211 Modules[ID].FileName.assign(Record.begin() + Idx,
212 Record.begin() + Idx + NameLen);
213 Idx += NameLen;
214
215 // Dependencies
216 unsigned NumDeps = Record[Idx++];
217 Modules[ID].Dependencies.insert(Modules[ID].Dependencies.end(),
218 Record.begin() + Idx,
219 Record.begin() + Idx + NumDeps);
220 Idx += NumDeps;
221
222 // Make sure we're at the end of the record.
223 assert(Idx == Record.size() && "More module info?")(static_cast <bool> (Idx == Record.size() && "More module info?"
) ? void (0) : __assert_fail ("Idx == Record.size() && \"More module info?\""
, "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/clang/lib/Serialization/GlobalModuleIndex.cpp"
, 223, __extension__ __PRETTY_FUNCTION__));
224
225 // Record this module as an unresolved module.
226 // FIXME: this doesn't work correctly for module names containing path
227 // separators.
228 StringRef ModuleName = llvm::sys::path::stem(Modules[ID].FileName);
229 // Remove the -<hash of ModuleMapPath>
230 ModuleName = ModuleName.rsplit('-').first;
231 UnresolvedModules[ModuleName] = ID;
232 break;
233 }
234
235 case IDENTIFIER_INDEX:
236 // Wire up the identifier index.
237 if (Record[0]) {
238 IdentifierIndex = IdentifierIndexTable::Create(
239 (const unsigned char *)Blob.data() + Record[0],
240 (const unsigned char *)Blob.data() + sizeof(uint32_t),
241 (const unsigned char *)Blob.data(), IdentifierIndexReaderTrait());
242 }
243 break;
244 }
245 }
246}
247
248GlobalModuleIndex::~GlobalModuleIndex() {
249 delete static_cast<IdentifierIndexTable *>(IdentifierIndex);
250}
251
252std::pair<GlobalModuleIndex *, llvm::Error>
253GlobalModuleIndex::readIndex(StringRef Path) {
254 // Load the index file, if it's there.
255 llvm::SmallString<128> IndexPath;
256 IndexPath += Path;
257 llvm::sys::path::append(IndexPath, IndexFileName);
258
259 llvm::ErrorOr<std::unique_ptr<llvm::MemoryBuffer>> BufferOrErr =
260 llvm::MemoryBuffer::getFile(IndexPath.c_str());
261 if (!BufferOrErr)
262 return std::make_pair(nullptr,
263 llvm::errorCodeToError(BufferOrErr.getError()));
264 std::unique_ptr<llvm::MemoryBuffer> Buffer = std::move(BufferOrErr.get());
265
266 /// The main bitstream cursor for the main block.
267 llvm::BitstreamCursor Cursor(*Buffer);
268
269 // Sniff for the signature.
270 for (unsigned char C : {'B', 'C', 'G', 'I'}) {
271 if (Expected<llvm::SimpleBitstreamCursor::word_t> Res = Cursor.Read(8)) {
272 if (Res.get() != C)
273 return std::make_pair(
274 nullptr, llvm::createStringError(std::errc::illegal_byte_sequence,
275 "expected signature BCGI"));
276 } else
277 return std::make_pair(nullptr, Res.takeError());
278 }
279
280 return std::make_pair(new GlobalModuleIndex(std::move(Buffer), Cursor),
281 llvm::Error::success());
282}
283
284void
285GlobalModuleIndex::getKnownModules(SmallVectorImpl<ModuleFile *> &ModuleFiles) {
286 ModuleFiles.clear();
287 for (unsigned I = 0, N = Modules.size(); I != N; ++I) {
288 if (ModuleFile *MF = Modules[I].File)
289 ModuleFiles.push_back(MF);
290 }
291}
292
293void GlobalModuleIndex::getModuleDependencies(
294 ModuleFile *File,
295 SmallVectorImpl<ModuleFile *> &Dependencies) {
296 // Look for information about this module file.
297 llvm::DenseMap<ModuleFile *, unsigned>::iterator Known
298 = ModulesByFile.find(File);
299 if (Known == ModulesByFile.end())
300 return;
301
302 // Record dependencies.
303 Dependencies.clear();
304 ArrayRef<unsigned> StoredDependencies = Modules[Known->second].Dependencies;
305 for (unsigned I = 0, N = StoredDependencies.size(); I != N; ++I) {
306 if (ModuleFile *MF = Modules[I].File)
307 Dependencies.push_back(MF);
308 }
309}
310
311bool GlobalModuleIndex::lookupIdentifier(StringRef Name, HitSet &Hits) {
312 Hits.clear();
313
314 // If there's no identifier index, there is nothing we can do.
315 if (!IdentifierIndex)
316 return false;
317
318 // Look into the identifier index.
319 ++NumIdentifierLookups;
320 IdentifierIndexTable &Table
321 = *static_cast<IdentifierIndexTable *>(IdentifierIndex);
322 IdentifierIndexTable::iterator Known = Table.find(Name);
323 if (Known == Table.end()) {
324 return false;
325 }
326
327 SmallVector<unsigned, 2> ModuleIDs = *Known;
328 for (unsigned I = 0, N = ModuleIDs.size(); I != N; ++I) {
329 if (ModuleFile *MF = Modules[ModuleIDs[I]].File)
330 Hits.insert(MF);
331 }
332
333 ++NumIdentifierLookupHits;
334 return true;
335}
336
337bool GlobalModuleIndex::loadedModuleFile(ModuleFile *File) {
338 // Look for the module in the global module index based on the module name.
339 StringRef Name = File->ModuleName;
340 llvm::StringMap<unsigned>::iterator Known = UnresolvedModules.find(Name);
341 if (Known == UnresolvedModules.end()) {
342 return true;
343 }
344
345 // Rectify this module with the global module index.
346 ModuleInfo &Info = Modules[Known->second];
347
348 // If the size and modification time match what we expected, record this
349 // module file.
350 bool Failed = true;
351 if (File->File->getSize() == Info.Size &&
352 File->File->getModificationTime() == Info.ModTime) {
353 Info.File = File;
354 ModulesByFile[File] = Known->second;
355
356 Failed = false;
357 }
358
359 // One way or another, we have resolved this module file.
