diff --git a/bolt/include/bolt/Profile/YAMLProfileReader.h b/bolt/include/bolt/Profile/YAMLProfileReader.h
@@ -43,6 +43,59 @@ class YAMLProfileReader : public ProfileReaderBase {
   using ProfileLookupMap =
       DenseMap<uint32_t, yaml::bolt::BinaryFunctionProfile *>;
 
+  /// A class for matching binary functions in functions in the YAML profile.
+  /// First, a call graph is constructed for both profiled and binary functions.
+  /// Then functions are hashed based on the names of their callee/caller
+  /// functions. Finally, functions are matched based on these neighbor hashes.
+  class CallGraphMatcher {
+  public:
+    /// Constructs the call graphs for binary and profiled functions and
+    /// computes neighbor hashes for binary functions.
+    CallGraphMatcher(BinaryContext &BC, yaml::bolt::BinaryProfile &YamlBP,
+                     ProfileLookupMap &IdToYAMLBF);
+
+    /// Returns the YamlBFs adjacent to the parameter YamlBF in the call graph.
+    std::optional<std::set<yaml::bolt::BinaryFunctionProfile *>>
+    getAdjacentYamlBFs(yaml::bolt::BinaryFunctionProfile &YamlBF) {
+      auto It = YamlBFAdjacencyMap.find(&YamlBF);
+      return It == YamlBFAdjacencyMap.end() ? std::nullopt
+                                            : std::make_optional(It->second);
+    }
+
+    /// Returns the binary functions with the parameter neighbor hash.
+    std::optional<std::vector<BinaryFunction *>>
+    getBFsWithNeighborHash(uint64_t NeighborHash) {
+      auto It = NeighborHashToBFs.find(NeighborHash);
+      return It == NeighborHashToBFs.end() ? std::nullopt
+                                           : std::make_optional(It->second);
+    }
+
+  private:
+    /// Adds edges to the binary function call graph given the callsites of the
+    /// parameter function.
+    void constructBFCG(BinaryContext &BC, yaml::bolt::BinaryProfile &YamlBP);
+
+    /// Using the constructed binary function call graph, computes and creates
+    /// mappings from "neighbor hash" (composed of the function names of callee
+    /// and caller functions of a function) to binary functions.
+    void computeBFNeighborHashes(BinaryContext &BC);
+
+    /// Constructs the call graph for profile functions.
+    void constructYAMLFCG(yaml::bolt::BinaryProfile &YamlBP,
+                          ProfileLookupMap &IdToYAMLBF);
+
+    /// Adjacency map for binary functions in the call graph.
+    DenseMap<BinaryFunction *, std::set<BinaryFunction *>> BFAdjacencyMap;
+
+    /// Maps neighbor hashes to binary functions.
+    DenseMap<uint64_t, std::vector<BinaryFunction *>> NeighborHashToBFs;
+
+    /// Adjacency map for profile functions in the call graph.
+    DenseMap<yaml::bolt::BinaryFunctionProfile *,
+             std::set<yaml::bolt::BinaryFunctionProfile *>>
+        YamlBFAdjacencyMap;
+  };
+
 private:
   /// Adjustments for basic samples profiles (without LBR).
   bool NormalizeByInsnCount{false};
@@ -100,6 +153,9 @@ class YAMLProfileReader : public ProfileReaderBase {
   /// Matches functions using exact hash.
   size_t matchWithHash(BinaryContext &BC);
 
+  /// Matches functions using the call graph.
+  size_t matchWithCallGraph(BinaryContext &BC);
+
   /// Matches functions with similarly named profiled functions.
   size_t matchWithNameSimilarity(BinaryContext &BC);
 

diff --git a/bolt/include/bolt/RuntimeLibs/RuntimeLibrary.h b/bolt/include/bolt/RuntimeLibs/RuntimeLibrary.h
@@ -58,7 +58,16 @@ class RuntimeLibrary {
   uint64_t RuntimeFiniAddress{0};
   uint64_t RuntimeStartAddress{0};
 
-  /// Get the full path to a runtime library specified by \p LibFileName.
+  /// Get the full path to a runtime library specified by \p LibFileName and \p
+  /// ToolPath.
+  static std::string getLibPathByToolPath(StringRef ToolPath,
+                                          StringRef LibFileName);
+
+  /// Get the full path to a runtime library by the install directory.
+  static std::string getLibPathByInstalled(StringRef LibFileName);
+
+  /// Gets the full path to a runtime library based on whether it exists
+  /// in the install libdir or runtime libdir.
   static std::string getLibPath(StringRef ToolPath, StringRef LibFileName);
 
   /// Load a static runtime library specified by \p LibPath.

diff --git a/bolt/lib/CMakeLists.txt b/bolt/lib/CMakeLists.txt
@@ -1,3 +1,5 @@
+add_compile_definitions(CMAKE_INSTALL_FULL_LIBDIR="${CMAKE_INSTALL_FULL_LIBDIR}")
+
 add_subdirectory(Core)
 add_subdirectory(Passes)
 add_subdirectory(Profile)

diff --git a/bolt/lib/Core/BinaryContext.cpp b/bolt/lib/Core/BinaryContext.cpp
@@ -71,7 +71,7 @@ PrintMemData("print-mem-data",
 
 cl::opt<std::string> CompDirOverride(
     "comp-dir-override",
-    cl::desc("overrides DW_AT_comp_dir, and provides an alterantive base "
+    cl::desc("overrides DW_AT_comp_dir, and provides an alternative base "
              "location, which is used with DW_AT_dwo_name to construct a path "
              "to *.dwo files."),
     cl::Hidden, cl::init(""), cl::cat(BoltCategory));
@@ -646,7 +646,7 @@ bool BinaryContext::analyzeJumpTable(const uint64_t Address,
     const BinaryFunction *TargetBF = getBinaryFunctionContainingAddress(Value);
     const bool DoesBelongToFunction =
         BF.containsAddress(Value) ||
-        (TargetBF && TargetBF->isParentOrChildOf(BF));
+        (TargetBF && areRelatedFragments(TargetBF, &BF));
     if (!DoesBelongToFunction) {
       LLVM_DEBUG({
         if (!BF.containsAddress(Value)) {
@@ -839,9 +839,11 @@ BinaryContext::getOrCreateJumpTable(BinaryFunction &Function, uint64_t Address,
     assert(Address == JT->getAddress() && "unexpected non-empty jump table");
 
     // Prevent associating a jump table to a specific fragment twice.
-    // This simple check arises from the assumption: no more than 2 fragments.
-    if (JT->Parents.size() == 1 && JT->Parents[0] != &Function) {
-      assert(JT->Parents[0]->isParentOrChildOf(Function) &&
+    if (!llvm::is_contained(JT->Parents, &Function)) {
+      assert(llvm::all_of(JT->Parents,
+                          [&](const BinaryFunction *BF) {
+                            return areRelatedFragments(&Function, BF);
+                          }) &&
              "cannot re-use jump table of a different function");
       // Duplicate the entry for the parent function for easy access
       JT->Parents.push_back(&Function);
@@ -852,8 +854,8 @@ BinaryContext::getOrCreateJumpTable(BinaryFunction &Function, uint64_t Address,
         JT->print(this->outs());
       }
       Function.JumpTables.emplace(Address, JT);
-      JT->Parents[0]->setHasIndirectTargetToSplitFragment(true);
-      JT->Parents[1]->setHasIndirectTargetToSplitFragment(true);
+      for (BinaryFunction *Parent : JT->Parents)
+        Parent->setHasIndirectTargetToSplitFragment(true);
     }
 
     bool IsJumpTableParent = false;
@@ -1209,12 +1211,13 @@ void BinaryContext::generateSymbolHashes() {
 }
 
 bool BinaryContext::registerFragment(BinaryFunction &TargetFunction,
-                                     BinaryFunction &Function) const {
+                                     BinaryFunction &Function) {
   assert(TargetFunction.isFragment() && "TargetFunction must be a fragment");
   if (TargetFunction.isChildOf(Function))
     return true;
   TargetFunction.addParentFragment(Function);
   Function.addFragment(TargetFunction);
+  FragmentClasses.unionSets(&TargetFunction, &Function);
   if (!HasRelocations) {
     TargetFunction.setSimple(false);
     Function.setSimple(false);
@@ -1336,7 +1339,7 @@ void BinaryContext::processInterproceduralReferences() {
 
     if (TargetFunction) {
       if (TargetFunction->isFragment() &&
-          !TargetFunction->isChildOf(Function)) {
+          !areRelatedFragments(TargetFunction, &Function)) {
         this->errs()
             << "BOLT-WARNING: interprocedural reference between unrelated "
                "fragments: "
@@ -2367,10 +2370,7 @@ BinaryContext::calculateEmittedSize(BinaryFunction &BF, bool FixBranches) {
   std::unique_ptr<MCObjectWriter> OW = MAB->createObjectWriter(VecOS);
   std::unique_ptr<MCStreamer> Streamer(TheTarget->createMCObjectStreamer(
       *TheTriple, *LocalCtx, std::unique_ptr<MCAsmBackend>(MAB), std::move(OW),
-      std::unique_ptr<MCCodeEmitter>(MCEInstance.MCE.release()), *STI,
-      /*RelaxAll=*/false,
-      /*IncrementalLinkerCompatible=*/false,
-      /*DWARFMustBeAtTheEnd=*/false));
+      std::unique_ptr<MCCodeEmitter>(MCEInstance.MCE.release()), *STI));
 
   Streamer->initSections(false, *STI);
 

diff --git a/bolt/lib/Core/Exceptions.cpp b/bolt/lib/Core/Exceptions.cpp
@@ -207,7 +207,7 @@ Error BinaryFunction::parseLSDA(ArrayRef<uint8_t> LSDASectionData,
                "BOLT-ERROR: cannot find landing pad fragment");
         BC.addInterproceduralReference(this, Fragment->getAddress());
         BC.processInterproceduralReferences();
-        assert(isParentOrChildOf(*Fragment) &&
+        assert(BC.areRelatedFragments(this, Fragment) &&
                "BOLT-ERROR: cannot have landing pads in different functions");
         setHasIndirectTargetToSplitFragment(true);
         BC.addFragmentsToSkip(this);

diff --git a/bolt/lib/Passes/AsmDump.cpp b/bolt/lib/Passes/AsmDump.cpp
@@ -139,11 +139,8 @@ void dumpFunction(const BinaryFunction &BF) {
   auto FOut = std::make_unique<formatted_raw_ostream>(OS);
   FOut->SetUnbuffered();
   std::unique_ptr<MCStreamer> AsmStreamer(
-      createAsmStreamer(*LocalCtx, std::move(FOut),
-                        /*isVerboseAsm=*/true,
-                        /*useDwarfDirectory=*/false, InstructionPrinter,
-                        std::move(MCEInstance.MCE), std::move(MAB),
-                        /*ShowInst=*/false));
+      createAsmStreamer(*LocalCtx, std::move(FOut), InstructionPrinter,
+                        std::move(MCEInstance.MCE), std::move(MAB)));
   AsmStreamer->initSections(true, *BC.STI);
   std::unique_ptr<TargetMachine> TM(BC.TheTarget->createTargetMachine(
       BC.TripleName, "", "", TargetOptions(), std::nullopt));

diff --git a/bolt/lib/Profile/DataAggregator.cpp b/bolt/lib/Profile/DataAggregator.cpp
@@ -88,6 +88,7 @@ MaxSamples("max-samples",
   cl::cat(AggregatorCategory));
 
 extern cl::opt<opts::ProfileFormatKind> ProfileFormat;
+extern cl::opt<bool> ProfileUsePseudoProbes;
 extern cl::opt<std::string> SaveProfile;
 
 cl::opt<bool> ReadPreAggregated(
@@ -2298,7 +2299,8 @@ std::error_code DataAggregator::writeBATYAML(BinaryContext &BC,
 
   yaml::bolt::BinaryProfile BP;
 
-  const MCPseudoProbeDecoder *PseudoProbeDecoder = BC.getPseudoProbeDecoder();
+  const MCPseudoProbeDecoder *PseudoProbeDecoder =
+      opts::ProfileUsePseudoProbes ? BC.getPseudoProbeDecoder() : nullptr;
 
   // Fill out the header info.
   BP.Header.Version = 1;

diff --git a/bolt/lib/Profile/YAMLProfileReader.cpp b/bolt/lib/Profile/YAMLProfileReader.cpp
@@ -41,15 +41,87 @@ llvm::cl::opt<bool>
     MatchProfileWithFunctionHash("match-profile-with-function-hash",
                                  cl::desc("Match profile with function hash"),
                                  cl::Hidden, cl::cat(BoltOptCategory));
+llvm::cl::opt<bool>
+    MatchWithCallGraph("match-with-call-graph",
+                       cl::desc("Match functions with call graph"), cl::Hidden,
+                       cl::cat(BoltOptCategory));
 
 llvm::cl::opt<bool> ProfileUseDFS("profile-use-dfs",
                                   cl::desc("use DFS order for YAML profile"),
                                   cl::Hidden, cl::cat(BoltOptCategory));
+
+llvm::cl::opt<bool> ProfileUsePseudoProbes(
+    "profile-use-pseudo-probes",
+    cl::desc("Use pseudo probes for profile generation and matching"),
+    cl::Hidden, cl::cat(BoltOptCategory));
 } // namespace opts
 
 namespace llvm {
 namespace bolt {
 
+YAMLProfileReader::CallGraphMatcher::CallGraphMatcher(
+    BinaryContext &BC, yaml::bolt::BinaryProfile &YamlBP,
+    ProfileLookupMap &IdToYAMLBF) {
+  constructBFCG(BC, YamlBP);
+  constructYAMLFCG(YamlBP, IdToYAMLBF);
+  computeBFNeighborHashes(BC);
+}
+
+void YAMLProfileReader::CallGraphMatcher::constructBFCG(
+    BinaryContext &BC, yaml::bolt::BinaryProfile &YamlBP) {
+  for (BinaryFunction *BF : BC.getAllBinaryFunctions()) {
+    for (const BinaryBasicBlock &BB : BF->blocks()) {
+      for (const MCInst &Instr : BB) {
+        if (!BC.MIB->isCall(Instr))
+          continue;
+        const MCSymbol *CallSymbol = BC.MIB->getTargetSymbol(Instr);
+        if (!CallSymbol)
+          continue;
+        BinaryData *BD = BC.getBinaryDataByName(CallSymbol->getName());
+        if (!BD)
+          continue;
+        BinaryFunction *CalleeBF = BC.getFunctionForSymbol(BD->getSymbol());
+        if (!CalleeBF)
+          continue;
+
+        BFAdjacencyMap[CalleeBF].insert(BF);
+        BFAdjacencyMap[BF].insert(CalleeBF);
+      }
+    }
+  }
+}
+
+void YAMLProfileReader::CallGraphMatcher::computeBFNeighborHashes(
+    BinaryContext &BC) {
+  for (BinaryFunction *BF : BC.getAllBinaryFunctions()) {
+    auto It = BFAdjacencyMap.find(BF);
+    if (It == BFAdjacencyMap.end())
+      continue;
+    auto &AdjacentBFs = It->second;
+    std::string HashStr;
+    for (BinaryFunction *BF : AdjacentBFs)
+      HashStr += BF->getOneName();
+    uint64_t Hash = std::hash<std::string>{}(HashStr);
+    NeighborHashToBFs[Hash].push_back(BF);
+  }
+}
+
+void YAMLProfileReader::CallGraphMatcher::constructYAMLFCG(
+    yaml::bolt::BinaryProfile &YamlBP, ProfileLookupMap &IdToYAMLBF) {
+
+  for (auto &CallerYamlBF : YamlBP.Functions) {
+    for (auto &YamlBB : CallerYamlBF.Blocks) {
+      for (auto &CallSite : YamlBB.CallSites) {
+        auto IdToYAMLBFIt = IdToYAMLBF.find(CallSite.DestId);
+        if (IdToYAMLBFIt == IdToYAMLBF.end())
+          continue;
+        YamlBFAdjacencyMap[&CallerYamlBF].insert(IdToYAMLBFIt->second);
+        YamlBFAdjacencyMap[IdToYAMLBFIt->second].insert(&CallerYamlBF);
+      }
+    }
+  }
+}
+
 bool YAMLProfileReader::isYAML(const StringRef Filename) {
   if (auto MB = MemoryBuffer::getFileOrSTDIN(Filename)) {
     StringRef Buffer = (*MB)->getBuffer();
@@ -350,7 +422,7 @@ bool YAMLProfileReader::profileMatches(
 }
 
 bool YAMLProfileReader::mayHaveProfileData(const BinaryFunction &BF) {
-  if (opts::MatchProfileWithFunctionHash)
+  if (opts::MatchProfileWithFunctionHash || opts::MatchWithCallGraph)
     return true;
   for (StringRef Name : BF.getNames())
     if (ProfileFunctionNames.contains(Name))
@@ -446,6 +518,79 @@ size_t YAMLProfileReader::matchWithLTOCommonName() {
   return MatchedWithLTOCommonName;
 }
 