360 UnresolvedModules.erase(Known);
361 return Failed;
362}
363
364void GlobalModuleIndex::printStats() {
365 std::fprintf(stderrstderr, "*** Global Module Index Statistics:\n");
366 if (NumIdentifierLookups) {
367 fprintf(stderrstderr, " %u / %u identifier lookups succeeded (%f%%)\n",
368 NumIdentifierLookupHits, NumIdentifierLookups,
369 (double)NumIdentifierLookupHits*100.0/NumIdentifierLookups);
370 }
371 std::fprintf(stderrstderr, "\n");
372}
373
374LLVM_DUMP_METHOD__attribute__((noinline)) __attribute__((__used__)) void GlobalModuleIndex::dump() {
375 llvm::errs() << "*** Global Module Index Dump:\n";
376 llvm::errs() << "Module files:\n";
377 for (auto &MI : Modules) {
378 llvm::errs() << "** " << MI.FileName << "\n";
379 if (MI.File)
380 MI.File->dump();
381 else
382 llvm::errs() << "\n";
383 }
384 llvm::errs() << "\n";
385}
386
387//----------------------------------------------------------------------------//
388// Global module index writer.
389//----------------------------------------------------------------------------//
390
391namespace {
392 /// Provides information about a specific module file.
393 struct ModuleFileInfo {
394 /// The numberic ID for this module file.
395 unsigned ID;
396
397 /// The set of modules on which this module depends. Each entry is
398 /// a module ID.
399 SmallVector<unsigned, 4> Dependencies;
400 ASTFileSignature Signature;
401 };
402
403 struct ImportedModuleFileInfo {
404 off_t StoredSize;
405 time_t StoredModTime;
406 ASTFileSignature StoredSignature;
407 ImportedModuleFileInfo(off_t Size, time_t ModTime, ASTFileSignature Sig)
408 : StoredSize(Size), StoredModTime(ModTime), StoredSignature(Sig) {}
409 };
410
411 /// Builder that generates the global module index file.
412 class GlobalModuleIndexBuilder {
413 FileManager &FileMgr;
414 const PCHContainerReader &PCHContainerRdr;
415
416 /// Mapping from files to module file information.
417 typedef llvm::MapVector<const FileEntry *, ModuleFileInfo> ModuleFilesMap;
418
419 /// Information about each of the known module files.
420 ModuleFilesMap ModuleFiles;
421
422 /// Mapping from the imported module file to the imported
423 /// information.
424 typedef std::multimap<const FileEntry *, ImportedModuleFileInfo>
425 ImportedModuleFilesMap;
426
427 /// Information about each importing of a module file.
428 ImportedModuleFilesMap ImportedModuleFiles;
429
430 /// Mapping from identifiers to the list of module file IDs that
431 /// consider this identifier to be interesting.
432 typedef llvm::StringMap<SmallVector<unsigned, 2> > InterestingIdentifierMap;
433
434 /// A mapping from all interesting identifiers to the set of module
435 /// files in which those identifiers are considered interesting.
436 InterestingIdentifierMap InterestingIdentifiers;
437
438 /// Write the block-info block for the global module index file.
439 void emitBlockInfoBlock(llvm::BitstreamWriter &Stream);
440
441 /// Retrieve the module file information for the given file.
442 ModuleFileInfo &getModuleFileInfo(const FileEntry *File) {
443 llvm::MapVector<const FileEntry *, ModuleFileInfo>::iterator Known
444 = ModuleFiles.find(File);
445 if (Known != ModuleFiles.end())
446 return Known->second;
447
448 unsigned NewID = ModuleFiles.size();
449 ModuleFileInfo &Info = ModuleFiles[File];
450 Info.ID = NewID;
451 return Info;
452 }
453
454 public:
455 explicit GlobalModuleIndexBuilder(
456 FileManager &FileMgr, const PCHContainerReader &PCHContainerRdr)
457 : FileMgr(FileMgr), PCHContainerRdr(PCHContainerRdr) {}
458
459 /// Load the contents of the given module file into the builder.
460 llvm::Error loadModuleFile(const FileEntry *File);
461
462 /// Write the index to the given bitstream.
463 /// \returns true if an error occurred, false otherwise.
464 bool writeIndex(llvm::BitstreamWriter &Stream);
465 };
466}
467
468static void emitBlockID(unsigned ID, const char *Name,
469 llvm::BitstreamWriter &Stream,
470 SmallVectorImpl<uint64_t> &Record) {
471 Record.clear();
472 Record.push_back(ID);
473 Stream.EmitRecord(llvm::bitc::BLOCKINFO_CODE_SETBID, Record);
474
475 // Emit the block name if present.
476 if (!Name || Name[0] == 0) return;
477 Record.clear();
478 while (*Name)
479 Record.push_back(*Name++);
480 Stream.EmitRecord(llvm::bitc::BLOCKINFO_CODE_BLOCKNAME, Record);
481}
482
483static void emitRecordID(unsigned ID, const char *Name,
484 llvm::BitstreamWriter &Stream,
485 SmallVectorImpl<uint64_t> &Record) {
486 Record.clear();
487 Record.push_back(ID);
488 while (*Name)
489 Record.push_back(*Name++);
490 Stream.EmitRecord(llvm::bitc::BLOCKINFO_CODE_SETRECORDNAME, Record);
491}
492
493void
494GlobalModuleIndexBuilder::emitBlockInfoBlock(llvm::BitstreamWriter &Stream) {
495 SmallVector<uint64_t, 64> Record;
496 Stream.EnterBlockInfoBlock();
497
498#define BLOCK(X) emitBlockID(X ## _ID, #X, Stream, Record)
499#define RECORD(X) emitRecordID(X, #X, Stream, Record)
500 BLOCK(GLOBAL_INDEX_BLOCK);
501 RECORD(INDEX_METADATA);
502 RECORD(MODULE);
503 RECORD(IDENTIFIER_INDEX);
504#undef RECORD
505#undef BLOCK
506
507 Stream.ExitBlock();
508}
509
510namespace {
511 class InterestingASTIdentifierLookupTrait
512 : public serialization::reader::ASTIdentifierLookupTraitBase {
513
514 public:
515 /// The identifier and whether it is "interesting".
516 typedef std::pair<StringRef, bool> data_type;
517
518 data_type ReadData(const internal_key_type& k,
519 const unsigned char* d,
520 unsigned DataLen) {
521 // The first bit indicates whether this identifier is interesting.
522 // That's all we care about.
523 using namespace llvm::support;
524 unsigned RawID = endian::readNext<uint32_t, little, unaligned>(d);
525 bool IsInteresting = RawID & 0x01;
526 return std::make_pair(k, IsInteresting);
527 }
528 };
529}
530
531llvm::Error GlobalModuleIndexBuilder::loadModuleFile(const FileEntry *File) {
532 // Open the module file.
533
534 auto Buffer = FileMgr.getBufferForFile(File, /*isVolatile=*/true);
535 if (!Buffer)
7
←
Taking false branch
→
536 return llvm::createStringError(Buffer.getError(),
537 "failed getting buffer for module file");
538
539 // Initialize the input stream
540 llvm::BitstreamCursor InStream(PCHContainerRdr.ExtractPCH(**Buffer));
541
542 // Sniff for the signature.
543 for (unsigned char C : {'C', 'P', 'C', 'H'})
8
←
Assuming '__begin1' is equal to '__end1'
→
544 if (Expected<llvm::SimpleBitstreamCursor::word_t> Res = InStream.Read(8)) {
545 if (Res.get() != C)
546 return llvm::createStringError(std::errc::illegal_byte_sequence,
547 "expected signature CPCH");
548 } else
549 return Res.takeError();
550
551 // Record this module file and assign it a unique ID (if it doesn't have
552 // one already).
553 unsigned ID = getModuleFileInfo(File).ID;
554
555 // Search for the blocks and records we care about.
556 enum { Other, ControlBlock, ASTBlock, DiagnosticOptionsBlock } State = Other;
557 bool Done = false;
558 while (!Done) {
9
←
Loop condition is true. Entering loop body
→
559 Expected<llvm::BitstreamEntry> MaybeEntry = InStream.advance();
10
←
Calling 'BitstreamCursor::advance'
→
560 if (!MaybeEntry)
561 return MaybeEntry.takeError();
562 llvm::BitstreamEntry Entry = MaybeEntry.get();
563
564 switch (Entry.Kind) {
565 case llvm::BitstreamEntry::Error:
566 Done = true;
567 continue;
568
569 case llvm::BitstreamEntry::Record:
570 // In the 'other' state, just skip the record. We don't care.
571 if (State == Other) {
572 if (llvm::Expected<unsigned> Skipped = InStream.skipRecord(Entry.ID))
573 continue;
574 else
575 return Skipped.takeError();
576 }
577
578 // Handle potentially-interesting records below.
579 break;
580
581 case llvm::BitstreamEntry::SubBlock:
582 if (Entry.ID == CONTROL_BLOCK_ID) {
583 if (llvm::Error Err = InStream.EnterSubBlock(CONTROL_BLOCK_ID))
584 return Err;
585
586 // Found the control block.
587 State = ControlBlock;
588 continue;
589 }
590
591 if (Entry.ID == AST_BLOCK_ID) {
592 if (llvm::Error Err = InStream.EnterSubBlock(AST_BLOCK_ID))
593 return Err;
594
595 // Found the AST block.
596 State = ASTBlock;
597 continue;
598 }
599
600 if (Entry.ID == UNHASHED_CONTROL_BLOCK_ID) {
601 if (llvm::Error Err = InStream.EnterSubBlock(UNHASHED_CONTROL_BLOCK_ID))
602 return Err;
603
604 // Found the Diagnostic Options block.
605 State = DiagnosticOptionsBlock;
606 continue;
607 }
608
609 if (llvm::Error Err = InStream.SkipBlock())
610 return Err;
611
612 continue;
613
614 case llvm::BitstreamEntry::EndBlock:
615 State = Other;
616 continue;
617 }
618
619 // Read the given record.
620 SmallVector<uint64_t, 64> Record;
621 StringRef Blob;
622 Expected<unsigned> MaybeCode = InStream.readRecord(Entry.ID, Record, &Blob);
623 if (!MaybeCode)
624 return MaybeCode.takeError();
625 unsigned Code = MaybeCode.get();
626
627 // Handle module dependencies.
628 if (State == ControlBlock && Code == IMPORTS) {
629 // Load each of the imported PCH files.
630 unsigned Idx = 0, N = Record.size();
631 while (Idx < N) {
632 // Read information about the AST file.
633
634 // Skip the imported kind
635 ++Idx;
636
637 // Skip the import location
638 ++Idx;
639
640 // Load stored size/modification time.
641 off_t StoredSize = (off_t)Record[Idx++];
642 time_t StoredModTime = (time_t)Record[Idx++];
643
644 // Skip the stored signature.
645 // FIXME: we could read the signature out of the import and validate it.
646 auto FirstSignatureByte = Record.begin() + Idx;
647 ASTFileSignature StoredSignature = ASTFileSignature::create(
648 FirstSignatureByte, FirstSignatureByte + ASTFileSignature::size);
649 Idx += ASTFileSignature::size;
650
651 // Skip the module name (currently this is only used for prebuilt
652 // modules while here we are only dealing with cached).
653 Idx += Record[Idx] + 1;
654
655 // Retrieve the imported file name.
656 unsigned Length = Record[Idx++];
657 SmallString<128> ImportedFile(Record.begin() + Idx,
658 Record.begin() + Idx + Length);
659 Idx += Length;
660
661 // Find the imported module file.
662 auto DependsOnFile
663 = FileMgr.getFile(ImportedFile, /*OpenFile=*/false,
664 /*CacheFailure=*/false);
665
666 if (!DependsOnFile)
667 return llvm::createStringError(std::errc::bad_file_descriptor,
668 "imported file \"%s\" not found",
669 ImportedFile.c_str());
670
671 // Save the information in ImportedModuleFileInfo so we can verify after
672 // loading all pcms.
673 ImportedModuleFiles.insert(std::make_pair(
674 *DependsOnFile, ImportedModuleFileInfo(StoredSize, StoredModTime,
675 StoredSignature)));
676
677 // Record the dependency.
678 unsigned DependsOnID = getModuleFileInfo(*DependsOnFile).ID;
679 getModuleFileInfo(File).Dependencies.push_back(DependsOnID);
680 }
681
682 continue;
683 }
684
685 // Handle the identifier table
686 if (State == ASTBlock && Code == IDENTIFIER_TABLE && Record[0] > 0) {
687 typedef llvm::OnDiskIterableChainedHashTable<
688 InterestingASTIdentifierLookupTrait> InterestingIdentifierTable;
689 std::unique_ptr<InterestingIdentifierTable> Table(
690 InterestingIdentifierTable::Create(
691 (const unsigned char *)Blob.data() + Record[0],
692 (const unsigned char *)Blob.data() + sizeof(uint32_t),
693 (const unsigned char *)Blob.data()));
694 for (InterestingIdentifierTable::data_iterator D = Table->data_begin(),
695 DEnd = Table->data_end();
696 D != DEnd; ++D) {
697 std::pair<StringRef, bool> Ident = *D;
698 if (Ident.second)
699 InterestingIdentifiers[Ident.first].push_back(ID);
700 else
701 (void)InterestingIdentifiers[Ident.first];
702 }
703 }
704
705 // Get Signature.
706 if (State == DiagnosticOptionsBlock && Code == SIGNATURE)
707 getModuleFileInfo(File).Signature = ASTFileSignature::create(
708 Record.begin(), Record.begin() + ASTFileSignature::size);
709
710 // We don't care about this record.
711 }
712
713 return llvm::Error::success();
714}
715
716namespace {
717
718/// Trait used to generate the identifier index as an on-disk hash
719/// table.