+size_t YAMLProfileReader::matchWithCallGraph(BinaryContext &BC) {
+  if (!opts::MatchWithCallGraph)
+    return 0;
+
+  size_t MatchedWithCallGraph = 0;
+  CallGraphMatcher CGMatcher(BC, YamlBP, IdToYamLBF);
+
+  ItaniumPartialDemangler Demangler;
+  auto GetBaseName = [&](std::string &FunctionName) {
+    if (Demangler.partialDemangle(FunctionName.c_str()))
+      return std::string("");
+    size_t BufferSize = 1;
+    char *Buffer = static_cast<char *>(std::malloc(BufferSize));
+    char *BaseName = Demangler.getFunctionBaseName(Buffer, &BufferSize);
+    if (!BaseName) {
+      std::free(Buffer);
+      return std::string("");
+    }
+    if (Buffer != BaseName)
+      Buffer = BaseName;
+    std::string BaseNameStr(Buffer, BufferSize);
+    std::free(Buffer);
+    return BaseNameStr;
+  };
+
+  // Matches YAMLBF to BFs with neighbor hashes.
+  for (yaml::bolt::BinaryFunctionProfile &YamlBF : YamlBP.Functions) {
+    if (YamlBF.Used)
+      continue;
+    auto AdjacentYamlBFsOpt = CGMatcher.getAdjacentYamlBFs(YamlBF);
+    if (!AdjacentYamlBFsOpt)
+      continue;
+    std::set<yaml::bolt::BinaryFunctionProfile *> AdjacentYamlBFs =
+        AdjacentYamlBFsOpt.value();
+    std::string AdjacentYamlBFsHashStr;
+    for (auto *AdjacentYamlBF : AdjacentYamlBFs)
+      AdjacentYamlBFsHashStr += AdjacentYamlBF->Name;
+    uint64_t Hash = std::hash<std::string>{}(AdjacentYamlBFsHashStr);
+    auto BFsWithSameHashOpt = CGMatcher.getBFsWithNeighborHash(Hash);
+    if (!BFsWithSameHashOpt)
+      continue;
+    std::vector<BinaryFunction *> BFsWithSameHash = BFsWithSameHashOpt.value();
+    // Finds the binary function with the longest common prefix to the profiled
+    // function and matches.
+    BinaryFunction *ClosestBF = nullptr;
+    size_t LCP = 0;
+    std::string YamlBFBaseName = GetBaseName(YamlBF.Name);
+    for (BinaryFunction *BF : BFsWithSameHash) {
+      if (ProfiledFunctions.count(BF))
+        continue;
+      std::string BFName = std::string(BF->getOneName());
+      std::string BFBaseName = GetBaseName(BFName);
+      size_t PrefixLength = 0;
+      size_t N = std::min(YamlBFBaseName.size(), BFBaseName.size());
+      for (size_t I = 0; I < N; ++I) {
+        if (YamlBFBaseName[I] != BFBaseName[I])
+          break;
+        ++PrefixLength;
+      }
+      if (PrefixLength >= LCP) {
+        LCP = PrefixLength;
+        ClosestBF = BF;
+      }
+    }
+    if (ClosestBF) {
+      matchProfileToFunction(YamlBF, *ClosestBF);
+      ++MatchedWithCallGraph;
+    }
+  }
+
+  return MatchedWithCallGraph;
+}
+
 size_t YAMLProfileReader::matchWithNameSimilarity(BinaryContext &BC) {
   if (opts::NameSimilarityFunctionMatchingThreshold == 0)
     return 0;
@@ -581,9 +726,14 @@ Error YAMLProfileReader::readProfile(BinaryContext &BC) {
     }
   }
 
+  // Map profiled function ids to names.
+  for (yaml::bolt::BinaryFunctionProfile &YamlBF : YamlBP.Functions)
+    IdToYamLBF[YamlBF.Id] = &YamlBF;
+
   const size_t MatchedWithExactName = matchWithExactName();
   const size_t MatchedWithHash = matchWithHash(BC);
   const size_t MatchedWithLTOCommonName = matchWithLTOCommonName();
+  const size_t MatchedWithCallGraph = matchWithCallGraph(BC);
   const size_t MatchedWithNameSimilarity = matchWithNameSimilarity(BC);
 
   for (auto [YamlBF, BF] : llvm::zip_equal(YamlBP.Functions, ProfileBFs))
@@ -603,18 +753,15 @@ Error YAMLProfileReader::readProfile(BinaryContext &BC) {
            << " functions with hash\n";
     outs() << "BOLT-INFO: matched " << MatchedWithLTOCommonName
            << " functions with matching LTO common names\n";
+    outs() << "BOLT-INFO: matched " << MatchedWithCallGraph
+           << " functions with call graph\n";
     outs() << "BOLT-INFO: matched " << MatchedWithNameSimilarity
            << " functions with similar names\n";
   }
 
   // Set for parseFunctionProfile().
   NormalizeByInsnCount = usesEvent("cycles") || usesEvent("instructions");
   NormalizeByCalls = usesEvent("branches");
-
-  // Map profiled function ids to names.
-  for (yaml::bolt::BinaryFunctionProfile &YamlBF : YamlBP.Functions)
-    IdToYamLBF[YamlBF.Id] = &YamlBF;
-
   uint64_t NumUnused = 0;
   for (yaml::bolt::BinaryFunctionProfile &YamlBF : YamlBP.Functions) {
     if (YamlBF.Id >= YamlProfileToFunction.size()) {
@@ -630,7 +777,8 @@ Error YAMLProfileReader::readProfile(BinaryContext &BC) {
 
   BC.setNumUnusedProfiledObjects(NumUnused);
 
-  if (opts::Lite && opts::MatchProfileWithFunctionHash) {
+  if (opts::Lite &&
+      (opts::MatchProfileWithFunctionHash || opts::MatchWithCallGraph)) {
     for (BinaryFunction *BF : BC.getAllBinaryFunctions())
       if (!BF->hasProfile())
         BF->setIgnored();

diff --git a/bolt/lib/Profile/YAMLProfileWriter.cpp b/bolt/lib/Profile/YAMLProfileWriter.cpp
@@ -22,6 +22,7 @@
 
 namespace opts {
 extern llvm::cl::opt<bool> ProfileUseDFS;
+extern llvm::cl::opt<bool> ProfileUsePseudoProbes;
 } // namespace opts
 
 namespace llvm {
@@ -57,7 +58,8 @@ YAMLProfileWriter::convert(const BinaryFunction &BF, bool UseDFS,
                            const BoltAddressTranslation *BAT) {
   yaml::bolt::BinaryFunctionProfile YamlBF;
   const BinaryContext &BC = BF.getBinaryContext();
-  const MCPseudoProbeDecoder *PseudoProbeDecoder = BC.getPseudoProbeDecoder();
+  const MCPseudoProbeDecoder *PseudoProbeDecoder =
+      opts::ProfileUsePseudoProbes ? BC.getPseudoProbeDecoder() : nullptr;
 
   const uint16_t LBRProfile = BF.getProfileFlags() & BinaryFunction::PF_LBR;
 

diff --git a/bolt/lib/Rewrite/BinaryPassManager.cpp b/bolt/lib/Rewrite/BinaryPassManager.cpp
@@ -263,6 +263,10 @@ static cl::opt<bool> CMOVConversionFlag("cmov-conversion",
                                         cl::ReallyHidden,
                                         cl::cat(BoltOptCategory));
 
+static cl::opt<bool> ShortenInstructions("shorten-instructions",
+                                         cl::desc("shorten instructions"),
+                                         cl::init(true),
+                                         cl::cat(BoltOptCategory));
 } // namespace opts
 
 namespace llvm {
@@ -378,7 +382,8 @@ Error BinaryFunctionPassManager::runAllPasses(BinaryContext &BC) {
   else if (opts::Hugify)
     Manager.registerPass(std::make_unique<HugePage>(NeverPrint));
 
-  Manager.registerPass(std::make_unique<ShortenInstructions>(NeverPrint));
+  Manager.registerPass(std::make_unique<ShortenInstructions>(NeverPrint),
+                       opts::ShortenInstructions);
 
   Manager.registerPass(std::make_unique<RemoveNops>(NeverPrint),
                        !opts::KeepNops);

diff --git a/bolt/lib/Rewrite/DWARFRewriter.cpp b/bolt/lib/Rewrite/DWARFRewriter.cpp
@@ -326,12 +326,6 @@ static cl::opt<bool> KeepARanges(
         "keep or generate .debug_aranges section if .gdb_index is written"),
     cl::Hidden, cl::cat(BoltCategory));
 
-static cl::opt<bool> DeterministicDebugInfo(
-    "deterministic-debuginfo",
-    cl::desc("disables parallel execution of tasks that may produce "
-             "nondeterministic debug info"),
-    cl::init(true), cl::cat(BoltCategory));
-
 static cl::opt<std::string> DwarfOutputPath(
     "dwarf-output-path",
     cl::desc("Path to where .dwo files or dwp file will be written out to."),
@@ -607,11 +601,6 @@ void DWARFRewriter::updateDebugInfo() {
   StrWriter = std::make_unique<DebugStrWriter>(*BC.DwCtx, false);
   StrOffstsWriter = std::make_unique<DebugStrOffsetsWriter>(BC);
 
-  if (!opts::DeterministicDebugInfo) {
-    opts::DeterministicDebugInfo = true;
-    errs() << "BOLT-WARNING: --deterministic-debuginfo is being deprecated\n";
-  }
-
   /// Stores and serializes information that will be put into the
   /// .debug_addr DWARF section.
   std::unique_ptr<DebugAddrWriter> FinalAddrWriter;
@@ -631,8 +620,8 @@ void DWARFRewriter::updateDebugInfo() {
   uint32_t CUIndex = 0;
   std::mutex AccessMutex;
   // Needs to be invoked in the same order as CUs are processed.
-  auto createRangeLocListAddressWriters =
-      [&](DWARFUnit &CU) -> DebugLocWriter * {
+  llvm::DenseMap<uint64_t, uint64_t> LocListWritersIndexByCU;
+  auto createRangeLocListAddressWriters = [&](DWARFUnit &CU) {
     std::lock_guard<std::mutex> Lock(AccessMutex);
     const uint16_t DwarfVersion = CU.getVersion();
     if (DwarfVersion >= 5) {
@@ -652,7 +641,6 @@ void DWARFRewriter::updateDebugInfo() {
         RangeListsWritersByCU[*DWOId] = std::move(DWORangeListsSectionWriter);
       }
       AddressWritersByCU[CU.getOffset()] = std::move(AddrW);
-
     } else {
       auto AddrW =
           std::make_unique<DebugAddrWriter>(&BC, CU.getAddressByteSize());
@@ -668,7 +656,7 @@ void DWARFRewriter::updateDebugInfo() {
             std::move(LegacyRangesSectionWriterByCU);
       }
     }
-    return LocListWritersByCU[CUIndex++].get();
+    LocListWritersIndexByCU[CU.getOffset()] = CUIndex++;
   };
 
   DWARF5AcceleratorTable DebugNamesTable(opts::CreateDebugNames, BC,
@@ -677,74 +665,62 @@ void DWARFRewriter::updateDebugInfo() {
   DWPState State;
   if (opts::WriteDWP)
     initDWPState(State);
-  auto processUnitDIE = [&](DWARFUnit *Unit, DIEBuilder *DIEBlder) {
-    // Check if the unit is a skeleton and we need special updates for it and
-    // its matching split/DWO CU.
+  auto processSplitCU = [&](DWARFUnit &Unit, DWARFUnit &SplitCU,
+                            DIEBuilder &DIEBlder,
+                            DebugRangesSectionWriter &TempRangesSectionWriter,
+                            DebugAddrWriter &AddressWriter,
+                            const std::string &DWOName,
+                            const std::optional<std::string> &DwarfOutputPath) {
+    DIEBuilder DWODIEBuilder(BC, &(SplitCU).getContext(), DebugNamesTable,
+                             &Unit);
+    DWODIEBuilder.buildDWOUnit(SplitCU);
+    DebugStrOffsetsWriter DWOStrOffstsWriter(BC);
+    DebugStrWriter DWOStrWriter((SplitCU).getContext(), true);
+    DWODIEBuilder.updateDWONameCompDirForTypes(
+        DWOStrOffstsWriter, DWOStrWriter, SplitCU, DwarfOutputPath, DWOName);
+    DebugLoclistWriter DebugLocDWoWriter(Unit, Unit.getVersion(), true,
+                                         AddressWriter);
+
+    updateUnitDebugInfo(SplitCU, DWODIEBuilder, DebugLocDWoWriter,
+                        TempRangesSectionWriter, AddressWriter);
+    DebugLocDWoWriter.finalize(DWODIEBuilder,
+                               *DWODIEBuilder.getUnitDIEbyUnit(SplitCU));
+    if (Unit.getVersion() >= 5)
+      TempRangesSectionWriter.finalizeSection();
+
+    emitDWOBuilder(DWOName, DWODIEBuilder, *this, SplitCU, Unit, State,
+                   DebugLocDWoWriter, DWOStrOffstsWriter, DWOStrWriter,
+                   GDBIndexSection);
+  };
+  auto processMainBinaryCU = [&](DWARFUnit &Unit, DIEBuilder &DIEBlder) {
     std::optional<DWARFUnit *> SplitCU;
     std::optional<uint64_t> RangesBase;
-    std::optional<uint64_t> DWOId = Unit->getDWOId();
+    std::optional<uint64_t> DWOId = Unit.getDWOId();
     if (DWOId)
       SplitCU = BC.getDWOCU(*DWOId);
-    DebugLocWriter *DebugLocWriter = createRangeLocListAddressWriters(*Unit);
-    DebugRangesSectionWriter *RangesSectionWriter =
-        Unit->getVersion() >= 5 ? RangeListsSectionWriter.get()
-                                : LegacyRangesSectionWriter.get();
-    DebugAddrWriter *AddressWriter =
-        AddressWritersByCU[Unit->getOffset()].get();
-    // Skipping CUs that failed to load.
-    if (SplitCU) {
-      DIEBuilder DWODIEBuilder(BC, &(*SplitCU)->getContext(), DebugNamesTable,
-                               Unit);
-      DWODIEBuilder.buildDWOUnit(**SplitCU);
-      std::string DWOName = "";
-      std::optional<std::string> DwarfOutputPath =
-          opts::DwarfOutputPath.empty()
-              ? std::nullopt
-              : std::optional<std::string>(opts::DwarfOutputPath.c_str());
-      {
-        std::lock_guard<std::mutex> Lock(AccessMutex);
-        DWOName = DIEBlder->updateDWONameCompDir(
-            *StrOffstsWriter, *StrWriter, *Unit, DwarfOutputPath, std::nullopt);
-      }
-      DebugStrOffsetsWriter DWOStrOffstsWriter(BC);
-      DebugStrWriter DWOStrWriter((*SplitCU)->getContext(), true);
-      DWODIEBuilder.updateDWONameCompDirForTypes(DWOStrOffstsWriter,
-                                                 DWOStrWriter, **SplitCU,
-                                                 DwarfOutputPath, DWOName);
-      DebugLoclistWriter DebugLocDWoWriter(*Unit, Unit->getVersion(), true,
-                                           *AddressWriter);
-      DebugRangesSectionWriter *TempRangesSectionWriter = RangesSectionWriter;
-      if (Unit->getVersion() >= 5) {
-        TempRangesSectionWriter = RangeListsWritersByCU[*DWOId].get();
-      } else {
-        TempRangesSectionWriter = LegacyRangesWritersByCU[*DWOId].get();
-        RangesBase = RangesSectionWriter->getSectionOffset();
-      }
-
-      updateUnitDebugInfo(*(*SplitCU), DWODIEBuilder, DebugLocDWoWriter,
-                          *TempRangesSectionWriter, *AddressWriter);
-      DebugLocDWoWriter.finalize(DWODIEBuilder,
-                                 *DWODIEBuilder.getUnitDIEbyUnit(**SplitCU));
-      if (Unit->getVersion() >= 5)
-        TempRangesSectionWriter->finalizeSection();
-
-      emitDWOBuilder(DWOName, DWODIEBuilder, *this, **SplitCU, *Unit, State,
-                     DebugLocDWoWriter, DWOStrOffstsWriter, DWOStrWriter,
-                     GDBIndexSection);
-    }
-
-    if (Unit->getVersion() >= 5) {
-      RangesBase = RangesSectionWriter->getSectionOffset() +
+    DebugLocWriter &DebugLocWriter =
+        *LocListWritersByCU[LocListWritersIndexByCU[Unit.getOffset()]].get();
+    DebugRangesSectionWriter &RangesSectionWriter =
+        Unit.getVersion() >= 5 ? *RangeListsSectionWriter.get()
+                               : *LegacyRangesSectionWriter.get();
+    DebugAddrWriter &AddressWriter =
+        *AddressWritersByCU[Unit.getOffset()].get();
+    if (Unit.getVersion() >= 5)
+      RangeListsSectionWriter->setAddressWriter(&AddressWriter);
+    if (Unit.getVersion() >= 5) {
+      RangesBase = RangesSectionWriter.getSectionOffset() +
                    getDWARF5RngListLocListHeaderSize();
-      RangesSectionWriter->initSection(*Unit);
-      StrOffstsWriter->finalizeSection(*Unit, *DIEBlder);
+      RangesSectionWriter.initSection(Unit);
+      StrOffstsWriter->finalizeSection(Unit, DIEBlder);
+    } else if (SplitCU) {
+      RangesBase = LegacyRangesSectionWriter.get()->getSectionOffset();
     }
 
-    updateUnitDebugInfo(*Unit, *DIEBlder, *DebugLocWriter, *RangesSectionWriter,
-                        *AddressWriter, RangesBase);
-    DebugLocWriter->finalize(*DIEBlder, *DIEBlder->getUnitDIEbyUnit(*Unit));
-    if (Unit->getVersion() >= 5)
-      RangesSectionWriter->finalizeSection();
+    updateUnitDebugInfo(Unit, DIEBlder, DebugLocWriter, RangesSectionWriter,
+                        AddressWriter, RangesBase);
+    DebugLocWriter.finalize(DIEBlder, *DIEBlder.getUnitDIEbyUnit(Unit));
+    if (Unit.getVersion() >= 5)
+      RangesSectionWriter.finalizeSection();
   };
 
   DIEBuilder DIEBlder(BC, BC.DwCtx.get(), DebugNamesTable);
@@ -759,25 +735,35 @@ void DWARFRewriter::updateDebugInfo() {
   CUOffsetMap OffsetMap =
       finalizeTypeSections(DIEBlder, *Streamer, GDBIndexSection);
 
-  const bool SingleThreadedMode =
-      opts::NoThreads || opts::DeterministicDebugInfo;
-  if (!SingleThreadedMode)
-    DIEBlder.buildCompileUnits();
-  if (SingleThreadedMode) {
-    CUPartitionVector PartVec = partitionCUs(*BC.DwCtx);
-    for (std::vector<DWARFUnit *> &Vec : PartVec) {
-      DIEBlder.buildCompileUnits(Vec);
-      for (DWARFUnit *CU : DIEBlder.getProcessedCUs())
-        processUnitDIE(CU, &DIEBlder);
-      finalizeCompileUnits(DIEBlder, *Streamer, OffsetMap,
-                           DIEBlder.getProcessedCUs(), *FinalAddrWriter);
+  CUPartitionVector PartVec = partitionCUs(*BC.DwCtx);
+  for (std::vector<DWARFUnit *> &Vec : PartVec) {
+    DIEBlder.buildCompileUnits(Vec);
+    for (DWARFUnit *CU : DIEBlder.getProcessedCUs()) {
+      createRangeLocListAddressWriters(*CU);
+      std::optional<DWARFUnit *> SplitCU;
+      std::optional<uint64_t> DWOId = CU->getDWOId();
+      if (DWOId)
+        SplitCU = BC.getDWOCU(*DWOId);
+      if (!SplitCU)
+        continue;
+      DebugAddrWriter &AddressWriter =
+          *AddressWritersByCU[CU->getOffset()].get();
+      DebugRangesSectionWriter *TempRangesSectionWriter =
+          CU->getVersion() >= 5 ? RangeListsWritersByCU[*DWOId].get()
+                                : LegacyRangesWritersByCU[*DWOId].get();
+      std::optional<std::string> DwarfOutputPath =
+          opts::DwarfOutputPath.empty()
+              ? std::nullopt
+              : std::optional<std::string>(opts::DwarfOutputPath.c_str());
+      std::string DWOName = DIEBlder.updateDWONameCompDir(
+          *StrOffstsWriter, *StrWriter, *CU, DwarfOutputPath, std::nullopt);
+      processSplitCU(*CU, **SplitCU, DIEBlder, *TempRangesSectionWriter,
+                     AddressWriter, DWOName, DwarfOutputPath);
     }
-  } else {
-    // Update unit debug info in parallel
-    ThreadPoolInterface &ThreadPool = ParallelUtilities::getThreadPool();
-    for (std::unique_ptr<DWARFUnit> &CU : BC.DwCtx->compile_units())
-      ThreadPool.async(processUnitDIE, CU.get(), &DIEBlder);
-    ThreadPool.wait();
+    for (DWARFUnit *CU : DIEBlder.getProcessedCUs())
+      processMainBinaryCU(*CU, DIEBlder);
+    finalizeCompileUnits(DIEBlder, *Streamer, OffsetMap,
+                         DIEBlder.getProcessedCUs(), *FinalAddrWriter);
   }
 
   DebugNamesTable.emitAccelTable();

diff --git a/bolt/lib/Rewrite/PseudoProbeRewriter.cpp b/bolt/lib/Rewrite/PseudoProbeRewriter.cpp
@@ -49,6 +49,7 @@ static cl::opt<PrintPseudoProbesOptions> PrintPseudoProbes(
                clEnumValN(PPP_All, "all", "enable all debugging printout")),
     cl::Hidden, cl::cat(BoltCategory));
 
+extern cl::opt<bool> ProfileUsePseudoProbes;
 } // namespace opts
 
 namespace {
@@ -89,12 +90,15 @@ class PseudoProbeRewriter final : public MetadataRewriter {
 };
 
 Error PseudoProbeRewriter::preCFGInitializer() {
-  parsePseudoProbe();
+  if (opts::ProfileUsePseudoProbes)
+    parsePseudoProbe();
 
   return Error::success();
 }
 
 Error PseudoProbeRewriter::postEmitFinalizer() {
+  if (!opts::ProfileUsePseudoProbes)
+    parsePseudoProbe();
   updatePseudoProbes();
 
   return Error::success();

diff --git a/bolt/lib/RuntimeLibs/HugifyRuntimeLibrary.cpp b/bolt/lib/RuntimeLibs/HugifyRuntimeLibrary.cpp
@@ -32,10 +32,10 @@ cl::opt<bool>
                     "(which is what --hot-text relies on)."),
            cl::cat(BoltOptCategory));
 
-static cl::opt<std::string> RuntimeHugifyLib(
-    "runtime-hugify-lib",
-    cl::desc("specify file name of the runtime hugify library"),
-    cl::init("libbolt_rt_hugify.a"), cl::cat(BoltOptCategory));
+static cl::opt<std::string>
+    RuntimeHugifyLib("runtime-hugify-lib",
+                     cl::desc("specify path of the runtime hugify library"),
+                     cl::init("libbolt_rt_hugify.a"), cl::cat(BoltOptCategory));
 
 } // namespace opts
 

diff --git a/bolt/lib/RuntimeLibs/InstrumentationRuntimeLibrary.cpp b/bolt/lib/RuntimeLibs/InstrumentationRuntimeLibrary.cpp
@@ -26,7 +26,7 @@ namespace opts {
 
 cl::opt<std::string> RuntimeInstrumentationLib(
     "runtime-instrumentation-lib",
-    cl::desc("specify file name of the runtime instrumentation library"),
+    cl::desc("specify path of the runtime instrumentation library"),
     cl::init("libbolt_rt_instr.a"), cl::cat(BoltOptCategory));
 
 extern cl::opt<bool> InstrumentationFileAppendPID;

diff --git a/bolt/lib/RuntimeLibs/RuntimeLibrary.cpp b/bolt/lib/RuntimeLibs/RuntimeLibrary.cpp
@@ -26,8 +26,8 @@ using namespace bolt;
 
 void RuntimeLibrary::anchor() {}
 
-std::string RuntimeLibrary::getLibPath(StringRef ToolPath,
-                                       StringRef LibFileName) {
+std::string RuntimeLibrary::getLibPathByToolPath(StringRef ToolPath,
+                                                 StringRef LibFileName) {
   StringRef Dir = llvm::sys::path::parent_path(ToolPath);
   SmallString<128> LibPath = llvm::sys::path::parent_path(Dir);
   llvm::sys::path::append(LibPath, "lib" LLVM_LIBDIR_SUFFIX);
@@ -38,13 +38,36 @@ std::string RuntimeLibrary::getLibPath(StringRef ToolPath,
     llvm::sys::path::append(LibPath, "lib" LLVM_LIBDIR_SUFFIX);
   }
   llvm::sys::path::append(LibPath, LibFileName);
-  if (!llvm::sys::fs::exists(LibPath)) {
-    errs() << "BOLT-ERROR: library not found: " << LibPath << "\n";
-    exit(1);
-  }
   return std::string(LibPath);
 }
 
+std::string RuntimeLibrary::getLibPathByInstalled(StringRef LibFileName) {
+  SmallString<128> LibPath(CMAKE_INSTALL_FULL_LIBDIR);
+  llvm::sys::path::append(LibPath, LibFileName);
+  return std::string(LibPath);
+}
+
+std::string RuntimeLibrary::getLibPath(StringRef ToolPath,
+                                       StringRef LibFileName) {
+  if (llvm::sys::fs::exists(LibFileName)) {
+    return std::string(LibFileName);
+  }
+
+  std::string ByTool = getLibPathByToolPath(ToolPath, LibFileName);
+  if (llvm::sys::fs::exists(ByTool)) {
+    return ByTool;
+  }
+
+  std::string ByInstalled = getLibPathByInstalled(LibFileName);
+  if (llvm::sys::fs::exists(ByInstalled)) {
+    return ByInstalled;
+  }
+
+  errs() << "BOLT-ERROR: library not found: " << ByTool << ", " << ByInstalled
+         << ", or " << LibFileName << "\n";
+  exit(1);
+}
+
 void RuntimeLibrary::loadLibrary(StringRef LibPath, BOLTLinker &Linker,
                                  BOLTLinker::SectionsMapper MapSections) {
   ErrorOr<std::unique_ptr<MemoryBuffer>> MaybeBuf =

diff --git a/bolt/lib/Target/AArch64/CMakeLists.txt b/bolt/lib/Target/AArch64/CMakeLists.txt
@@ -4,6 +4,18 @@ set(LLVM_LINK_COMPONENTS
   AArch64Desc
   )
 
+if(BOLT_BUILT_STANDALONE)
+  set(LLVM_TARGET_DEFINITIONS ${LLVM_MAIN_SRC_DIR}/lib/Target/AArch64/AArch64.td)
+  list(APPEND LLVM_TABLEGEN_FLAGS -I ${LLVM_MAIN_SRC_DIR}/lib/Target/AArch64)
+  tablegen(LLVM AArch64GenInstrInfo.inc -gen-instr-info)
+  tablegen(LLVM AArch64GenRegisterInfo.inc -gen-register-info)
+  tablegen(LLVM AArch64GenSystemOperands.inc -gen-searchable-tables)
+  tablegen(LLVM AArch64GenSubtargetInfo.inc -gen-subtarget)
+
+  add_public_tablegen_target(AArch64CommonTableGen)
+  include_directories(${CMAKE_CURRENT_BINARY_DIR})
+endif()
+
 add_llvm_library(LLVMBOLTTargetAArch64
   AArch64MCPlusBuilder.cpp
 

diff --git a/bolt/lib/Target/RISCV/CMakeLists.txt b/bolt/lib/Target/RISCV/CMakeLists.txt
@@ -4,6 +4,19 @@ set(LLVM_LINK_COMPONENTS
   RISCVDesc
   )
 
+if(BOLT_BUILT_STANDALONE)
+  # tablegen, copied from llvm/lib/Target/RISCV/CMakeLists.txt
+  set(LLVM_TARGET_DEFINITIONS ${LLVM_MAIN_SRC_DIR}/lib/Target/RISCV/RISCV.td)
+  list(APPEND LLVM_TABLEGEN_FLAGS -I ${LLVM_MAIN_SRC_DIR}/lib/Target/RISCV)
+  tablegen(LLVM RISCVGenInstrInfo.inc -gen-instr-info)
+  tablegen(LLVM RISCVGenRegisterInfo.inc -gen-register-info)
+  tablegen(LLVM RISCVGenSearchableTables.inc -gen-searchable-tables)
+  tablegen(LLVM RISCVGenSubtargetInfo.inc -gen-subtarget)
+
+  add_public_tablegen_target(RISCVCommonTableGen)
+  include_directories(${CMAKE_CURRENT_BINARY_DIR})
+endif()
+
 add_llvm_library(LLVMBOLTTargetRISCV
   RISCVMCPlusBuilder.cpp
 

diff --git a/bolt/lib/Target/X86/CMakeLists.txt b/bolt/lib/Target/X86/CMakeLists.txt
@@ -5,6 +5,18 @@ set(LLVM_LINK_COMPONENTS
   X86Desc
   )
 
+if(BOLT_BUILT_STANDALONE)
+  set(LLVM_TARGET_DEFINITIONS ${LLVM_MAIN_SRC_DIR}/lib/Target/X86/X86.td)
+  list(APPEND LLVM_TABLEGEN_FLAGS -I ${LLVM_MAIN_SRC_DIR}/lib/Target/X86)
+  tablegen(LLVM X86GenInstrInfo.inc -gen-instr-info -instr-info-expand-mi-operand-info=0)
+  tablegen(LLVM X86GenMnemonicTables.inc -gen-x86-mnemonic-tables -asmwriternum=1)
+  tablegen(LLVM X86GenRegisterInfo.inc -gen-register-info)
+  tablegen(LLVM X86GenSubtargetInfo.inc -gen-subtarget)
+
+  add_public_tablegen_target(X86CommonTableGen)
+  include_directories(${CMAKE_CURRENT_BINARY_DIR})
+endif()
+
 add_llvm_library(LLVMBOLTTargetX86
   X86MCPlusBuilder.cpp
   X86MCSymbolizer.cpp

diff --git a/bolt/lib/Utils/CMakeLists.txt b/bolt/lib/Utils/CMakeLists.txt
@@ -1,15 +1,39 @@
+find_first_existing_vc_file("${LLVM_MAIN_SRC_DIR}" llvm_vc)
+find_first_existing_vc_file("${BOLT_SOURCE_DIR}" bolt_vc)
+
+# The VC revision include that we want to generate.
+set(version_inc "${CMAKE_CURRENT_BINARY_DIR}/VCSVersion.inc")
+
+set(generate_vcs_version_script "${LLVM_CMAKE_DIR}/GenerateVersionFromVCS.cmake")
+
+# Create custom target to generate the VC revision include.
+add_custom_command(OUTPUT "${version_inc}"
+  DEPENDS "${llvm_vc}" "${bolt_vc}" "${generate_vcs_version_script}"
+  COMMAND ${CMAKE_COMMAND} "-DNAMES=BOLT"
+                           "-DHEADER_FILE=${version_inc}"
+                           "-DBOLT_SOURCE_DIR=${BOLT_SOURCE_DIR}"
+                           "-DLLVM_VC_REPOSITORY=${llvm_vc_repository}"
+                           "-DLLVM_VC_REVISION=${llvm_vc_revision}"
+                           "-DLLVM_FORCE_VC_REVISION=${LLVM_FORCE_VC_REVISION}"
+                           "-DLLVM_FORCE_VC_REPOSITORY=${LLVM_FORCE_VC_REPOSITORY}"
+                           -P "${generate_vcs_version_script}")
+
+# Mark the generated header as being generated.
+set_source_files_properties("${version_inc}"
+  PROPERTIES GENERATED TRUE
+             HEADER_FILE_ONLY TRUE)
+
+include_directories(${CMAKE_CURRENT_BINARY_DIR})
+
 add_llvm_library(LLVMBOLTUtils
   CommandLineOpts.cpp
   Utils.cpp
-
+  ${version_inc}
   DISABLE_LLVM_LINK_LLVM_DYLIB
 
   LINK_LIBS
   ${LLVM_PTHREAD_LIB}
 
-  DEPENDS
-  llvm_vcsrevision_h
-
   LINK_COMPONENTS
   Support
   )
diff --git a/bolt/lib/Utils/CommandLineOpts.cpp b/bolt/lib/Utils/CommandLineOpts.cpp
@@ -11,15 +11,15 @@
 //===----------------------------------------------------------------------===//
 