720class IdentifierIndexWriterTrait {
721public:
722 typedef StringRef key_type;
723 typedef StringRef key_type_ref;
724 typedef SmallVector<unsigned, 2> data_type;
725 typedef const SmallVector<unsigned, 2> &data_type_ref;
726 typedef unsigned hash_value_type;
727 typedef unsigned offset_type;
728
729 static hash_value_type ComputeHash(key_type_ref Key) {
730 return llvm::djbHash(Key);
731 }
732
733 std::pair<unsigned,unsigned>
734 EmitKeyDataLength(raw_ostream& Out, key_type_ref Key, data_type_ref Data) {
735 using namespace llvm::support;
736 endian::Writer LE(Out, little);
737 unsigned KeyLen = Key.size();
738 unsigned DataLen = Data.size() * 4;
739 LE.write<uint16_t>(KeyLen);
740 LE.write<uint16_t>(DataLen);
741 return std::make_pair(KeyLen, DataLen);
742 }
743
744 void EmitKey(raw_ostream& Out, key_type_ref Key, unsigned KeyLen) {
745 Out.write(Key.data(), KeyLen);
746 }
747
748 void EmitData(raw_ostream& Out, key_type_ref Key, data_type_ref Data,
749 unsigned DataLen) {
750 using namespace llvm::support;
751 for (unsigned I = 0, N = Data.size(); I != N; ++I)
752 endian::write<uint32_t>(Out, Data[I], little);
753 }
754};
755
756}
757
758bool GlobalModuleIndexBuilder::writeIndex(llvm::BitstreamWriter &Stream) {
759 for (auto MapEntry : ImportedModuleFiles) {
760 auto *File = MapEntry.first;
761 ImportedModuleFileInfo &Info = MapEntry.second;
762 if (getModuleFileInfo(File).Signature) {
763 if (getModuleFileInfo(File).Signature != Info.StoredSignature)
764 // Verify Signature.
765 return true;
766 } else if (Info.StoredSize != File->getSize() ||
767 Info.StoredModTime != File->getModificationTime())
768 // Verify Size and ModTime.
769 return true;
770 }
771
772 using namespace llvm;
773 llvm::TimeTraceScope TimeScope("Module WriteIndex");
774
775 // Emit the file header.
776 Stream.Emit((unsigned)'B', 8);
777 Stream.Emit((unsigned)'C', 8);
778 Stream.Emit((unsigned)'G', 8);
779 Stream.Emit((unsigned)'I', 8);
780
781 // Write the block-info block, which describes the records in this bitcode
782 // file.
783 emitBlockInfoBlock(Stream);
784
785 Stream.EnterSubblock(GLOBAL_INDEX_BLOCK_ID, 3);
786
787 // Write the metadata.
788 SmallVector<uint64_t, 2> Record;
789 Record.push_back(CurrentVersion);
790 Stream.EmitRecord(INDEX_METADATA, Record);
791
792 // Write the set of known module files.
793 for (ModuleFilesMap::iterator M = ModuleFiles.begin(),
794 MEnd = ModuleFiles.end();
795 M != MEnd; ++M) {
796 Record.clear();
797 Record.push_back(M->second.ID);
798 Record.push_back(M->first->getSize());
799 Record.push_back(M->first->getModificationTime());
800
801 // File name
802 StringRef Name(M->first->getName());
803 Record.push_back(Name.size());
804 Record.append(Name.begin(), Name.end());
805
806 // Dependencies
807 Record.push_back(M->second.Dependencies.size());
808 Record.append(M->second.Dependencies.begin(), M->second.Dependencies.end());
809 Stream.EmitRecord(MODULE, Record);
810 }
811
812 // Write the identifier -> module file mapping.
813 {
814 llvm::OnDiskChainedHashTableGenerator<IdentifierIndexWriterTrait> Generator;
815 IdentifierIndexWriterTrait Trait;
816
817 // Populate the hash table.
818 for (InterestingIdentifierMap::iterator I = InterestingIdentifiers.begin(),
819 IEnd = InterestingIdentifiers.end();
820 I != IEnd; ++I) {
821 Generator.insert(I->first(), I->second, Trait);
822 }
823
824 // Create the on-disk hash table in a buffer.
825 SmallString<4096> IdentifierTable;
826 uint32_t BucketOffset;
827 {
828 using namespace llvm::support;
829 llvm::raw_svector_ostream Out(IdentifierTable);
830 // Make sure that no bucket is at offset 0
831 endian::write<uint32_t>(Out, 0, little);
832 BucketOffset = Generator.Emit(Out, Trait);
833 }
834
835 // Create a blob abbreviation
836 auto Abbrev = std::make_shared<BitCodeAbbrev>();
837 Abbrev->Add(BitCodeAbbrevOp(IDENTIFIER_INDEX));
838 Abbrev->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Fixed, 32));
839 Abbrev->Add(BitCodeAbbrevOp(BitCodeAbbrevOp::Blob));
840 unsigned IDTableAbbrev = Stream.EmitAbbrev(std::move(Abbrev));
841
842 // Write the identifier table
843 uint64_t Record[] = {IDENTIFIER_INDEX, BucketOffset};
844 Stream.EmitRecordWithBlob(IDTableAbbrev, Record, IdentifierTable);
845 }
846
847 Stream.ExitBlock();
848 return false;
849}
850
851llvm::Error
852GlobalModuleIndex::writeIndex(FileManager &FileMgr,
853 const PCHContainerReader &PCHContainerRdr,
854 StringRef Path) {
855 llvm::SmallString<128> IndexPath;
856 IndexPath += Path;
857 llvm::sys::path::append(IndexPath, IndexFileName);
858
859 // Coordinate building the global index file with other processes that might
860 // try to do the same.
861 llvm::LockFileManager Locked(IndexPath);
862 switch (Locked) {
1
Control jumps to 'case LFS_Owned:' at line 866
→
863 case llvm::LockFileManager::LFS_Error:
864 return llvm::createStringError(std::errc::io_error, "LFS error");
865
866 case llvm::LockFileManager::LFS_Owned:
867 // We're responsible for building the index ourselves. Do so below.
868 break;
2
←
 Execution continues on line 878
→
869
870 case llvm::LockFileManager::LFS_Shared:
871 // Someone else is responsible for building the index. We don't care
872 // when they finish, so we're done.
873 return llvm::createStringError(std::errc::device_or_resource_busy,
874 "someone else is building the index");
875 }
876
877 // The module index builder.
878 GlobalModuleIndexBuilder Builder(FileMgr, PCHContainerRdr);
879
880 // Load each of the module files.
881 std::error_code EC;
882 for (llvm::sys::fs::directory_iterator D(Path, EC), DEnd;
3
←
Loop condition is true. Entering loop body
→
883 D != DEnd && !EC;
884 D.increment(EC)) {
885 // If this isn't a module file, we don't care.