 #include "bolt/Utils/CommandLineOpts.h"
-#include "llvm/Support/VCSRevision.h"
+#include "VCSVersion.inc"
 
 using namespace llvm;
 
 namespace llvm {
 namespace bolt {
 const char *BoltRevision =
-#ifdef LLVM_REVISION
-    LLVM_REVISION;
+#ifdef BOLT_REVISION
+    BOLT_REVISION;
 #else
     "<unknown>";
 #endif

diff --git a/bolt/runtime/CMakeLists.txt b/bolt/runtime/CMakeLists.txt
@@ -16,12 +16,19 @@ add_library(bolt_rt_instr STATIC
   instr.cpp
   ${CMAKE_CURRENT_BINARY_DIR}/config.h
   )
-set_target_properties(bolt_rt_instr PROPERTIES ARCHIVE_OUTPUT_DIRECTORY "${LLVM_LIBRARY_DIR}")
+set_target_properties(bolt_rt_instr PROPERTIES ARCHIVE_OUTPUT_DIRECTORY "${CMAKE_INSTALL_LIBDIR}${LLVM_LIBDIR_SUFFIX}")
 add_library(bolt_rt_hugify STATIC
   hugify.cpp
   ${CMAKE_CURRENT_BINARY_DIR}/config.h
   )
-set_target_properties(bolt_rt_hugify PROPERTIES ARCHIVE_OUTPUT_DIRECTORY "${LLVM_LIBRARY_DIR}")
+set_target_properties(bolt_rt_hugify PROPERTIES ARCHIVE_OUTPUT_DIRECTORY "${CMAKE_INSTALL_LIBDIR}${LLVM_LIBDIR_SUFFIX}")
+
+if(NOT BOLT_BUILT_STANDALONE)
+  add_custom_command(TARGET bolt_rt_instr POST_BUILD
+    COMMAND ${CMAKE_COMMAND} -E copy "${CMAKE_CURRENT_BINARY_DIR}/lib/libbolt_rt_instr.a" "${LLVM_LIBRARY_DIR}")
+  add_custom_command(TARGET bolt_rt_hugify POST_BUILD
+    COMMAND ${CMAKE_COMMAND} -E copy "${CMAKE_CURRENT_BINARY_DIR}/lib/libbolt_rt_hugify.a" "${LLVM_LIBRARY_DIR}")
+endif()
 
 set(BOLT_RT_FLAGS
   -ffreestanding
@@ -46,18 +53,23 @@ target_include_directories(bolt_rt_instr PRIVATE ${CMAKE_CURRENT_BINARY_DIR})
 target_compile_options(bolt_rt_hugify PRIVATE ${BOLT_RT_FLAGS})
 target_include_directories(bolt_rt_hugify PRIVATE ${CMAKE_CURRENT_BINARY_DIR})
 
-install(TARGETS bolt_rt_instr DESTINATION "lib${LLVM_LIBDIR_SUFFIX}")
-install(TARGETS bolt_rt_hugify DESTINATION "lib${LLVM_LIBDIR_SUFFIX}")
+install(TARGETS bolt_rt_instr DESTINATION "${CMAKE_INSTALL_LIBDIR}${LLVM_LIBDIR_SUFFIX}")
+install(TARGETS bolt_rt_hugify DESTINATION "${CMAKE_INSTALL_LIBDIR}${LLVM_LIBDIR_SUFFIX}")
 
 if (CMAKE_CXX_COMPILER_ID MATCHES ".*Clang.*" AND CMAKE_SYSTEM_NAME STREQUAL "Darwin")
   add_library(bolt_rt_instr_osx STATIC
     instr.cpp
     ${CMAKE_CURRENT_BINARY_DIR}/config.h
   )
-  set_target_properties(bolt_rt_instr_osx PROPERTIES ARCHIVE_OUTPUT_DIRECTORY "${LLVM_LIBRARY_DIR}")
+  set_target_properties(bolt_rt_instr_osx PROPERTIES ARCHIVE_OUTPUT_DIRECTORY "${CMAKE_INSTALL_LIBDIR}${LLVM_LIBDIR_SUFFIX}")
   target_include_directories(bolt_rt_instr_osx PRIVATE ${CMAKE_CURRENT_BINARY_DIR})
   target_compile_options(bolt_rt_instr_osx PRIVATE
     -target x86_64-apple-darwin19.6.0
     ${BOLT_RT_FLAGS})
-  install(TARGETS bolt_rt_instr_osx DESTINATION "lib${LLVM_LIBDIR_SUFFIX}")
+  install(TARGETS bolt_rt_instr_osx DESTINATION "${CMAKE_INSTALL_LIBDIR}${LLVM_LIBDIR_SUFFIX}")
+
+  if(NOT BOLT_BUILT_STANDALONE)
+    add_custom_command(TARGET bolt_rt_instr_osx POST_BUILD
+      COMMAND ${CMAKE_COMMAND} -E copy "${CMAKE_CURRENT_BINARY_DIR}/lib/libbolt_rt_instr_osx.a" "${LLVM_LIBRARY_DIR}")
+  endif()
 endif()
diff --git a/bolt/test/AArch64/update-debug-reloc.test b/bolt/test/AArch64/update-debug-reloc.test
@@ -2,7 +2,7 @@
 # update-debug-sections option.
 
 RUN: %clang %cflags -g %p/../Inputs/asm_foo.s %p/../Inputs/asm_main.c -o %t.exe
-RUN: llvm-bolt %t.exe -o %t --update-debug-sections
+RUN: llvm-bolt %t.exe -o %t --update-debug-sections 2>&1 | FileCheck %s
 
 CHECK: BOLT-INFO: Target architecture: aarch64
 CHECK-NOT: Reloc num: 10

diff --git a/bolt/test/AArch64/veneer-gold.s b/bolt/test/AArch64/veneer-gold.s
@@ -29,7 +29,7 @@ dummy:
 .type foo, %function
 foo:
 # CHECK: <foo>:
-# CHECK-NEXT : {{.*}} bl {{.*}} <foo2>
+# CHECK-NEXT: {{.*}} bl {{.*}} <foo2>
   bl .L2
   ret
 .size foo, .-foo
@@ -38,7 +38,7 @@ foo:
 .type foo2, %function
 foo2:
 # CHECK: <foo2>:
-# CHECK-NEXT : {{.*}} bl {{.*}} <foo2>
+# CHECK-NEXT: {{.*}} bl {{.*}} <foo2>
   bl .L2
   ret
 .size foo2, .-foo2

diff --git a/bolt/test/X86/dwarf5-df-types-modify-dwo-name-mixed.test b/bolt/test/X86/dwarf5-df-types-modify-dwo-name-mixed.test
@@ -90,7 +90,7 @@
 ; BOLT-DWP: DW_TAG_compile_unit
 ; BOLT-DWP: DW_AT_dwo_name  ("main.dwo.dwo")
 ; BOLT-DWP: DW_TAG_type_unit
-; BOLT-DW-NOT: DW_AT_dwo_name
+; BOLT-DWP-NOT: DW_AT_dwo_name
 ; BOLT-DWP:       Contribution size = 68, Format = DWARF32, Version = 5
 ; BOLT-DWP-NEXT: "main"
 ; BOLT-DWP-NEXT: "int"

diff --git a/bolt/test/X86/dwarf5-dwarf4-types-backward-forward-cross-reference.test b/bolt/test/X86/dwarf5-dwarf4-types-backward-forward-cross-reference.test
@@ -5,10 +5,11 @@
 # RUN: %clang %cflags %tmain.o %thelper.o -o %t.exe
 # RUN: llvm-bolt %t.exe -o %t.bolt --update-debug-sections
 # RUN: llvm-dwarfdump --show-form --verbose --debug-info %t.bolt | FileCheck --check-prefix=POSTCHECK %s
+# RUN: llvm-dwarfdump --show-form --verbose --debug-addr %t.bolt | FileCheck --check-prefix=POSTCHECKADDR %s
 # RUN: llvm-dwarfdump --show-form --verbose --debug-types %t.bolt | FileCheck --check-prefix=POSTCHECKTU %s
 
 ## This test checks that BOLT handles correctly backward and forward cross CU references
-## for DWARF5 and DWARF4 with -fdebug-types-section
+## for DWARF5 and DWARF4 with -fdebug-types-section and checks the address table is correct.
 
 # POSTCHECK: version = 0x0005
 # POSTCHECK: DW_TAG_type_unit
@@ -29,6 +30,15 @@
 # POSTCHECK: DW_TAG_variable [20]
 # POSTCHECK: DW_AT_type [DW_FORM_ref_addr] (0x{{[0-9a-f]+}} "Foo3a")
 
+# POSTCHECKADDR: Addrs: [
+# POSTCHECKADDR-NEXT: 0x0000000000001360
+# POSTCHECKADDR-NEXT: 0x0000000000000000
+# POSTCHECKADDR-NEXT: ]
+# POSTCHECKADDR: Addrs: [
+# POSTCHECKADDR-NEXT: 0x00000000000013e0
+# POSTCHECKADDR-NEXT: 0x0000000000000000
+# POSTCHECKADDR-NEXT: ]
+
 # POSTCHECKTU: version = 0x0004
 # POSTCHECKTU: DW_TAG_type_unit
 # POSTCHECKTU: DW_TAG_structure_type

diff --git a/bolt/test/X86/dwarf5-locexpr-referrence.test b/bolt/test/X86/dwarf5-locexpr-referrence.test
@@ -5,8 +5,10 @@
 # RUN: %clang %cflags -dwarf-5 %tmain.o %thelper.o -o %t.exe -Wl,-q
 # RUN: llvm-bolt %t.exe -o %t.bolt --update-debug-sections
 # RUN: llvm-dwarfdump --show-form --verbose --debug-info %t.bolt | FileCheck --check-prefix=CHECK %s
+# RUN: llvm-dwarfdump --show-form --verbose --debug-addr %t.bolt | FileCheck --check-prefix=CHECKADDR %s
 
-## This test checks that we update relative DIE references with DW_OP_convert that are in locexpr.
+## This test checks that we update relative DIE references with DW_OP_convert that are in locexpr
+## and checks the address table is correct.
 
 # CHECK: version = 0x0005
 # CHECK: DW_TAG_variable
@@ -19,3 +21,18 @@
 # CHECK-SAME: DW_OP_convert (0x00000028 -> 0x00000092)
 # CHECK-SAME: DW_OP_convert (0x0000002c -> 0x00000096)
 # CHECK: version = 0x0005
+
+# CHECKADDR: Addrs: [
+# CHECKADDR-NEXT: 0x0000000000001330
+# CHECKADDR-NEXT: 0x0000000000000000
+# CHECKADDR-NEXT: 0x0000000000001333
+# CHECKADDR-NEXT: ]
+# CHECKADDR: Addrs: [
+# CHECKADDR-NEXT: 0x0000000000001340
+# CHECKADDR-NEXT: 0x0000000000000000
+# CHECKADDR-NEXT: 0x0000000000001343
+# CHECKADDR-NEXT: ]
+# CHECKADDR: Addrs: [
+# CHECKADDR-NEXT: 0x0000000000001320
+# CHECKADDR-NEXT: 0x0000000000000000
+# CHECKADDR-NEXT: ]
diff --git a/bolt/test/X86/dwarf5-one-loclists-two-bases.test b/bolt/test/X86/dwarf5-one-loclists-two-bases.test
@@ -34,7 +34,7 @@
 # POSTCHECK: version = 0x0005
 # POSTCHECK: DW_AT_loclists_base [DW_FORM_sec_offset]	(0x0000000c)
 # POSTCHECK: DW_AT_rnglists_base [DW_FORM_sec_offset]	(0x0000000c)
-# POSTCHECK-EMPTY
+# POSTCHECK-EMPTY:
 # POSTCHECK: DW_TAG_variable
 # POSTCHECK: DW_AT_location [DW_FORM_loclistx]
 # POSTCHECK-SAME: indexed (0x0)

diff --git a/bolt/test/X86/dwarf5-two-loclists.test b/bolt/test/X86/dwarf5-two-loclists.test
@@ -45,7 +45,7 @@
 # POSTCHECK: version = 0x0005
 # POSTCHECK: DW_AT_loclists_base [DW_FORM_sec_offset]	(0x0000000c)
 # POSTCHECK: DW_AT_rnglists_base [DW_FORM_sec_offset]	(0x0000000c)
-# POSTCHECK-EMPTY
+# POSTCHECK-EMPTY:
 # POSTCHECK: DW_TAG_variable
 # POSTCHECK: DW_AT_location [DW_FORM_loclistx]
 # POSTCHECK-SAME: indexed (0x0)

diff --git a/bolt/test/X86/dwarf5-two-rnglists.test b/bolt/test/X86/dwarf5-two-rnglists.test
@@ -52,7 +52,7 @@
 # POSTCHECK-NEXT: DW_AT_addr_base [DW_FORM_sec_offset]  (0x00000008)
 # POSTCHECK-NEXT: DW_AT_loclists_base [DW_FORM_sec_offset]	(0x0000000c)
 # POSTCHECK-NEXT: DW_AT_rnglists_base [DW_FORM_sec_offset]	(0x0000000c)
-# POSTCHECK-EMPTY
+# POSTCHECK-EMPTY:
 # POSTCHECK: DW_TAG_subprogram
 # POSTCHECK-NEXT: DW_AT_ranges [DW_FORM_rnglistx]
 # POSTCHECK-SAME: indexed (0x1)
@@ -75,7 +75,7 @@
 # POSTCHECK-NEXT: DW_AT_addr_base [DW_FORM_sec_offset]  (0x00000030)
 # POSTCHECK-NEXT: DW_AT_loclists_base [DW_FORM_sec_offset]	(0x00000045)
 # POSTCHECK-NEXT: DW_AT_rnglists_base [DW_FORM_sec_offset]	(0x00000035)
-# POSTCHECK-EMPTY
+# POSTCHECK-EMPTY:
 
 # POSTCHECK: DW_TAG_subprogram
 # POSTCHECK-NEXT: DW_AT_ranges [DW_FORM_rnglistx]

diff --git a/bolt/test/X86/match-functions-with-call-graph.test b/bolt/test/X86/match-functions-with-call-graph.test
@@ -0,0 +1,103 @@
+## Tests blocks matching by called function names in inferStaleProfile.
+
+# REQUIRES: system-linux
+# RUN: split-file %s %t
+# RUN: %clang %cflags %t/main.cpp -o %t.exe -Wl,-q -nostdlib
+# RUN: llvm-bolt %t.exe -o %t.out --data %t/yaml --profile-ignore-hash -v=1 \
+# RUN:   --dyno-stats --print-cfg --infer-stale-profile=1 --match-with-call-graph 2>&1 | FileCheck %s
+
+# CHECK: BOLT-INFO: matched 1 functions with call graph
+
+#--- main.cpp
+void foo() {}
+
+void bar() {}
+
+void qux() {
+    foo();
+    bar();
+}
+
+void fred() {
+    foo();
+    qux();
+    bar();
+    bar();
+    foo();
+}
+
+int main() {
+    return 0;
+}
+
+#--- yaml
+---
+header:
+  profile-version: 1
+  binary-name:     'match-functions-with-calls-as-anchors.s.tmp.exe'
+  binary-build-id: '<unknown>'
+  profile-flags:   [ lbr ]
+  profile-origin:  branch profile reader
+  profile-events:  ''
+  dfs-order:       false
+  hash-func:       xxh3
+functions:
+  - name:            main
+    fid:             0
+    hash:            0x0000000000000001
+    exec:            1
+    nblocks:         6
+    blocks:
+      - bid:             1
+        hash:            0x0000000000000001
+        insns:           1
+        succ:            [ { bid: 3, cnt: 1} ]
+  - name:            _Z3foov
+    fid:             1
+    hash:            0x0000000000000002
+    exec:            1
+    nblocks:         6
+    blocks:
+      - bid:             1
+        hash:            0x0000000000000002
+        insns:           1
+        succ:            [ { bid: 3, cnt: 1} ]
+
+  - name:            _Z3barv
+    fid:             2
+    hash:            0x0000000000000003
+    exec:            1
+    nblocks:         6
+    blocks:
+      - bid:             1
+        hash:            0x0000000000000003
+        insns:           1
+        succ:            [ { bid: 3, cnt: 1} ]
+  - name:            _Z3quxv
+    fid:             3
+    hash:            0x0000000000000004
+    exec:            4
+    nblocks:         6
+    blocks:
+      - bid:             1
+        hash:            0x0000000000000004
+        insns:           1
+        succ:            [ { bid: 3, cnt: 1} ]
+        calls:           [ { off : 0, fid : 1, cnt : 0},
+                           { off : 0, fid : 2, cnt : 0} ]
+  - name:            _Z4bazv
+    fid:             4
+    hash:            0x0000000000000005
+    exec:            1
+    nblocks:         6
+    blocks:
+      - bid:             1
+        hash:            0x0000000000000005
+        insns:           1
+        succ:            [ { bid: 3, cnt: 1} ]
+        calls:           [ { off : 0, fid : 3, cnt : 0},
+                           { off : 0, fid : 1, cnt : 0},
+                           { off : 0, fid : 2, cnt : 0},
+                           { off : 0, fid : 1, cnt : 0},
+                           { off : 0, fid : 2, cnt : 0} ]
+...
diff --git a/bolt/test/X86/pseudoprobe-decoding-inline.test b/bolt/test/X86/pseudoprobe-decoding-inline.test
@@ -6,11 +6,11 @@
 # PREAGG: B X:0 #main# 1 0
 ## Check pseudo-probes in regular YAML profile (non-BOLTed binary)
 # RUN: link_fdata %s %S/../../../llvm/test/tools/llvm-profgen/Inputs/inline-cs-pseudoprobe.perfbin %t.preagg PREAGG
-# RUN: perf2bolt %S/../../../llvm/test/tools/llvm-profgen/Inputs/inline-cs-pseudoprobe.perfbin -p %t.preagg --pa -w %t.yaml -o %t.fdata
+# RUN: perf2bolt %S/../../../llvm/test/tools/llvm-profgen/Inputs/inline-cs-pseudoprobe.perfbin -p %t.preagg --pa -w %t.yaml -o %t.fdata --profile-use-pseudo-probes
 # RUN: FileCheck --input-file %t.yaml %s --check-prefix CHECK-YAML
 ## Check pseudo-probes in BAT YAML profile (BOLTed binary)
 # RUN: link_fdata %s %t.bolt %t.preagg2 PREAGG
-# RUN: perf2bolt %t.bolt -p %t.preagg2 --pa -w %t.yaml2 -o %t.fdata2
+# RUN: perf2bolt %t.bolt -p %t.preagg2 --pa -w %t.yaml2 -o %t.fdata2 --profile-use-pseudo-probes
 # RUN: FileCheck --input-file %t.yaml2 %s --check-prefix CHECK-YAML
 # CHECK-YAML: name: bar
 # CHECK-YAML: - bid: 0
@@ -29,6 +29,14 @@
 # CHECK-YAML:   pseudo_probes: [ { guid: 0xDB956436E78DD5FA, id: 1, type: 0 }, { guid: 0x5CF8C24CDB18BDAC, id: 1, type: 0 }, { guid: 0x5CF8C24CDB18BDAC, id: 2, type: 0 } ]
 # CHECK-YAML: guid: 0xDB956436E78DD5FA
 # CHECK-YAML: pseudo_probe_desc_hash: 0x10000FFFFFFFF
+#
+## Check that without --profile-use-pseudo-probes option, no pseudo probes are
+## generated
+# RUN: perf2bolt %S/../../../llvm/test/tools/llvm-profgen/Inputs/inline-cs-pseudoprobe.perfbin -p %t.preagg --pa -w %t.yaml -o %t.fdata
+# RUN: FileCheck --input-file %t.yaml %s --check-prefix CHECK-NO-OPT
+# CHECK-NO-OPT-NOT: pseudo_probes
+# CHECK-NO-OPT-NOT: guid
+# CHECK-NO-OPT-NOT: pseudo_probe_desc_hash
 