886 if (llvm::sys::path::extension(D->path()) != ".pcm") {
4
←
Taking false branch
→
887 // ... unless it's a .pcm.lock file, which indicates that someone is
888 // in the process of rebuilding a module. They'll rebuild the index
889 // at the end of that translation unit, so we don't have to.
890 if (llvm::sys::path::extension(D->path()) == ".pcm.lock")
891 return llvm::createStringError(std::errc::device_or_resource_busy,
892 "someone else is building the index");
893
894 continue;
895 }
896
897 // If we can't find the module file, skip it.
898 auto ModuleFile = FileMgr.getFile(D->path());
899 if (!ModuleFile)
5
←
Taking false branch
→
900 continue;
901
902 // Load this module file.
903 if (llvm::Error Err = Builder.loadModuleFile(*ModuleFile))
6
←
Calling 'GlobalModuleIndexBuilder::loadModuleFile'
→
904 return Err;
905 }
906
907 // The output buffer, into which the global index will be written.
908 SmallString<16> OutputBuffer;
909 {
910 llvm::BitstreamWriter OutputStream(OutputBuffer);
911 if (Builder.writeIndex(OutputStream))
912 return llvm::createStringError(std::errc::io_error,
913 "failed writing index");
914 }
915
916 return llvm::writeFileAtomically((IndexPath + "-%%%%%%%%").str(), IndexPath,
917 OutputBuffer);
918}
919
920namespace {
921 class GlobalIndexIdentifierIterator : public IdentifierIterator {
922 /// The current position within the identifier lookup table.
923 IdentifierIndexTable::key_iterator Current;
924
925 /// The end position within the identifier lookup table.
926 IdentifierIndexTable::key_iterator End;
927
928 public:
929 explicit GlobalIndexIdentifierIterator(IdentifierIndexTable &Idx) {
930 Current = Idx.key_begin();
931 End = Idx.key_end();
932 }
933
934 StringRef Next() override {
935 if (Current == End)
936 return StringRef();
937
938 StringRef Result = *Current;
939 ++Current;
940 return Result;
941 }
942 };
943}
944
945IdentifierIterator *GlobalModuleIndex::createIdentifierIterator() const {
946 IdentifierIndexTable &Table =
947 *static_cast<IdentifierIndexTable *>(IdentifierIndex);
948 return new GlobalIndexIdentifierIterator(Table);
949}
←

/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/Bitstream/BitstreamReader.h

1//===- BitstreamReader.h - Low-level bitstream reader interface -*- C++ -*-===//
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 header defines the BitstreamReader class. This class can be used to
10// read an arbitrary bitstream, regardless of its contents.
11//
12//===----------------------------------------------------------------------===//
13
14#ifndef LLVM_BITSTREAM_BITSTREAMREADER_H
15#define LLVM_BITSTREAM_BITSTREAMREADER_H
16
17#include "llvm/ADT/ArrayRef.h"
18#include "llvm/ADT/SmallVector.h"
19#include "llvm/Bitstream/BitCodes.h"
20#include "llvm/Support/Endian.h"
21#include "llvm/Support/Error.h"
22#include "llvm/Support/ErrorHandling.h"
23#include "llvm/Support/MathExtras.h"
24#include "llvm/Support/MemoryBuffer.h"
25#include <algorithm>
26#include <cassert>
27#include <climits>
28#include <cstddef>
29#include <cstdint>
30#include <memory>
31#include <string>
32#include <utility>
33#include <vector>
34
35namespace llvm {
36
37/// This class maintains the abbreviations read from a block info block.
38class BitstreamBlockInfo {
39public:
40 /// This contains information emitted to BLOCKINFO_BLOCK blocks. These
41 /// describe abbreviations that all blocks of the specified ID inherit.
42 struct BlockInfo {
43 unsigned BlockID = 0;
44 std::vector<std::shared_ptr<BitCodeAbbrev>> Abbrevs;
45 std::string Name;
46 std::vector<std::pair<unsigned, std::string>> RecordNames;
47 };
48
49private:
50 std::vector<BlockInfo> BlockInfoRecords;
51
52public:
53 /// If there is block info for the specified ID, return it, otherwise return
54 /// null.
55 const BlockInfo *getBlockInfo(unsigned BlockID) const {
56 // Common case, the most recent entry matches BlockID.
57 if (!BlockInfoRecords.empty() && BlockInfoRecords.back().BlockID == BlockID)
58 return &BlockInfoRecords.back();
59
60 for (unsigned i = 0, e = static_cast<unsigned>(BlockInfoRecords.size());
61 i != e; ++i)
62 if (BlockInfoRecords[i].BlockID == BlockID)
63 return &BlockInfoRecords[i];
64 return nullptr;
65 }
66
67 BlockInfo &getOrCreateBlockInfo(unsigned BlockID) {
68 if (const BlockInfo *BI = getBlockInfo(BlockID))
69 return *const_cast<BlockInfo*>(BI);
70
71 // Otherwise, add a new record.
72 BlockInfoRecords.emplace_back();
73 BlockInfoRecords.back().BlockID = BlockID;
74 return BlockInfoRecords.back();
75 }
76};
77
78/// This represents a position within a bitstream. There may be multiple
79/// independent cursors reading within one bitstream, each maintaining their
80/// own local state.
81class SimpleBitstreamCursor {
82 ArrayRef<uint8_t> BitcodeBytes;
83 size_t NextChar = 0;
84
85public:
86 /// This is the current data we have pulled from the stream but have not
87 /// returned to the client. This is specifically and intentionally defined to
88 /// follow the word size of the host machine for efficiency. We use word_t in
89 /// places that are aware of this to make it perfectly explicit what is going
90 /// on.
91 using word_t = size_t;
92
93private:
94 word_t CurWord = 0;
95
96 /// This is the number of bits in CurWord that are valid. This is always from
97 /// [0...bits_of(size_t)-1] inclusive.
98 unsigned BitsInCurWord = 0;
99
100public:
101 static const constexpr size_t MaxChunkSize = sizeof(word_t) * 8;
102
103 SimpleBitstreamCursor() = default;
104 explicit SimpleBitstreamCursor(ArrayRef<uint8_t> BitcodeBytes)
105 : BitcodeBytes(BitcodeBytes) {}
106 explicit SimpleBitstreamCursor(StringRef BitcodeBytes)
107 : BitcodeBytes(arrayRefFromStringRef(BitcodeBytes)) {}
108 explicit SimpleBitstreamCursor(MemoryBufferRef BitcodeBytes)
109 : SimpleBitstreamCursor(BitcodeBytes.getBuffer()) {}
110
111 bool canSkipToPos(size_t pos) const {
112 // pos can be skipped to if it is a valid address or one byte past the end.
113 return pos <= BitcodeBytes.size();
114 }
115
116 bool AtEndOfStream() {
117 return BitsInCurWord == 0 && BitcodeBytes.size() <= NextChar;
118 }
119
120 /// Return the bit # of the bit we are reading.
121 uint64_t GetCurrentBitNo() const {
122 return NextChar*CHAR_BIT8 - BitsInCurWord;
123 }
124
125 // Return the byte # of the current bit.
126 uint64_t getCurrentByteNo() const { return GetCurrentBitNo() / 8; }
127
128 ArrayRef<uint8_t> getBitcodeBytes() const { return BitcodeBytes; }
129
130 /// Reset the stream to the specified bit number.
131 Error JumpToBit(uint64_t BitNo) {
132 size_t ByteNo = size_t(BitNo/8) & ~(sizeof(word_t)-1);
133 unsigned WordBitNo = unsigned(BitNo & (sizeof(word_t)*8-1));
134 assert(canSkipToPos(ByteNo) && "Invalid location")(static_cast <bool> (canSkipToPos(ByteNo) && "Invalid location"
) ? void (0) : __assert_fail ("canSkipToPos(ByteNo) && \"Invalid location\""
, "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/Bitstream/BitstreamReader.h"
, 134, __extension__ __PRETTY_FUNCTION__));
135
136 // Move the cursor to the right word.
137 NextChar = ByteNo;
138 BitsInCurWord = 0;
139
140 // Skip over any bits that are already consumed.
141 if (WordBitNo) {
142 if (Expected<word_t> Res = Read(WordBitNo))
143 return Error::success();
144 else
145 return Res.takeError();
146 }
147
148 return Error::success();
149 }
150
151 /// Get a pointer into the bitstream at the specified byte offset.
152 const uint8_t *getPointerToByte(uint64_t ByteNo, uint64_t NumBytes) {
153 return BitcodeBytes.data() + ByteNo;
154 }
155
156 /// Get a pointer into the bitstream at the specified bit offset.
157 ///
158 /// The bit offset must be on a byte boundary.
159 const uint8_t *getPointerToBit(uint64_t BitNo, uint64_t NumBytes) {
160 assert(!(BitNo % 8) && "Expected bit on byte boundary")(static_cast <bool> (!(BitNo % 8) && "Expected bit on byte boundary"
) ? void (0) : __assert_fail ("!(BitNo % 8) && \"Expected bit on byte boundary\""
, "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/Bitstream/BitstreamReader.h"
, 160, __extension__ __PRETTY_FUNCTION__));
161 return getPointerToByte(BitNo / 8, NumBytes);
162 }
163
164 Error fillCurWord() {
165 if (NextChar >= BitcodeBytes.size())
166 return createStringError(std::errc::io_error,
167 "Unexpected end of file reading %u of %u bytes",
168 NextChar, BitcodeBytes.size());
169
170 // Read the next word from the stream.
171 const uint8_t *NextCharPtr = BitcodeBytes.data() + NextChar;
172 unsigned BytesRead;
173 if (BitcodeBytes.size() >= NextChar + sizeof(word_t)) {
174 BytesRead = sizeof(word_t);
175 CurWord =
176 support::endian::read<word_t, support::little, support::unaligned>(
177 NextCharPtr);
178 } else {
179 // Short read.
180 BytesRead = BitcodeBytes.size() - NextChar;
181 CurWord = 0;
182 for (unsigned B = 0; B != BytesRead; ++B)
183 CurWord |= uint64_t(NextCharPtr[B]) << (B * 8);
184 }
185 NextChar += BytesRead;
186 BitsInCurWord = BytesRead * 8;
187 return Error::success();
188 }
189
190 Expected<word_t> Read(unsigned NumBits) {
191 static const unsigned BitsInWord = MaxChunkSize;
192
193 assert(NumBits && NumBits <= BitsInWord &&(static_cast <bool> (NumBits && NumBits <= BitsInWord
&& "Cannot return zero or more than BitsInWord bits!"
) ? void (0) : __assert_fail ("NumBits && NumBits <= BitsInWord && \"Cannot return zero or more than BitsInWord bits!\""
, "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/Bitstream/BitstreamReader.h"
, 194, __extension__ __PRETTY_FUNCTION__))
15
←
Assuming 'NumBits' is not equal to 0
→
16
←
Assuming 'NumBits' is <= 'BitsInWord'
→
17
←
'?' condition is true
→
194 "Cannot return zero or more than BitsInWord bits!")(static_cast <bool> (NumBits && NumBits <= BitsInWord
&& "Cannot return zero or more than BitsInWord bits!"
) ? void (0) : __assert_fail ("NumBits && NumBits <= BitsInWord && \"Cannot return zero or more than BitsInWord bits!\""
, "/build/llvm-toolchain-snapshot-14~++20210828111110+16086d47c0d0/llvm/include/llvm/Bitstream/BitstreamReader.h"
, 194, __extension__ __PRETTY_FUNCTION__));
195
196 static const unsigned Mask = sizeof(word_t) > 4 ? 0x3f : 0x1f;
18
←
'?' condition is true
→
197
198 // If the field is fully contained by CurWord, return it quickly.
199 if (BitsInCurWord >= NumBits) {
19
←
Assuming 'NumBits' is > field 'BitsInCurWord'
→
200 word_t R = CurWord & (~word_t(0) >> (BitsInWord - NumBits));
201
202 // Use a mask to avoid undefined behavior.
203 CurWord >>= (NumBits & Mask);
204
205 BitsInCurWord -= NumBits;
206 return R;
207 }
208
209 word_t R = BitsInCurWord
20.1
Field 'BitsInCurWord' is not equal to 0
20.1
Field 'BitsInCurWord' is not equal to 0
 ? CurWord : 0;
20
←
Taking false branch
→
21
←
'?' condition is true
→
210 unsigned BitsLeft = NumBits - BitsInCurWord;
211
212 if (Error fillResult = fillCurWord())
22
←
Assuming the condition is false
→
23
←
Taking false branch
→
213 return std::move(fillResult);
214
215 // If we run out of data, abort.
216 if (BitsLeft > BitsInCurWord)
24
←
Assuming 'BitsLeft' is <= field 'BitsInCurWord'
→
25
←
Taking false branch
→
217 return createStringError(std::errc::io_error,
218 "Unexpected end of file reading %u of %u bits",
219 BitsInCurWord, BitsLeft);
220
221 word_t R2 = CurWord & (~word_t(0) >> (BitsInWord - BitsLeft));
26
←
The result of the right shift is undefined due to shifting by '64', which is greater or equal to the width of type 'llvm::SimpleBitstreamCursor::word_t'
222
223 // Use a mask to avoid undefined behavior.