 CHECK: Report of decoding input pseudo probe binaries
 

diff --git a/bolt/test/X86/three-way-split-jt.s b/bolt/test/X86/three-way-split-jt.s
@@ -0,0 +1,95 @@
+## This reproduces an issue where the function is split into three fragments
+## and all fragments access the same jump table.
+
+# REQUIRES: system-linux
+
+# RUN: llvm-mc -filetype=obj -triple x86_64-unknown-unknown %s -o %t.o
+# RUN: llvm-strip --strip-unneeded %t.o
+# RUN: %clang %cflags %t.o -o %t.exe -Wl,-q
+# RUN: llvm-bolt %t.exe -o %t.out -v=1 -print-only=main.warm -print-cfg  2>&1 | FileCheck %s
+
+# CHECK-DAG: BOLT-INFO: marking main.warm as a fragment of main
+# CHECK-DAG: BOLT-INFO: marking main.cold as a fragment of main
+# CHECK-DAG: BOLT-INFO: processing main.warm as a sibling of non-ignored function
+# CHECK-DAG: BOLT-INFO: processing main.cold as a sibling of non-ignored function
+# CHECK-DAG: BOLT-WARNING: Ignoring main.cold
+# CHECK-DAG: BOLT-WARNING: Ignoring main.warm
+# CHECK-DAG: BOLT-WARNING: Ignoring main
+# CHECK: BOLT-WARNING: skipped 3 functions due to cold fragments
+
+# CHECK: PIC Jump table JUMP_TABLE for function main, main.warm, main.cold
+# CHECK-NEXT:   0x0000 : __ENTRY_main@0x[[#]]
+# CHECK-NEXT:   0x0004 : __ENTRY_main@0x[[#]]
+# CHECK-NEXT:   0x0008 : __ENTRY_main.cold@0x[[#]]
+# CHECK-NEXT:   0x000c : __ENTRY_main@0x[[#]]
+  .globl main
+  .type main, %function
+  .p2align 2
+main:
+LBB0:
+  andl $0xf, %ecx
+  cmpb $0x4, %cl
+  ## exit through ret
+  ja LBB3
+
+## jump table dispatch, jumping to label indexed by val in %ecx
+LBB1:
+  leaq JUMP_TABLE(%rip), %r8
+  movzbl %cl, %ecx
+  movslq (%r8,%rcx,4), %rax
+  addq %rax, %r8
+  jmpq *%r8
+
+LBB2:
+  xorq %rax, %rax
+LBB3:
+  addq $0x8, %rsp
+  ret
+.size main, .-main
+
+  .globl main.warm
+  .type main.warm, %function
+  .p2align 2
+main.warm:
+LBB20:
+  andl $0xb, %ebx
+  cmpb $0x1, %cl
+  # exit through ret
+  ja LBB23
+
+## jump table dispatch, jumping to label indexed by val in %ecx
+LBB21:
+  leaq JUMP_TABLE(%rip), %r8
+  movzbl %cl, %ecx
+  movslq (%r8,%rcx,4), %rax
+  addq %rax, %r8
+  jmpq *%r8
+
+LBB22:
+  xorq %rax, %rax
+LBB23:
+  addq $0x8, %rsp
+  ret
+.size main.warm, .-main.warm
+
+## cold fragment is only reachable through jump table
+  .globl main.cold
+  .type main.cold, %function
+main.cold:
+  leaq JUMP_TABLE(%rip), %r8
+  movzbl %cl, %ecx
+  movslq (%r8,%rcx,4), %rax
+  addq %rax, %r8
+  jmpq *%r8
+LBB4:
+  callq abort
+.size main.cold, .-main.cold
+
+  .rodata
+## jmp table, entries must be R_X86_64_PC32 relocs
+  .globl JUMP_TABLE
+JUMP_TABLE:
+  .long LBB2-JUMP_TABLE
+  .long LBB3-JUMP_TABLE
+  .long LBB4-JUMP_TABLE
+  .long LBB3-JUMP_TABLE
diff --git a/bolt/test/lit.cfg.py b/bolt/test/lit.cfg.py
@@ -92,10 +92,22 @@
 
 tool_dirs = [config.llvm_tools_dir, config.test_source_root]
 
+llvm_bolt_args = []
+
+if config.libbolt_rt_instr:
+    llvm_bolt_args.append(f"--runtime-instrumentation-lib={config.libbolt_rt_instr}")
+
+if config.libbolt_rt_hugify:
+    llvm_bolt_args.append(f"--runtime-hugify-lib={config.libbolt_rt_hugify}")
+
 tools = [
     ToolSubst("llc", unresolved="fatal"),
     ToolSubst("llvm-dwarfdump", unresolved="fatal"),
-    ToolSubst("llvm-bolt", unresolved="fatal"),
+    ToolSubst(
+        "llvm-bolt",
+        unresolved="fatal",
+        extra_args=llvm_bolt_args,
+    ),
     ToolSubst("llvm-boltdiff", unresolved="fatal"),
     ToolSubst("llvm-bolt-heatmap", unresolved="fatal"),
     ToolSubst("llvm-bat-dump", unresolved="fatal"),

diff --git a/bolt/test/lit.site.cfg.py.in b/bolt/test/lit.site.cfg.py.in
@@ -19,6 +19,8 @@ config.bolt_clang = "@BOLT_CLANG_EXE@"
 config.bolt_lld = "@BOLT_LLD_EXE@"
 config.targets_to_build = "@BOLT_TARGETS_TO_BUILD@"
 config.gnu_ld = "@GNU_LD_EXECUTABLE@"
+config.libbolt_rt_instr = "@LIBBOLT_RT_INSTR@"
+config.libbolt_rt_hugify = "@LIBBOLT_RT_HUGIFY@"
 
 import lit.llvm
 lit.llvm.initialize(lit_config, config)

diff --git a/bolt/test/perf2bolt/lit.local.cfg b/bolt/test/perf2bolt/lit.local.cfg
@@ -1,4 +1,4 @@
 import shutil
 
-if shutil.which("perf") != None:
+if shutil.which("perf") is not None:
     config.available_features.add("perf")
diff --git a/bolt/utils/docker/Dockerfile b/bolt/utils/docker/Dockerfile
@@ -1,4 +1,4 @@
-FROM ubuntu:20.04 AS builder
+FROM ubuntu:24.04 AS builder
 
 ARG DEBIAN_FRONTEND=noninteractive
 ENV TZ=UTC
@@ -26,6 +26,6 @@ RUN mkdir build && \
     ninja install-llvm-bolt install-perf2bolt install-merge-fdata \
       install-llvm-boltdiff install-bolt_rt
 
-FROM ubuntu:20.04
+FROM ubuntu:24.04
 
 COPY --from=builder /home/bolt/install /usr/local
diff --git a/clang-tools-extra/clang-doc/HTMLGenerator.cpp b/clang-tools-extra/clang-doc/HTMLGenerator.cpp
@@ -16,6 +16,7 @@
 #include "llvm/Support/JSON.h"
 #include "llvm/Support/Path.h"
 #include "llvm/Support/raw_ostream.h"
+#include <algorithm>
 #include <optional>
 #include <string>
 
@@ -979,6 +980,18 @@ static llvm::Error serializeIndex(ClangDocContext &CDCtx) {
                                    "error creating index file: " +
                                        FileErr.message());
   }
+  llvm::SmallString<128> RootPath(CDCtx.OutDirectory);
+  if (llvm::sys::path::is_relative(RootPath)) {
+    llvm::sys::fs::make_absolute(RootPath);
+  }
+  // Replace the escaped characters with a forward slash. It shouldn't matter
+  // when rendering the webpage in a web browser. This helps to prevent the
+  // JavaScript from escaping characters incorrectly, and introducing  bad paths
+  // in the URLs.
+  std::string RootPathEscaped = RootPath.str().str();
+  std::replace(RootPathEscaped.begin(), RootPathEscaped.end(), '\\', '/');
+  OS << "var RootPath = \"" << RootPathEscaped << "\";\n";
+
   CDCtx.Idx.sort();
   llvm::json::OStream J(OS, 2);
   std::function<void(Index)> IndexToJSON = [&](const Index &I) {

diff --git a/clang-tools-extra/clang-doc/Mapper.cpp b/clang-tools-extra/clang-doc/Mapper.cpp
@@ -12,16 +12,28 @@
 #include "clang/AST/Comment.h"
 #include "clang/Index/USRGeneration.h"
 #include "llvm/ADT/StringExtras.h"
-#include "llvm/Support/Error.h"
+#include "llvm/ADT/StringSet.h"
+#include "llvm/Support/Mutex.h"
 
 namespace clang {
 namespace doc {
 
+static llvm::StringSet<> USRVisited;
+static llvm::sys::Mutex USRVisitedGuard;
+
+template <typename T> bool isTypedefAnonRecord(const T *D) {
+  if (const auto *C = dyn_cast<CXXRecordDecl>(D)) {
+    return C->getTypedefNameForAnonDecl();
+  }
+  return false;
+}
+
 void MapASTVisitor::HandleTranslationUnit(ASTContext &Context) {
   TraverseDecl(Context.getTranslationUnitDecl());
 }
 
-template <typename T> bool MapASTVisitor::mapDecl(const T *D) {
+template <typename T>
+bool MapASTVisitor::mapDecl(const T *D, bool IsDefinition) {
   // If we're looking a decl not in user files, skip this decl.
   if (D->getASTContext().getSourceManager().isInSystemHeader(D->getLocation()))
     return true;
@@ -34,6 +46,16 @@ template <typename T> bool MapASTVisitor::mapDecl(const T *D) {
   // If there is an error generating a USR for the decl, skip this decl.
   if (index::generateUSRForDecl(D, USR))
     return true;
+  // Prevent Visiting USR twice
+  {
+    std::lock_guard<llvm::sys::Mutex> Guard(USRVisitedGuard);
+    StringRef Visited = USR.str();
+    if (USRVisited.count(Visited) && !isTypedefAnonRecord<T>(D))
+      return true;
+    // We considered a USR to be visited only when its defined
+    if (IsDefinition)
+      USRVisited.insert(Visited);
+  }
   bool IsFileInRootDir;
   llvm::SmallString<128> File =
       getFile(D, D->getASTContext(), CDCtx.SourceRoot, IsFileInRootDir);
@@ -53,30 +75,34 @@ template <typename T> bool MapASTVisitor::mapDecl(const T *D) {
 }
 
 bool MapASTVisitor::VisitNamespaceDecl(const NamespaceDecl *D) {
-  return mapDecl(D);
+  return mapDecl(D, /*isDefinition=*/true);
 }
 
-bool MapASTVisitor::VisitRecordDecl(const RecordDecl *D) { return mapDecl(D); }
+bool MapASTVisitor::VisitRecordDecl(const RecordDecl *D) {
+  return mapDecl(D, D->isThisDeclarationADefinition());
+}
 
-bool MapASTVisitor::VisitEnumDecl(const EnumDecl *D) { return mapDecl(D); }
+bool MapASTVisitor::VisitEnumDecl(const EnumDecl *D) {
+  return mapDecl(D, D->isThisDeclarationADefinition());
+}
 
 bool MapASTVisitor::VisitCXXMethodDecl(const CXXMethodDecl *D) {
-  return mapDecl(D);
+  return mapDecl(D, D->isThisDeclarationADefinition());
 }
 
 bool MapASTVisitor::VisitFunctionDecl(const FunctionDecl *D) {
   // Don't visit CXXMethodDecls twice
   if (isa<CXXMethodDecl>(D))
     return true;
-  return mapDecl(D);
+  return mapDecl(D, D->isThisDeclarationADefinition());
 }
 
 bool MapASTVisitor::VisitTypedefDecl(const TypedefDecl *D) {
-  return mapDecl(D);
+  return mapDecl(D, /*isDefinition=*/true);
 }
 
 bool MapASTVisitor::VisitTypeAliasDecl(const TypeAliasDecl *D) {
-  return mapDecl(D);
+  return mapDecl(D, /*isDefinition=*/true);
 }
 
 comments::FullComment *

diff --git a/clang-tools-extra/clang-doc/Mapper.h b/clang-tools-extra/clang-doc/Mapper.h
@@ -43,7 +43,7 @@ class MapASTVisitor : public clang::RecursiveASTVisitor<MapASTVisitor>,
   bool VisitTypeAliasDecl(const TypeAliasDecl *D);
 
 private:
-  template <typename T> bool mapDecl(const T *D);
+  template <typename T> bool mapDecl(const T *D, bool IsDefinition);
 
   int getLine(const NamedDecl *D, const ASTContext &Context) const;
   llvm::SmallString<128> getFile(const NamedDecl *D, const ASTContext &Context,

diff --git a/clang-tools-extra/clang-doc/assets/index.js b/clang-tools-extra/clang-doc/assets/index.js
@@ -1,42 +1,17 @@
-// Append using posix-style a file name or directory to Base
-function append(Base, New) {
-  if (!New)
-    return Base;
-  if (Base)
-    Base += "/";
-  Base += New;
-  return Base;
-}
-
-// Get relative path to access FilePath from CurrentDirectory
-function computeRelativePath(FilePath, CurrentDirectory) {
-  var Path = FilePath;
-  while (Path) {
-    if (CurrentDirectory == Path)
-      return FilePath.substring(Path.length + 1);
-    Path = Path.substring(0, Path.lastIndexOf("/"));
-  }
-
-  var Dir = CurrentDirectory;
-  var Result = "";
-  while (Dir) {
-    if (Dir == FilePath)
-      break;
-    Dir = Dir.substring(0, Dir.lastIndexOf("/"));
-    Result = append(Result, "..")
+function genLink(Ref) {
+  // we treat the file paths different depending on if we're
+  // serving via a http server or viewing from a local
+  var Path = window.location.protocol.startsWith("file") ?
+      `${window.location.protocol}//${window.location.host}/${Ref.Path}` :
+      `${window.location.protocol}//${RootPath}/${Ref.Path}`;
+  if (Ref.RefType === "namespace") {
+    Path = `${Path}/index.html`
+  } else if (Ref.Path === "") {
+      Path = `${Path}${Ref.Name}.html`;
+  } else {
+    Path = `${Path}/${Ref.Name}.html`;
   }
-  Result = append(Result, FilePath.substring(Dir.length))
-  return Result;
-}
-
-function genLink(Ref, CurrentDirectory) {
-  var Path = computeRelativePath(Ref.Path, CurrentDirectory);
-  if (Ref.RefType == "namespace")
-    Path = append(Path, "index.html");
-  else
-    Path = append(Path, Ref.Name + ".html")
-
-    ANode = document.createElement("a");
+  ANode = document.createElement("a");
   ANode.setAttribute("href", Path);
   var TextNode = document.createTextNode(Ref.Name);
   ANode.appendChild(TextNode);

diff --git a/clang-tools-extra/clang-doc/tool/ClangDocMain.cpp b/clang-tools-extra/clang-doc/tool/ClangDocMain.cpp
@@ -303,7 +303,6 @@ Example usage for a project using a compile commands database:
   for (auto &Group : USRToBitcode) {
     Pool.async([&]() {
       std::vector<std::unique_ptr<doc::Info>> Infos;
-
       for (auto &Bitcode : Group.getValue()) {
         llvm::BitstreamCursor Stream(Bitcode);
         doc::ClangDocBitcodeReader Reader(Stream);

diff --git a/clang-tools-extra/clang-tidy/add_new_check.py b/clang-tools-extra/clang-tidy/add_new_check.py
@@ -16,6 +16,7 @@
 import os
 import re
 import sys
+import textwrap
 