224 CurWord >>= (BitsLeft & Mask);
225
226 BitsInCurWord -= BitsLeft;
227
228 R |= R2 << (NumBits - BitsLeft);
229
230 return R;
231 }
232
233 Expected<uint32_t> ReadVBR(unsigned NumBits) {
234 Expected<unsigned> MaybeRead = Read(NumBits);
235 if (!MaybeRead)
236 return MaybeRead;
237 uint32_t Piece = MaybeRead.get();
238
239 if ((Piece & (1U << (NumBits-1))) == 0)
240 return Piece;
241
242 uint32_t Result = 0;
243 unsigned NextBit = 0;
244 while (true) {
245 Result |= (Piece & ((1U << (NumBits-1))-1)) << NextBit;
246
247 if ((Piece & (1U << (NumBits-1))) == 0)
248 return Result;
249
250 NextBit += NumBits-1;
251 MaybeRead = Read(NumBits);
252 if (!MaybeRead)
253 return MaybeRead;
254 Piece = MaybeRead.get();
255 }
256 }
257
258 // Read a VBR that may have a value up to 64-bits in size. The chunk size of
259 // the VBR must still be <= 32 bits though.
260 Expected<uint64_t> ReadVBR64(unsigned NumBits) {
261 Expected<uint64_t> MaybeRead = Read(NumBits);
262 if (!MaybeRead)
263 return MaybeRead;
264 uint32_t Piece = MaybeRead.get();
265
266 if ((Piece & (1U << (NumBits-1))) == 0)
267 return uint64_t(Piece);
268
269 uint64_t Result = 0;
270 unsigned NextBit = 0;
271 while (true) {
272 Result |= uint64_t(Piece & ((1U << (NumBits-1))-1)) << NextBit;
273
274 if ((Piece & (1U << (NumBits-1))) == 0)
275 return Result;
276
277 NextBit += NumBits-1;
278 MaybeRead = Read(NumBits);
279 if (!MaybeRead)
280 return MaybeRead;
281 Piece = MaybeRead.get();
282 }
283 }
284
285 void SkipToFourByteBoundary() {
286 // If word_t is 64-bits and if we've read less than 32 bits, just dump
287 // the bits we have up to the next 32-bit boundary.
288 if (sizeof(word_t) > 4 &&
289 BitsInCurWord >= 32) {
290 CurWord >>= BitsInCurWord-32;
291 BitsInCurWord = 32;
292 return;
293 }
294
295 BitsInCurWord = 0;
296 }
297
298 /// Return the size of the stream in bytes.
299 size_t SizeInBytes() const { return BitcodeBytes.size(); }
300
301 /// Skip to the end of the file.
302 void skipToEnd() { NextChar = BitcodeBytes.size(); }
303};
304
305/// When advancing through a bitstream cursor, each advance can discover a few
306/// different kinds of entries:
307struct BitstreamEntry {
308 enum {
309 Error, // Malformed bitcode was found.
310 EndBlock, // We've reached the end of the current block, (or the end of the
311 // file, which is treated like a series of EndBlock records.
312 SubBlock, // This is the start of a new subblock of a specific ID.
313 Record // This is a record with a specific AbbrevID.
314 } Kind;
315
316 unsigned ID;
317
318 static BitstreamEntry getError() {
319 BitstreamEntry E; E.Kind = Error; return E;
320 }
321
322 static BitstreamEntry getEndBlock() {
323 BitstreamEntry E; E.Kind = EndBlock; return E;
324 }
325
326 static BitstreamEntry getSubBlock(unsigned ID) {
327 BitstreamEntry E; E.Kind = SubBlock; E.ID = ID; return E;
328 }
329
330 static BitstreamEntry getRecord(unsigned AbbrevID) {
331 BitstreamEntry E; E.Kind = Record; E.ID = AbbrevID; return E;
332 }
333};
334
335/// This represents a position within a bitcode file, implemented on top of a
336/// SimpleBitstreamCursor.
337///
338/// Unlike iterators, BitstreamCursors are heavy-weight objects that should not
339/// be passed by value.
340class BitstreamCursor : SimpleBitstreamCursor {
341 // This is the declared size of code values used for the current block, in
342 // bits.
343 unsigned CurCodeSize = 2;
344
345 /// Abbrevs installed at in this block.
346 std::vector<std::shared_ptr<BitCodeAbbrev>> CurAbbrevs;
347
348 struct Block {
349 unsigned PrevCodeSize;
350 std::vector<std::shared_ptr<BitCodeAbbrev>> PrevAbbrevs;
351
352 explicit Block(unsigned PCS) : PrevCodeSize(PCS) {}
353 };
354
355 /// This tracks the codesize of parent blocks.
356 SmallVector<Block, 8> BlockScope;
357
358 BitstreamBlockInfo *BlockInfo = nullptr;
359
360public:
361 static const size_t MaxChunkSize = sizeof(word_t) * 8;
362
363 BitstreamCursor() = default;
364 explicit BitstreamCursor(ArrayRef<uint8_t> BitcodeBytes)
365 : SimpleBitstreamCursor(BitcodeBytes) {}
366 explicit BitstreamCursor(StringRef BitcodeBytes)
367 : SimpleBitstreamCursor(BitcodeBytes) {}
368 explicit BitstreamCursor(MemoryBufferRef BitcodeBytes)
369 : SimpleBitstreamCursor(BitcodeBytes) {}
370
371 using SimpleBitstreamCursor::AtEndOfStream;
372 using SimpleBitstreamCursor::canSkipToPos;
373 using SimpleBitstreamCursor::fillCurWord;
374 using SimpleBitstreamCursor::getBitcodeBytes;
375 using SimpleBitstreamCursor::GetCurrentBitNo;
376 using SimpleBitstreamCursor::getCurrentByteNo;
377 using SimpleBitstreamCursor::getPointerToByte;
378 using SimpleBitstreamCursor::JumpToBit;
379 using SimpleBitstreamCursor::Read;
380 using SimpleBitstreamCursor::ReadVBR;
381 using SimpleBitstreamCursor::ReadVBR64;
382 using SimpleBitstreamCursor::SizeInBytes;
383 using SimpleBitstreamCursor::skipToEnd;
384
385 /// Return the number of bits used to encode an abbrev #.
386 unsigned getAbbrevIDWidth() const { return CurCodeSize; }
387
388 /// Flags that modify the behavior of advance().
389 enum {
390 /// If this flag is used, the advance() method does not automatically pop
391 /// the block scope when the end of a block is reached.
392 AF_DontPopBlockAtEnd = 1,
393
394 /// If this flag is used, abbrev entries are returned just like normal
395 /// records.
396 AF_DontAutoprocessAbbrevs = 2
397 };
398
399 /// Advance the current bitstream, returning the next entry in the stream.
400 Expected<BitstreamEntry> advance(unsigned Flags = 0) {
401 while (true) {
11
←
Loop condition is true. Entering loop body
→
402 if (AtEndOfStream())
12
←
Taking false branch
→
403 return BitstreamEntry::getError();
404
405 Expected<unsigned> MaybeCode = ReadCode();
13
←
Calling 'BitstreamCursor::ReadCode'
→
406 if (!MaybeCode)
407 return MaybeCode.takeError();
408 unsigned Code = MaybeCode.get();
409
410 if (Code == bitc::END_BLOCK) {
411 // Pop the end of the block unless Flags tells us not to.
412 if (!(Flags & AF_DontPopBlockAtEnd) && ReadBlockEnd())
413 return BitstreamEntry::getError();
414 return BitstreamEntry::getEndBlock();
415 }
416
417 if (Code == bitc::ENTER_SUBBLOCK) {
418 if (Expected<unsigned> MaybeSubBlock = ReadSubBlockID())
419 return BitstreamEntry::getSubBlock(MaybeSubBlock.get());
420 else
421 return MaybeSubBlock.takeError();
422 }
423
424 if (Code == bitc::DEFINE_ABBREV &&
425 !(Flags & AF_DontAutoprocessAbbrevs)) {
426 // We read and accumulate abbrev's, the client can't do anything with
427 // them anyway.
428 if (Error Err = ReadAbbrevRecord())
429 return std::move(Err);
430 continue;
431 }
432
433 return BitstreamEntry::getRecord(Code);
434 }
435 }
436
437 /// This is a convenience function for clients that don't expect any
438 /// subblocks. This just skips over them automatically.
439 Expected<BitstreamEntry> advanceSkippingSubblocks(unsigned Flags = 0) {
440 while (true) {
441 // If we found a normal entry, return it.
442 Expected<BitstreamEntry> MaybeEntry = advance(Flags);
443 if (!MaybeEntry)
444 return MaybeEntry;
445 BitstreamEntry Entry = MaybeEntry.get();
446
447 if (Entry.Kind != BitstreamEntry::SubBlock)
448 return Entry;
449
450 // If we found a sub-block, just skip over it and check the next entry.
451 if (Error Err = SkipBlock())
452 return std::move(Err);
453 }
454 }
455
456 Expected<unsigned> ReadCode() { return Read(CurCodeSize); }
14
←
Calling 'SimpleBitstreamCursor::Read'
→
457
458 // Block header:
459 // [ENTER_SUBBLOCK, blockid, newcodelen, <align4bytes>, blocklen]
460
461 /// Having read the ENTER_SUBBLOCK code, read the BlockID for the block.
462 Expected<unsigned> ReadSubBlockID() { return ReadVBR(bitc::BlockIDWidth); }
463
464 /// Having read the ENTER_SUBBLOCK abbrevid and a BlockID, skip over the body
465 /// of this block.
466 Error SkipBlock() {
467 // Read and ignore the codelen value.
468 if (Expected<uint32_t> Res = ReadVBR(bitc::CodeLenWidth))
469 ; // Since we are skipping this block, we don't care what code widths are
470 // used inside of it.
471 else
472 return Res.takeError();
473
474 SkipToFourByteBoundary();
475 Expected<unsigned> MaybeNum = Read(bitc::BlockSizeWidth);
476 if (!MaybeNum)
477 return MaybeNum.takeError();
478 size_t NumFourBytes = MaybeNum.get();
479
480 // Check that the block wasn't partially defined, and that the offset isn't
481 // bogus.
482 size_t SkipTo = GetCurrentBitNo() + NumFourBytes * 4 * 8;
483 if (AtEndOfStream())
484 return createStringError(std::errc::illegal_byte_sequence,
485 "can't skip block: already at end of stream");
486 if (!canSkipToPos(SkipTo / 8))
487 return createStringError(std::errc::illegal_byte_sequence,
488 "can't skip to bit %zu from %" PRIu64"l" "u", SkipTo,
489 GetCurrentBitNo());
490
491 if (Error Res = JumpToBit(SkipTo))
492 return Res;
493
494 return Error::success();
495 }
496
497 /// Having read the ENTER_SUBBLOCK abbrevid, and enter the block.
498 Error EnterSubBlock(unsigned BlockID, unsigned *NumWordsP = nullptr);
499
500 bool ReadBlockEnd() {
501 if (BlockScope.empty()) return true;
502
503 // Block tail:
504 // [END_BLOCK, <align4bytes>]
505 SkipToFourByteBoundary();
506
507 popBlockScope();
508 return false;
509 }
510
511private:
512 void popBlockScope() {
513 CurCodeSize = BlockScope.back().PrevCodeSize;
514
515 CurAbbrevs = std::move(BlockScope.back().PrevAbbrevs);
516 BlockScope.pop_back();
517 }
518
519 //===--------------------------------------------------------------------===//
520 // Record Processing
521 //===--------------------------------------------------------------------===//
522
523public:
524 /// Return the abbreviation for the specified AbbrevId.
525 const BitCodeAbbrev *getAbbrev(unsigned AbbrevID) {
526 unsigned AbbrevNo = AbbrevID - bitc::FIRST_APPLICATION_ABBREV;
527 if (AbbrevNo >= CurAbbrevs.size())
528 report_fatal_error("Invalid abbrev number");
529 return CurAbbrevs[AbbrevNo].get();
530 }
531
532 /// Read the current record and discard it, returning the code for the record.
533 Expected<unsigned> skipRecord(unsigned AbbrevID);
534
535 Expected<unsigned> readRecord(unsigned AbbrevID,
536 SmallVectorImpl<uint64_t> &Vals,
537 StringRef *Blob = nullptr);
538
539 //===--------------------------------------------------------------------===//
540 // Abbrev Processing
541 //===--------------------------------------------------------------------===//
542 Error ReadAbbrevRecord();
543
544 /// Read and return a block info block from the bitstream. If an error was
545 /// encountered, return None.
546 ///
547 /// \param ReadBlockInfoNames Whether to read block/record name information in
548 /// the BlockInfo block. Only llvm-bcanalyzer uses this.
549 Expected<Optional<BitstreamBlockInfo>>
550 ReadBlockInfoBlock(bool ReadBlockInfoNames = false);
551
552 /// Set the block info to be used by this BitstreamCursor to interpret
553 /// abbreviated records.
554 void setBlockInfo(BitstreamBlockInfo *BI) { BlockInfo = BI; }
555};
556
557} // end llvm namespace
558
559#endif // LLVM_BITSTREAM_BITSTREAMREADER_H