 # Adapts the module's CMakelist file. Returns 'True' if it could add a new
 # entry and 'False' if the entry already existed.
@@ -53,7 +54,14 @@ def adapt_cmake(module_path, check_name_camel):
 
 
 # Adds a header for the new check.
-def write_header(module_path, module, namespace, check_name, check_name_camel):
+def write_header(
+    module_path, module, namespace, check_name, check_name_camel, description
+):
+    wrapped_desc = "\n".join(
+        textwrap.wrap(
+            description, width=80, initial_indent="/// ", subsequent_indent="/// "
+        )
+    )
     filename = os.path.join(module_path, check_name_camel) + ".h"
     print("Creating %s..." % filename)
     with io.open(filename, "w", encoding="utf8", newline="\n") as f:
@@ -85,7 +93,7 @@ def write_header(module_path, module, namespace, check_name, check_name_camel):
 
 namespace clang::tidy::%(namespace)s {
 
-/// FIXME: Write a short description.
+%(description)s
 ///
 /// For the user-facing documentation see:
 /// http://clang.llvm.org/extra/clang-tidy/checks/%(module)s/%(check_name)s.html
@@ -107,6 +115,7 @@ class %(check_name_camel)s : public ClangTidyCheck {
                 "check_name": check_name,
                 "module": module,
                 "namespace": namespace,
+                "description": wrapped_desc,
             }
         )
 
@@ -235,7 +244,12 @@ def adapt_module(module_path, module, check_name, check_name_camel):
 
 
 # Adds a release notes entry.
-def add_release_notes(module_path, module, check_name):
+def add_release_notes(module_path, module, check_name, description):
+    wrapped_desc = "\n".join(
+        textwrap.wrap(
+            description, width=80, initial_indent="  ", subsequent_indent="  "
+        )
+    )
     check_name_dashes = module + "-" + check_name
     filename = os.path.normpath(
         os.path.join(module_path, "../../docs/ReleaseNotes.rst")
@@ -281,10 +295,10 @@ def add_release_notes(module_path, module, check_name):
                             """- New :doc:`%s
   <clang-tidy/checks/%s/%s>` check.
 
-  FIXME: add release notes.
+%s
 
 """
-                            % (check_name_dashes, module, check_name)
+                            % (check_name_dashes, module, check_name, wrapped_desc)
                         )
                         note_added = True
 
@@ -552,8 +566,8 @@ def format_link_alias(doc_file):
                 f.write('   :header: "Name", "Offers fixes"\n\n')
                 f.writelines(checks)
                 # and the aliases
-                f.write("\n\n")
-                f.write(".. csv-table:: Aliases..\n")
+                f.write("\nCheck aliases\n-------------\n\n")
+                f.write(".. csv-table::\n")
                 f.write('   :header: "Name", "Redirect", "Offers fixes"\n\n')
                 f.writelines(checks_alias)
                 break
@@ -612,6 +626,13 @@ def main():
         default="c++",
         metavar="LANG",
     )
+    parser.add_argument(
+        "--description",
+        "-d",
+        help="short description of what the check does",
+        default="FIXME: Write a short description",
+        type=str,
+    )
     parser.add_argument(
         "module",
         nargs="?",
@@ -652,10 +673,16 @@ def main():
     else:
         namespace = module
 
-    write_header(module_path, module, namespace, check_name, check_name_camel)
+    description = args.description
+    if not description.endswith("."):
+        description += "."
+
+    write_header(
+        module_path, module, namespace, check_name, check_name_camel, description
+    )
     write_implementation(module_path, module, namespace, check_name_camel)
     adapt_module(module_path, module, check_name, check_name_camel)
-    add_release_notes(module_path, module, check_name)
+    add_release_notes(module_path, module, check_name, description)
     test_extension = language_to_extension.get(args.language)
     write_test(module_path, module, check_name, test_extension)
     write_docs(module_path, module, check_name)

diff --git a/clang-tools-extra/clang-tidy/cert/CERTTidyModule.cpp b/clang-tools-extra/clang-tidy/cert/CERTTidyModule.cpp
diff --git a/clang-tools-extra/clang-tidy/misc/ConstCorrectnessCheck.cpp b/clang-tools-extra/clang-tidy/misc/ConstCorrectnessCheck.cpp
@@ -93,13 +93,12 @@ void ConstCorrectnessCheck::registerMatchers(MatchFinder *Finder) {
   // shall be run.
   const auto FunctionScope =
       functionDecl(
-          hasBody(
-              compoundStmt(forEachDescendant(
-                               declStmt(containsAnyDeclaration(
-                                            LocalValDecl.bind("local-value")),
-                                        unless(has(decompositionDecl())))
-                                   .bind("decl-stmt")))
-                  .bind("scope")))
+          hasBody(stmt(forEachDescendant(
+                           declStmt(containsAnyDeclaration(
+                                        LocalValDecl.bind("local-value")),
+                                    unless(has(decompositionDecl())))
+                               .bind("decl-stmt")))
+                      .bind("scope")))
           .bind("function-decl");
 
   Finder->addMatcher(FunctionScope, this);
@@ -109,7 +108,7 @@ void ConstCorrectnessCheck::registerMatchers(MatchFinder *Finder) {
 enum class VariableCategory { Value, Reference, Pointer };
 
 void ConstCorrectnessCheck::check(const MatchFinder::MatchResult &Result) {
-  const auto *LocalScope = Result.Nodes.getNodeAs<CompoundStmt>("scope");
+  const auto *LocalScope = Result.Nodes.getNodeAs<Stmt>("scope");
   const auto *Variable = Result.Nodes.getNodeAs<VarDecl>("local-value");
   const auto *Function = Result.Nodes.getNodeAs<FunctionDecl>("function-decl");
 
@@ -198,7 +197,7 @@ void ConstCorrectnessCheck::check(const MatchFinder::MatchResult &Result) {
   }
 }
 
-void ConstCorrectnessCheck::registerScope(const CompoundStmt *LocalScope,
+void ConstCorrectnessCheck::registerScope(const Stmt *LocalScope,
                                           ASTContext *Context) {
   auto &Analyzer = ScopesCache[LocalScope];
   if (!Analyzer)

diff --git a/clang-tools-extra/clang-tidy/misc/ConstCorrectnessCheck.h b/clang-tools-extra/clang-tidy/misc/ConstCorrectnessCheck.h
@@ -32,10 +32,10 @@ class ConstCorrectnessCheck : public ClangTidyCheck {
   void check(const ast_matchers::MatchFinder::MatchResult &Result) override;
 
 private:
-  void registerScope(const CompoundStmt *LocalScope, ASTContext *Context);
+  void registerScope(const Stmt *LocalScope, ASTContext *Context);
 
   using MutationAnalyzer = std::unique_ptr<ExprMutationAnalyzer>;
-  llvm::DenseMap<const CompoundStmt *, MutationAnalyzer> ScopesCache;
+  llvm::DenseMap<const Stmt *, MutationAnalyzer> ScopesCache;
   llvm::DenseSet<SourceLocation> TemplateDiagnosticsCache;
 
   const bool AnalyzeValues;

diff --git a/clang-tools-extra/clang-tidy/performance/UnnecessaryValueParamCheck.cpp b/clang-tools-extra/clang-tidy/performance/UnnecessaryValueParamCheck.cpp
@@ -119,65 +119,72 @@ void UnnecessaryValueParamCheck::check(const MatchFinder::MatchResult &Result) {
     }
   }
 
-  const size_t Index = llvm::find(Function->parameters(), Param) -
-                       Function->parameters().begin();
+  handleConstRefFix(*Function, *Param, *Result.Context);
+}
+
+void UnnecessaryValueParamCheck::registerPPCallbacks(
+    const SourceManager &SM, Preprocessor *PP, Preprocessor *ModuleExpanderPP) {
+  Inserter.registerPreprocessor(PP);
+}
+
+void UnnecessaryValueParamCheck::storeOptions(
+    ClangTidyOptions::OptionMap &Opts) {
+  Options.store(Opts, "IncludeStyle", Inserter.getStyle());
+  Options.store(Opts, "AllowedTypes",
+                utils::options::serializeStringList(AllowedTypes));
+}
+
+void UnnecessaryValueParamCheck::onEndOfTranslationUnit() {
+  MutationAnalyzerCache.clear();
+}
+
+void UnnecessaryValueParamCheck::handleConstRefFix(const FunctionDecl &Function,
+                                                   const ParmVarDecl &Param,
+                                                   ASTContext &Context) {
+  const size_t Index =
+      llvm::find(Function.parameters(), &Param) - Function.parameters().begin();
+  const bool IsConstQualified =
+      Param.getType().getCanonicalType().isConstQualified();
 
   auto Diag =
-      diag(Param->getLocation(),
+      diag(Param.getLocation(),
            "the %select{|const qualified }0parameter %1 is copied for each "
            "invocation%select{ but only used as a const reference|}0; consider "
            "making it a %select{const |}0reference")
-      << IsConstQualified << paramNameOrIndex(Param->getName(), Index);
+      << IsConstQualified << paramNameOrIndex(Param.getName(), Index);
   // Do not propose fixes when:
   // 1. the ParmVarDecl is in a macro, since we cannot place them correctly
   // 2. the function is virtual as it might break overrides
   // 3. the function is referenced outside of a call expression within the
   //    compilation unit as the signature change could introduce build errors.
   // 4. the function is an explicit template/ specialization.
-  const auto *Method = llvm::dyn_cast<CXXMethodDecl>(Function);
-  if (Param->getBeginLoc().isMacroID() || (Method && Method->isVirtual()) ||
-      isReferencedOutsideOfCallExpr(*Function, *Result.Context) ||
-      Function->getTemplateSpecializationKind() == TSK_ExplicitSpecialization)
+  const auto *Method = llvm::dyn_cast<CXXMethodDecl>(&Function);
+  if (Param.getBeginLoc().isMacroID() || (Method && Method->isVirtual()) ||
+      isReferencedOutsideOfCallExpr(Function, Context) ||
+      Function.getTemplateSpecializationKind() == TSK_ExplicitSpecialization)
     return;
-  for (const auto *FunctionDecl = Function; FunctionDecl != nullptr;
+  for (const auto *FunctionDecl = &Function; FunctionDecl != nullptr;
        FunctionDecl = FunctionDecl->getPreviousDecl()) {
     const auto &CurrentParam = *FunctionDecl->getParamDecl(Index);
-    Diag << utils::fixit::changeVarDeclToReference(CurrentParam,
-                                                   *Result.Context);
+    Diag << utils::fixit::changeVarDeclToReference(CurrentParam, Context);
     // The parameter of each declaration needs to be checked individually as to
     // whether it is const or not as constness can differ between definition and
     // declaration.
     if (!CurrentParam.getType().getCanonicalType().isConstQualified()) {
       if (std::optional<FixItHint> Fix = utils::fixit::addQualifierToVarDecl(
-              CurrentParam, *Result.Context, DeclSpec::TQ::TQ_const))
+              CurrentParam, Context, DeclSpec::TQ::TQ_const))
         Diag << *Fix;
     }
   }
 }
 
-void UnnecessaryValueParamCheck::registerPPCallbacks(
-    const SourceManager &SM, Preprocessor *PP, Preprocessor *ModuleExpanderPP) {
-  Inserter.registerPreprocessor(PP);
-}
-
-void UnnecessaryValueParamCheck::storeOptions(
-    ClangTidyOptions::OptionMap &Opts) {
-  Options.store(Opts, "IncludeStyle", Inserter.getStyle());
-  Options.store(Opts, "AllowedTypes",
-                utils::options::serializeStringList(AllowedTypes));
-}
-
-void UnnecessaryValueParamCheck::onEndOfTranslationUnit() {
-  MutationAnalyzerCache.clear();
-}
-
-void UnnecessaryValueParamCheck::handleMoveFix(const ParmVarDecl &Var,
+void UnnecessaryValueParamCheck::handleMoveFix(const ParmVarDecl &Param,
                                                const DeclRefExpr &CopyArgument,
-                                               const ASTContext &Context) {
+                                               ASTContext &Context) {
   auto Diag = diag(CopyArgument.getBeginLoc(),
                    "parameter %0 is passed by value and only copied once; "
                    "consider moving it to avoid unnecessary copies")
-              << &Var;
+              << &Param;
   // Do not propose fixes in macros since we cannot place them correctly.
   if (CopyArgument.getBeginLoc().isMacroID())
     return;

diff --git a/clang-tools-extra/clang-tidy/performance/UnnecessaryValueParamCheck.h b/clang-tools-extra/clang-tidy/performance/UnnecessaryValueParamCheck.h
@@ -33,10 +33,16 @@ class UnnecessaryValueParamCheck : public ClangTidyCheck {
   void storeOptions(ClangTidyOptions::OptionMap &Opts) override;
   void onEndOfTranslationUnit() override;
 
-private:
-  void handleMoveFix(const ParmVarDecl &Var, const DeclRefExpr &CopyArgument,
-                     const ASTContext &Context);
+protected:
+  // Create diagnostics. These are virtual so that derived classes can change
+  // behaviour.
+  virtual void handleMoveFix(const ParmVarDecl &Param,
+                             const DeclRefExpr &CopyArgument,
+                             ASTContext &Context);
+  virtual void handleConstRefFix(const FunctionDecl &Function,
+                                 const ParmVarDecl &Param, ASTContext &Context);
 
+private:
   ExprMutationAnalyzer::Memoized MutationAnalyzerCache;
   utils::IncludeInserter Inserter;
   const std::vector<StringRef> AllowedTypes;

diff --git a/clang-tools-extra/clang-tidy/readability/NonConstParameterCheck.cpp b/clang-tools-extra/clang-tidy/readability/NonConstParameterCheck.cpp
@@ -157,9 +157,12 @@ void NonConstParameterCheck::diagnoseNonConstParameters() {
     if (!Function)
       continue;
     unsigned Index = Par->getFunctionScopeIndex();
-    for (FunctionDecl *FnDecl : Function->redecls())
+    for (FunctionDecl *FnDecl : Function->redecls()) {
+      if (FnDecl->getNumParams() <= Index)
+        continue;
       Fixes.push_back(FixItHint::CreateInsertion(
           FnDecl->getParamDecl(Index)->getBeginLoc(), "const "));
+    }
 
     diag(Par->getLocation(), "pointer parameter '%0' can be pointer to const")
         << Par->getName() << Fixes;

diff --git a/clang-tools-extra/clang-tidy/tool/run-clang-tidy.py b/clang-tools-extra/clang-tidy/tool/run-clang-tidy.py
diff --git a/clang-tools-extra/clang-tidy/utils/ASTUtils.cpp b/clang-tools-extra/clang-tidy/utils/ASTUtils.cpp
@@ -96,7 +96,7 @@ bool areStatementsIdentical(const Stmt *FirstStmt, const Stmt *SecondStmt,
   if (FirstStmt == SecondStmt)
     return true;
 
-  if (FirstStmt->getStmtClass() != FirstStmt->getStmtClass())
+  if (FirstStmt->getStmtClass() != SecondStmt->getStmtClass())
     return false;
 
   if (isa<Expr>(FirstStmt) && isa<Expr>(SecondStmt)) {

diff --git a/clang-tools-extra/clang-tidy/utils/ExceptionAnalyzer.cpp b/clang-tools-extra/clang-tidy/utils/ExceptionAnalyzer.cpp
@@ -141,7 +141,10 @@ bool isStandardPointerConvertible(QualType From, QualType To) {
     if (RD->isCompleteDefinition() &&
         isBaseOf(From->getPointeeType().getTypePtr(),
                  To->getPointeeType().getTypePtr())) {
-      return true;
+      // If B is an inaccessible or ambiguous base class of D, a program
+      // that necessitates this conversion is ill-formed
+      return isUnambiguousPublicBaseClass(From->getPointeeType().getTypePtr(),
+                                          To->getPointeeType().getTypePtr());
     }
   }
 
@@ -375,10 +378,7 @@ bool ExceptionAnalyzer::ExceptionInfo::filterByCatch(
         isPointerOrPointerToMember(ExceptionCanTy->getTypePtr())) {
       // A standard pointer conversion not involving conversions to pointers to
       // private or protected or ambiguous classes ...
-      if (isStandardPointerConvertible(ExceptionCanTy, HandlerCanTy) &&
-          isUnambiguousPublicBaseClass(
-              ExceptionCanTy->getTypePtr()->getPointeeType().getTypePtr(),
-              HandlerCanTy->getTypePtr()->getPointeeType().getTypePtr())) {
+      if (isStandardPointerConvertible(ExceptionCanTy, HandlerCanTy)) {
         TypesToDelete.push_back(ExceptionTy);
       }
       // A function pointer conversion ...

diff --git a/clang-tools-extra/clang-tidy/utils/UseRangesCheck.cpp b/clang-tools-extra/clang-tidy/utils/UseRangesCheck.cpp
@@ -39,12 +39,6 @@ static constexpr const char ArgName[] = "ArgName";
 
 namespace clang::tidy::utils {
 
-static bool operator==(const UseRangesCheck::Indexes &L,
-                       const UseRangesCheck::Indexes &R) {
-  return std::tie(L.BeginArg, L.EndArg, L.ReplaceArg) ==
-         std::tie(R.BeginArg, R.EndArg, R.ReplaceArg);
-}
-
 static std::string getFullPrefix(ArrayRef<UseRangesCheck::Indexes> Signature) {
   std::string Output;
   llvm::raw_string_ostream OS(Output);
@@ -54,15 +48,6 @@ static std::string getFullPrefix(ArrayRef<UseRangesCheck::Indexes> Signature) {
   return Output;
 }
 
-static llvm::hash_code hash_value(const UseRangesCheck::Indexes &Indexes) {
-  return llvm::hash_combine(Indexes.BeginArg, Indexes.EndArg,
-                            Indexes.ReplaceArg);
-}
-
-static llvm::hash_code hash_value(const UseRangesCheck::Signature &Sig) {
-  return llvm::hash_combine_range(Sig.begin(), Sig.end());
-}
-
 namespace {
 
 AST_MATCHER(Expr, hasSideEffects) {
@@ -123,32 +108,34 @@ makeMatcherPair(StringRef State, const UseRangesCheck::Indexes &Indexes,
 }
 
 void UseRangesCheck::registerMatchers(MatchFinder *Finder) {
-  Replaces = getReplacerMap();
+  auto Replaces = getReplacerMap();
   ReverseDescriptor = getReverseDescriptor();
   auto BeginEndNames = getFreeBeginEndMethods();
   llvm::SmallVector<StringRef, 4> BeginNames{
       llvm::make_first_range(BeginEndNames)};
   llvm::SmallVector<StringRef, 4> EndNames{
       llvm::make_second_range(BeginEndNames)};
-  llvm::DenseSet<ArrayRef<Signature>> Seen;
+  Replacers.clear();
+  llvm::DenseSet<Replacer *> SeenRepl;
   for (auto I = Replaces.begin(), E = Replaces.end(); I != E; ++I) {
-    const ArrayRef<Signature> &Signatures =
-        I->getValue()->getReplacementSignatures();
-    if (!Seen.insert(Signatures).second)
+    auto Replacer = I->getValue();
+    if (!SeenRepl.insert(Replacer.get()).second)
       continue;
-    assert(!Signatures.empty() &&
-           llvm::all_of(Signatures, [](auto Index) { return !Index.empty(); }));
+    Replacers.push_back(Replacer);
+    assert(!Replacer->getReplacementSignatures().empty() &&
+           llvm::all_of(Replacer->getReplacementSignatures(),
+                        [](auto Index) { return !Index.empty(); }));
     std::vector<StringRef> Names(1, I->getKey());
     for (auto J = std::next(I); J != E; ++J)
-      if (J->getValue()->getReplacementSignatures() == Signatures)
+      if (J->getValue() == Replacer)
         Names.push_back(J->getKey());
 
     std::vector<ast_matchers::internal::DynTypedMatcher> TotalMatchers;
     // As we match on the first matched signature, we need to sort the
     // signatures in order of length(longest to shortest). This way any
     // signature that is a subset of another signature will be matched after the
     // other.
-    SmallVector<Signature> SigVec(Signatures);
+    SmallVector<Signature> SigVec(Replacer->getReplacementSignatures());
     llvm::sort(SigVec, [](auto &L, auto &R) { return R.size() < L.size(); });
     for (const auto &Signature : SigVec) {
       std::vector<ast_matchers::internal::DynTypedMatcher> Matchers;
@@ -163,7 +150,8 @@ void UseRangesCheck::registerMatchers(MatchFinder *Finder) {
     }
     Finder->addMatcher(
         callExpr(
-            callee(functionDecl(hasAnyName(std::move(Names))).bind(FuncDecl)),
+            callee(functionDecl(hasAnyName(std::move(Names)))
+                       .bind((FuncDecl + Twine(Replacers.size() - 1).str()))),
             ast_matchers::internal::DynTypedMatcher::constructVariadic(
                 ast_matchers::internal::DynTypedMatcher::VO_AnyOf,
                 ASTNodeKind::getFromNodeKind<CallExpr>(),
@@ -205,21 +193,33 @@ static void removeFunctionArgs(DiagnosticBuilder &Diag, const CallExpr &Call,
 }
 
 void UseRangesCheck::check(const MatchFinder::MatchResult &Result) {
-  const auto *Function = Result.Nodes.getNodeAs<FunctionDecl>(FuncDecl);
-  std::string Qualified = "::" + Function->getQualifiedNameAsString();
-  auto Iter = Replaces.find(Qualified);
-  assert(Iter != Replaces.end());
+  Replacer *Replacer = nullptr;
+  const FunctionDecl *Function = nullptr;
+  for (auto [Node, Value] : Result.Nodes.getMap()) {
+    StringRef NodeStr(Node);
+    if (!NodeStr.consume_front(FuncDecl))
+      continue;
+    Function = Value.get<FunctionDecl>();
+    size_t Index;
+    if (NodeStr.getAsInteger(10, Index)) {
+      llvm_unreachable("Unable to extract replacer index");
+    }
+    assert(Index < Replacers.size());
+    Replacer = Replacers[Index].get();
+    break;
+  }
+  assert(Replacer && Function);
   SmallString<64> Buffer;
-  for (const Signature &Sig : Iter->getValue()->getReplacementSignatures()) {
+  for (const Signature &Sig : Replacer->getReplacementSignatures()) {
     Buffer.assign({BoundCall, getFullPrefix(Sig)});
     const auto *Call = Result.Nodes.getNodeAs<CallExpr>(Buffer);
     if (!Call)
       continue;
     auto Diag = createDiag(*Call);
-    if (auto ReplaceName = Iter->getValue()->getReplaceName(*Function))
+    if (auto ReplaceName = Replacer->getReplaceName(*Function))
       Diag << FixItHint::CreateReplacement(Call->getCallee()->getSourceRange(),
                                            *ReplaceName);
-    if (auto Include = Iter->getValue()->getHeaderInclusion(*Function))
+    if (auto Include = Replacer->getHeaderInclusion(*Function))
       Diag << Inserter.createIncludeInsertion(
           Result.SourceManager->getFileID(Call->getBeginLoc()), *Include);
     llvm::SmallVector<unsigned, 3> ToRemove;

diff --git a/clang-tools-extra/clang-tidy/utils/UseRangesCheck.h b/clang-tools-extra/clang-tidy/utils/UseRangesCheck.h
@@ -85,7 +85,7 @@ class UseRangesCheck : public ClangTidyCheck {
   std::optional<TraversalKind> getCheckTraversalKind() const override;
 
 private:
-  ReplacerMap Replaces;
+  std::vector<llvm::IntrusiveRefCntPtr<Replacer>> Replacers;
   std::optional<ReverseIteratorDescriptor> ReverseDescriptor;
   IncludeInserter Inserter;
 };

diff --git a/clang-tools-extra/clangd/IncludeCleaner.cpp b/clang-tools-extra/clangd/IncludeCleaner.cpp
@@ -401,6 +401,26 @@ computeIncludeCleanerFindings(ParsedAST &AST, bool AnalyzeAngledIncludes) {
             Ref.RT != include_cleaner::RefType::Explicit)
           return;
 
+        // Check if we have any headers with the same spelling, in edge cases
+        // like `#include_next "foo.h"`, the user can't ever include the
+        // physical foo.h, but can have a spelling that refers to it.
+        // We postpone this check because spelling a header for every usage is
+        // expensive.
+        std::string Spelling = include_cleaner::spellHeader(
+            {Providers.front(), AST.getPreprocessor().getHeaderSearchInfo(),
+             MainFile});
+        for (auto *Inc :
+             ConvertedIncludes.match(include_cleaner::Header{Spelling})) {
+          Satisfied = true;
+          auto HeaderID =
+              AST.getIncludeStructure().getID(&Inc->Resolved->getFileEntry());
+          assert(HeaderID.has_value() &&
+                 "ConvertedIncludes only contains resolved includes.");
+          Used.insert(*HeaderID);
+        }
+        if (Satisfied)
+          return;
+
         // We actually always want to map usages to their spellings, but
         // spelling locations can point into preamble section. Using these
         // offsets could lead into crashes in presence of stale preambles. Hence

diff --git a/clang-tools-extra/clangd/unittests/IncludeCleanerTests.cpp b/clang-tools-extra/clangd/unittests/IncludeCleanerTests.cpp
@@ -644,6 +644,28 @@ TEST(IncludeCleaner, ResourceDirIsIgnored) {
   EXPECT_THAT(Findings.MissingIncludes, IsEmpty());
 }
 
+TEST(IncludeCleaner, DifferentHeaderSameSpelling) {
+  // `foo` is declared in foo_inner/foo.h, but there's no way to spell it
+  // directly. Make sure we don't generate unusued/missing include findings in
+  // such cases.
+  auto TU = TestTU::withCode(R"cpp(
+    #include <foo.h>
+    void baz() {
+      foo();
+    }
+  )cpp");
+  TU.AdditionalFiles["foo/foo.h"] = guard("#include_next <foo.h>");
+  TU.AdditionalFiles["foo_inner/foo.h"] = guard(R"cpp(
+    void foo();
+  )cpp");
+  TU.ExtraArgs.push_back("-Ifoo");
+  TU.ExtraArgs.push_back("-Ifoo_inner");
+
+  auto AST = TU.build();
+  auto Findings = computeIncludeCleanerFindings(AST);
+  EXPECT_THAT(Findings.UnusedIncludes, IsEmpty());
+  EXPECT_THAT(Findings.MissingIncludes, IsEmpty());
+}
 } // namespace
 } // namespace clangd
 } // namespace clang
diff --git a/clang-tools-extra/docs/ReleaseNotes.rst b/clang-tools-extra/docs/ReleaseNotes.rst
diff --git a/clang-tools-extra/docs/clang-tidy/checks/clang-analyzer/cplusplus.Move.rst b/clang-tools-extra/docs/clang-tidy/checks/clang-analyzer/cplusplus.Move.rst
@@ -1,9 +1,13 @@
 .. title:: clang-tidy - clang-analyzer-cplusplus.Move
+.. meta::
+   :http-equiv=refresh: 5;URL=https://clang.llvm.org/docs/analyzer/checkers.html#cplusplus-move
 
 clang-analyzer-cplusplus.Move
 =============================
 
 Find use-after-move bugs in C++.
 
-The clang-analyzer-cplusplus.Move check is an alias of
-Clang Static Analyzer cplusplus.Move.
+The `clang-analyzer-cplusplus.Move` check is an alias, please see
+`Clang Static Analyzer Available Checkers
+<https://clang.llvm.org/docs/analyzer/checkers.html#cplusplus-move>`_
+for more information.
diff --git a/...-tools-extra/docs/clang-tidy/checks/clang-analyzer/optin.taint.TaintedAlloc.rst b/...-tools-extra/docs/clang-tidy/checks/clang-analyzer/optin.taint.TaintedAlloc.rst
@@ -0,0 +1,14 @@
+.. title:: clang-tidy - clang-analyzer-optin.taint.TaintedAlloc
+.. meta::
+   :http-equiv=refresh: 5;URL=https://clang.llvm.org/docs/analyzer/checkers.html#optin-taint-taintedalloc
+
+clang-analyzer-optin.taint.TaintedAlloc
+=======================================
+
+Check for memory allocations, where the size parameter might be a tainted
+(attacker controlled) value.
+
+The `clang-analyzer-optin.taint.TaintedAlloc` check is an alias, please see
+`Clang Static Analyzer Available Checkers
+<https://clang.llvm.org/docs/analyzer/checkers.html#optin-taint-taintedalloc>`_
+for more information.
diff --git a/...tools-extra/docs/clang-tidy/checks/clang-analyzer/security.PutenvStackArray.rst b/...tools-extra/docs/clang-tidy/checks/clang-analyzer/security.PutenvStackArray.rst
@@ -0,0 +1,10 @@
+.. title:: clang-tidy - clang-analyzer-security.PutenvStackArray
+
+clang-analyzer-security.PutenvStackArray
+========================================
+
+Finds calls to the function 'putenv' which pass a pointer to an automatic
+(stack-allocated) array as the argument.
+
+The clang-analyzer-security.PutenvStackArray check is an alias of
+Clang Static Analyzer security.PutenvStackArray.
diff --git a/...ols-extra/docs/clang-tidy/checks/clang-analyzer/unix.BlockInCriticalSection.rst b/...ols-extra/docs/clang-tidy/checks/clang-analyzer/unix.BlockInCriticalSection.rst
@@ -0,0 +1,13 @@
+.. title:: clang-tidy - clang-analyzer-unix.BlockInCriticalSection
+.. meta::
+   :http-equiv=refresh: 5;URL=https://clang.llvm.org/docs/analyzer/checkers.html#unix-blockincriticalsection
+
+clang-analyzer-unix.BlockInCriticalSection
+==========================================
+
+Check for calls to blocking functions inside a critical section.
+
+The `clang-analyzer-unix.BlockInCriticalSection` check is an alias, please see
+`Clang Static Analyzer Available Checkers
+<https://clang.llvm.org/docs/analyzer/checkers.html#unix-blockincriticalsection>`_
+for more information.
diff --git a/clang-tools-extra/docs/clang-tidy/checks/list.rst b/clang-tools-extra/docs/clang-tidy/checks/list.rst
@@ -118,6 +118,7 @@ Clang-Tidy Checks
    :doc:`bugprone-not-null-terminated-result <bugprone/not-null-terminated-result>`, "Yes"
    :doc:`bugprone-optional-value-conversion <bugprone/optional-value-conversion>`, "Yes"
    :doc:`bugprone-parent-virtual-call <bugprone/parent-virtual-call>`, "Yes"
+   :doc:`bugprone-pointer-arithmetic-on-polymorphic-object <bugprone/pointer-arithmetic-on-polymorphic-object>`,
    :doc:`bugprone-posix-return <bugprone/posix-return>`, "Yes"
    :doc:`bugprone-redundant-branch-condition <bugprone/redundant-branch-condition>`, "Yes"
    :doc:`bugprone-reserved-identifier <bugprone/reserved-identifier>`, "Yes"
@@ -158,7 +159,6 @@ Clang-Tidy Checks
    :doc:`bugprone-unused-raii <bugprone/unused-raii>`, "Yes"
    :doc:`bugprone-unused-return-value <bugprone/unused-return-value>`,
    :doc:`bugprone-use-after-move <bugprone/use-after-move>`,
-   :doc:`bugprone-pointer-arithmetic-on-polymorphic-object <bugprone/pointer-arithmetic-on-polymorphic-object>`,
    :doc:`bugprone-virtual-near-miss <bugprone/virtual-near-miss>`, "Yes"
    :doc:`cert-dcl50-cpp <cert/dcl50-cpp>`,
    :doc:`cert-dcl58-cpp <cert/dcl58-cpp>`,
@@ -269,7 +269,7 @@ Clang-Tidy Checks
    :doc:`misc-unused-parameters <misc/unused-parameters>`, "Yes"
    :doc:`misc-unused-using-decls <misc/unused-using-decls>`, "Yes"
    :doc:`misc-use-anonymous-namespace <misc/use-anonymous-namespace>`,
-   :doc:`misc-use-internal-linkage <misc/use-internal-linkage>`,
+   :doc:`misc-use-internal-linkage <misc/use-internal-linkage>`, "Yes"
    :doc:`modernize-avoid-bind <modernize/avoid-bind>`, "Yes"
    :doc:`modernize-avoid-c-arrays <modernize/avoid-c-arrays>`,
    :doc:`modernize-concat-nested-namespaces <modernize/concat-nested-namespaces>`, "Yes"
@@ -409,6 +409,7 @@ Check aliases
    :doc:`bugprone-narrowing-conversions <bugprone/narrowing-conversions>`, :doc:`cppcoreguidelines-narrowing-conversions <cppcoreguidelines/narrowing-conversions>`,
    :doc:`cert-con36-c <cert/con36-c>`, :doc:`bugprone-spuriously-wake-up-functions <bugprone/spuriously-wake-up-functions>`,
    :doc:`cert-con54-cpp <cert/con54-cpp>`, :doc:`bugprone-spuriously-wake-up-functions <bugprone/spuriously-wake-up-functions>`,
+   :doc:`cert-ctr56-cpp <cert/ctr56-cpp>`, :doc:`bugprone-pointer-arithmetic-on-polymorphic-object <bugprone/pointer-arithmetic-on-polymorphic-object>`,
    :doc:`cert-dcl03-c <cert/dcl03-c>`, :doc:`misc-static-assert <misc/static-assert>`, "Yes"
    :doc:`cert-dcl16-c <cert/dcl16-c>`, :doc:`readability-uppercase-literal-suffix <readability/uppercase-literal-suffix>`, "Yes"
    :doc:`cert-dcl37-c <cert/dcl37-c>`, :doc:`bugprone-reserved-identifier <bugprone/reserved-identifier>`, "Yes"
@@ -448,7 +449,7 @@ Check aliases
    :doc:`clang-analyzer-core.uninitialized.UndefReturn <clang-analyzer/core.uninitialized.UndefReturn>`, `Clang Static Analyzer core.uninitialized.UndefReturn <https://clang.llvm.org/docs/analyzer/checkers.html#core-uninitialized-undefreturn>`_,
    :doc:`clang-analyzer-cplusplus.ArrayDelete <clang-analyzer/cplusplus.ArrayDelete>`, `Clang Static Analyzer cplusplus.ArrayDelete <https://clang.llvm.org/docs/analyzer/checkers.html#cplusplus-arraydelete>`_,
    :doc:`clang-analyzer-cplusplus.InnerPointer <clang-analyzer/cplusplus.InnerPointer>`, `Clang Static Analyzer cplusplus.InnerPointer <https://clang.llvm.org/docs/analyzer/checkers.html#cplusplus-innerpointer>`_,
-   :doc:`clang-analyzer-cplusplus.Move <clang-analyzer/cplusplus.Move>`, Clang Static Analyzer cplusplus.Move,
+   :doc:`clang-analyzer-cplusplus.Move <clang-analyzer/cplusplus.Move>`, `Clang Static Analyzer cplusplus.Move <https://clang.llvm.org/docs/analyzer/checkers.html#cplusplus-move>`_,
    :doc:`clang-analyzer-cplusplus.NewDelete <clang-analyzer/cplusplus.NewDelete>`, `Clang Static Analyzer cplusplus.NewDelete <https://clang.llvm.org/docs/analyzer/checkers.html#cplusplus-newdelete>`_,
    :doc:`clang-analyzer-cplusplus.NewDeleteLeaks <clang-analyzer/cplusplus.NewDeleteLeaks>`, `Clang Static Analyzer cplusplus.NewDeleteLeaks <https://clang.llvm.org/docs/analyzer/checkers.html#cplusplus-newdeleteleaks>`_,
    :doc:`clang-analyzer-cplusplus.PlacementNew <clang-analyzer/cplusplus.PlacementNew>`, `Clang Static Analyzer cplusplus.PlacementNew <https://clang.llvm.org/docs/analyzer/checkers.html#cplusplus-placementnew>`_,
@@ -471,6 +472,7 @@ Check aliases
    :doc:`clang-analyzer-optin.performance.GCDAntipattern <clang-analyzer/optin.performance.GCDAntipattern>`, `Clang Static Analyzer optin.performance.GCDAntipattern <https://clang.llvm.org/docs/analyzer/checkers.html#optin-performance-gcdantipattern>`_,
    :doc:`clang-analyzer-optin.performance.Padding <clang-analyzer/optin.performance.Padding>`, `Clang Static Analyzer optin.performance.Padding <https://clang.llvm.org/docs/analyzer/checkers.html#optin-performance-padding>`_,
    :doc:`clang-analyzer-optin.portability.UnixAPI <clang-analyzer/optin.portability.UnixAPI>`, `Clang Static Analyzer optin.portability.UnixAPI <https://clang.llvm.org/docs/analyzer/checkers.html#optin-portability-unixapi>`_,
+   :doc:`clang-analyzer-optin.taint.TaintedAlloc <clang-analyzer/optin.taint.TaintedAlloc>`, `Clang Static Analyzer optin.taint.TaintedAlloc <https://clang.llvm.org/docs/analyzer/checkers.html#optin-taint-taintedalloc>`_,
    :doc:`clang-analyzer-osx.API <clang-analyzer/osx.API>`, `Clang Static Analyzer osx.API <https://clang.llvm.org/docs/analyzer/checkers.html#osx-api>`_,
    :doc:`clang-analyzer-osx.MIG <clang-analyzer/osx.MIG>`, Clang Static Analyzer osx.MIG,
    :doc:`clang-analyzer-osx.NumberObjectConversion <clang-analyzer/osx.NumberObjectConversion>`, `Clang Static Analyzer osx.NumberObjectConversion <https://clang.llvm.org/docs/analyzer/checkers.html#osx-numberobjectconversion>`_,
@@ -501,6 +503,7 @@ Check aliases
    :doc:`clang-analyzer-osx.coreFoundation.containers.OutOfBounds <clang-analyzer/osx.coreFoundation.containers.OutOfBounds>`, `Clang Static Analyzer osx.coreFoundation.containers.OutOfBounds <https://clang.llvm.org/docs/analyzer/checkers.html#osx-corefoundation-containers-outofbounds>`_,
    :doc:`clang-analyzer-osx.coreFoundation.containers.PointerSizedValues <clang-analyzer/osx.coreFoundation.containers.PointerSizedValues>`, `Clang Static Analyzer osx.coreFoundation.containers.PointerSizedValues <https://clang.llvm.org/docs/analyzer/checkers.html#osx-corefoundation-containers-pointersizedvalues>`_,
    :doc:`clang-analyzer-security.FloatLoopCounter <clang-analyzer/security.FloatLoopCounter>`, `Clang Static Analyzer security.FloatLoopCounter <https://clang.llvm.org/docs/analyzer/checkers.html#security-floatloopcounter>`_,
+   :doc:`clang-analyzer-security.PutenvStackArray <clang-analyzer/security.PutenvStackArray>`, Clang Static Analyzer security.PutenvStackArray,
    :doc:`clang-analyzer-security.SetgidSetuidOrder <clang-analyzer/security.SetgidSetuidOrder>`, Clang Static Analyzer security.SetgidSetuidOrder,
    :doc:`clang-analyzer-security.cert.env.InvalidPtr <clang-analyzer/security.cert.env.InvalidPtr>`, `Clang Static Analyzer security.cert.env.InvalidPtr <https://clang.llvm.org/docs/analyzer/checkers.html#security-cert-env-invalidptr>`_,
    :doc:`clang-analyzer-security.insecureAPI.DeprecatedOrUnsafeBufferHandling <clang-analyzer/security.insecureAPI.DeprecatedOrUnsafeBufferHandling>`, `Clang Static Analyzer security.insecureAPI.DeprecatedOrUnsafeBufferHandling <https://clang.llvm.org/docs/analyzer/checkers.html#security-insecureapi-deprecatedorunsafebufferhandling>`_,
@@ -517,6 +520,7 @@ Check aliases
    :doc:`clang-analyzer-security.insecureAPI.strcpy <clang-analyzer/security.insecureAPI.strcpy>`, `Clang Static Analyzer security.insecureAPI.strcpy <https://clang.llvm.org/docs/analyzer/checkers.html#security-insecureapi-strcpy>`_,
    :doc:`clang-analyzer-security.insecureAPI.vfork <clang-analyzer/security.insecureAPI.vfork>`, `Clang Static Analyzer security.insecureAPI.vfork <https://clang.llvm.org/docs/analyzer/checkers.html#security-insecureapi-vfork>`_,
    :doc:`clang-analyzer-unix.API <clang-analyzer/unix.API>`, `Clang Static Analyzer unix.API <https://clang.llvm.org/docs/analyzer/checkers.html#unix-api>`_,
+   :doc:`clang-analyzer-unix.BlockInCriticalSection <clang-analyzer/unix.BlockInCriticalSection>`, `Clang Static Analyzer unix.BlockInCriticalSection <https://clang.llvm.org/docs/analyzer/checkers.html#unix-blockincriticalsection>`_,
    :doc:`clang-analyzer-unix.Errno <clang-analyzer/unix.Errno>`, `Clang Static Analyzer unix.Errno <https://clang.llvm.org/docs/analyzer/checkers.html#unix-errno>`_,
    :doc:`clang-analyzer-unix.Malloc <clang-analyzer/unix.Malloc>`, `Clang Static Analyzer unix.Malloc <https://clang.llvm.org/docs/analyzer/checkers.html#unix-malloc>`_,
    :doc:`clang-analyzer-unix.MallocSizeof <clang-analyzer/unix.MallocSizeof>`, `Clang Static Analyzer unix.MallocSizeof <https://clang.llvm.org/docs/analyzer/checkers.html#unix-mallocsizeof>`_,

diff --git a/clang-tools-extra/include-cleaner/lib/Analysis.cpp b/clang-tools-extra/include-cleaner/lib/Analysis.cpp
@@ -105,9 +105,20 @@ analyze(llvm::ArrayRef<Decl *> ASTRoots,
              }
              if (!Satisfied && !Providers.empty() &&
                  Ref.RT == RefType::Explicit &&
-                 !HeaderFilter(Providers.front().resolvedPath()))
-               Missing.insert(spellHeader(
-                   {Providers.front(), PP.getHeaderSearchInfo(), MainFile}));
+                 !HeaderFilter(Providers.front().resolvedPath())) {
+               // Check if we have any headers with the same spelling, in edge
+               // cases like `#include_next "foo.h"`, the user can't ever
+               // include the physical foo.h, but can have a spelling that
+               // refers to it.
+               auto Spelling = spellHeader(
+                   {Providers.front(), PP.getHeaderSearchInfo(), MainFile});
+               for (const Include *I : Inc.match(Header{Spelling})) {
+                 Used.insert(I);
+                 Satisfied = true;
+               }
+               if (!Satisfied)
+                 Missing.insert(std::move(Spelling));
+             }
            });
 
   AnalysisResults Results;

diff --git a/clang-tools-extra/include-cleaner/unittests/AnalysisTest.cpp b/clang-tools-extra/include-cleaner/unittests/AnalysisTest.cpp
@@ -12,6 +12,7 @@
 #include "clang-include-cleaner/Record.h"
 #include "clang-include-cleaner/Types.h"
 #include "clang/AST/ASTContext.h"
+#include "clang/AST/DeclBase.h"
 #include "clang/Basic/FileManager.h"
 #include "clang/Basic/IdentifierTable.h"
 #include "clang/Basic/SourceLocation.h"
@@ -296,6 +297,31 @@ TEST_F(AnalyzeTest, ResourceDirIsIgnored) {
   EXPECT_THAT(Results.Missing, testing::IsEmpty());
 }
 
+TEST_F(AnalyzeTest, DifferentHeaderSameSpelling) {
+  Inputs.ExtraArgs.push_back("-Ifoo");
+  Inputs.ExtraArgs.push_back("-Ifoo_inner");
+  // `foo` is declared in foo_inner/foo.h, but there's no way to spell it
+  // directly. Make sure we don't generate unusued/missing include findings in
+  // such cases.
+  Inputs.Code = R"cpp(
+    #include <foo.h>
+    void baz() {
+      foo();
+    }
+  )cpp";
+  Inputs.ExtraFiles["foo/foo.h"] = guard("#include_next <foo.h>");
+  Inputs.ExtraFiles["foo_inner/foo.h"] = guard(R"cpp(
+    void foo();
+  )cpp");
+  TestAST AST(Inputs);
+  std::vector<Decl *> DeclsInTU;
+  for (auto *D : AST.context().getTranslationUnitDecl()->decls())
+    DeclsInTU.push_back(D);
+  auto Results = analyze(DeclsInTU, {}, PP.Includes, &PI, AST.preprocessor());
+  EXPECT_THAT(Results.Unused, testing::IsEmpty());
+  EXPECT_THAT(Results.Missing, testing::IsEmpty());
+}
+
 TEST(FixIncludes, Basic) {
   llvm::StringRef Code = R"cpp(#include "d.h"
 #include "a.h"

diff --git a/clang-tools-extra/test/clang-doc/basic-project.test b/clang-tools-extra/test/clang-doc/basic-project.test
@@ -1,3 +1,6 @@
+// See https://github.com/llvm/llvm-project/issues/97507.
+// UNSUPPORTED: target={{.*}}
+
 // RUN: rm -rf %t && mkdir -p %t/docs %t/build
 // RUN: sed 's|$test_dir|%/S|g' %S/Inputs/basic-project/database_template.json > %t/build/compile_commands.json
 // RUN: clang-doc --format=html --output=%t/docs --executor=all-TUs %t/build/compile_commands.json

diff --git a/clang-tools-extra/test/clang-doc/test-path-abs.cpp b/clang-tools-extra/test/clang-doc/test-path-abs.cpp
@@ -0,0 +1,6 @@
+// RUN: rm -rf %t && mkdir %t
+// RUN: clang-doc --format=html --executor=standalone %s --output=%t
+// RUN: FileCheck %s -input-file=%t/index_json.js  -check-prefix=JSON-INDEX
+// RUN: rm -rf %t
+
+// JSON-INDEX: var RootPath = "{{.*}}test-path-abs.cpp.tmp";
diff --git a/clang-tools-extra/test/clang-tidy/checkers/bugprone/exception-escape.cpp b/clang-tools-extra/test/clang-tidy/checkers/bugprone/exception-escape.cpp
@@ -756,3 +756,21 @@ struct test_implicit_throw {
 };
 
 }}
+
+void pointer_exception_can_not_escape_with_const_void_handler() noexcept {
+  // CHECK-MESSAGES-NOT: :[[@LINE-1]]:6: warning: an exception may be thrown in function 'pointer_exception_can_not_escape_with_const_void_handler' which should not throw exceptions
+  const int value = 42;
+  try {
+    throw &value;
+  } catch (const void *) {
+  }
+}
+
+void pointer_exception_can_not_escape_with_void_handler() noexcept {
+  // CHECK-MESSAGES-NOT: :[[@LINE-1]]:6: warning: an exception may be thrown in function 'pointer_exception_can_not_escape_with_void_handler' which should not throw exceptions
+  int value = 42;
+  try {
+    throw &value;
+  } catch (void *) {
+  }
+}
diff --git a/clang-tools-extra/test/clang-tidy/checkers/misc/const-correctness-values.cpp b/clang-tools-extra/test/clang-tidy/checkers/misc/const-correctness-values.cpp
@@ -3,7 +3,7 @@
 // RUN:     misc-const-correctness.TransformValues: true, \
 // RUN:     misc-const-correctness.WarnPointersAsValues: false, \
 // RUN:     misc-const-correctness.TransformPointersAsValues: false \
-// RUN:   }}" -- -fno-delayed-template-parsing
+// RUN:   }}" -- -fno-delayed-template-parsing -fexceptions
 
 // ------- Provide test samples for primitive builtins ---------
 // - every 'p_*' variable is a 'potential_const_*' variable
@@ -56,6 +56,15 @@ void some_function(double np_arg0, wchar_t np_arg1) {
   np_local6--;
 }
 
+int function_try_block() try {
+  int p_local0 = 0;
+  // CHECK-MESSAGES: [[@LINE-1]]:3: warning: variable 'p_local0' of type 'int' can be declared 'const'
+  // CHECK-FIXES: int const p_local0
+  return p_local0;
+} catch (...) {
+  return 0;
+}
+
 void nested_scopes() {
   int p_local0 = 2;
   // CHECK-MESSAGES: [[@LINE-1]]:3: warning: variable 'p_local0' of type 'int' can be declared 'const'