diff --git a/bolt/docs/CommandLineArgumentReference.md b/bolt/docs/CommandLineArgumentReference.md
diff --git a/bolt/include/bolt/Core/BinaryFunction.h b/bolt/include/bolt/Core/BinaryFunction.h
@@ -930,6 +930,8 @@ class BinaryFunction {
     return const_cast<BinaryFunction *>(this)->getInstructionAtOffset(Offset);
   }
 
+  std::optional<MCInst> disassembleInstructionAtOffset(uint64_t Offset) const;
+
   /// Return offset for the first instruction. If there is data at the
   /// beginning of a function then offset of the first instruction could
   /// be different from 0

diff --git a/bolt/include/bolt/Core/DIEBuilder.h b/bolt/include/bolt/Core/DIEBuilder.h
@@ -129,6 +129,9 @@ class DIEBuilder {
   uint64_t UnitSize{0};
   llvm::DenseSet<uint64_t> AllProcessed;
   DWARF5AcceleratorTable &DebugNamesTable;
+  // Unordered map to handle name collision if output DWO directory is
+  // specified.
+  std::unordered_map<std::string, uint32_t> NameToIndexMap;
 
   /// Returns current state of the DIEBuilder
   State &getState() { return *BuilderState.get(); }
@@ -384,6 +387,17 @@ class DIEBuilder {
   bool deleteValue(DIEValueList *Die, dwarf::Attribute Attribute) {
     return Die->deleteValue(Attribute);
   }
+  /// Updates DWO Name and Compilation directory for Skeleton CU \p Unit.
+  std::string updateDWONameCompDir(DebugStrOffsetsWriter &StrOffstsWriter,
+                                   DebugStrWriter &StrWriter,
+                                   DWARFUnit &SkeletonCU,
+                                   std::optional<StringRef> DwarfOutputPath,
+                                   std::optional<StringRef> DWONameToUse);
+  /// Updates DWO Name and Compilation directory for Type Units.
+  void updateDWONameCompDirForTypes(DebugStrOffsetsWriter &StrOffstsWriter,
+                                    DebugStrWriter &StrWriter, DWARFUnit &Unit,
+                                    std::optional<StringRef> DwarfOutputPath,
+                                    const StringRef DWOName);
 };
 } // namespace bolt
 } // namespace llvm

diff --git a/bolt/include/bolt/Core/DebugData.h b/bolt/include/bolt/Core/DebugData.h
@@ -430,7 +430,7 @@ class DebugAddrWriterDwarf5 : public DebugAddrWriter {
 using DebugStrOffsetsBufferVector = SmallVector<char, 16>;
 class DebugStrOffsetsWriter {
 public:
-  DebugStrOffsetsWriter() {
+  DebugStrOffsetsWriter(BinaryContext &BC) : BC(BC) {
     StrOffsetsBuffer = std::make_unique<DebugStrOffsetsBufferVector>();
     StrOffsetsStream = std::make_unique<raw_svector_ostream>(*StrOffsetsBuffer);
   }
@@ -460,20 +460,27 @@ class DebugStrOffsetsWriter {
     StrOffsets.clear();
   }
 
+  bool isStrOffsetsSectionModified() const {
+    return StrOffsetSectionWasModified;
+  }
+
 private:
   std::unique_ptr<DebugStrOffsetsBufferVector> StrOffsetsBuffer;
   std::unique_ptr<raw_svector_ostream> StrOffsetsStream;
   std::map<uint32_t, uint32_t> IndexToAddressMap;
   SmallVector<uint32_t, 5> StrOffsets;
   std::unordered_map<uint64_t, uint64_t> ProcessedBaseOffsets;
   bool StrOffsetSectionWasModified = false;
+  BinaryContext &BC;
 };
 
 using DebugStrBufferVector = SmallVector<char, 16>;
 class DebugStrWriter {
 public:
   DebugStrWriter() = delete;
-  DebugStrWriter(BinaryContext &BC) : BC(BC) { create(); }
+  DebugStrWriter(DWARFContext &DwCtx, bool IsDWO) : DwCtx(DwCtx), IsDWO(IsDWO) {
+    create();
+  }
   std::unique_ptr<DebugStrBufferVector> releaseBuffer() {
     return std::move(StrBuffer);
   }
@@ -495,7 +502,8 @@ class DebugStrWriter {
   void create();
   std::unique_ptr<DebugStrBufferVector> StrBuffer;
   std::unique_ptr<raw_svector_ostream> StrStream;
-  BinaryContext &BC;
+  DWARFContext &DwCtx;
+  bool IsDWO;
 };
 
 enum class LocWriterKind { DebugLocWriter, DebugLoclistWriter };

diff --git a/bolt/include/bolt/Core/MCPlusBuilder.h b/bolt/include/bolt/Core/MCPlusBuilder.h
@@ -438,8 +438,8 @@ class MCPlusBuilder {
     return false;
   }
 
-  /// Check whether we support inverting this branch
-  virtual bool isUnsupportedBranch(const MCInst &Inst) const { return false; }
+  /// Check whether this conditional branch can be reversed
+  virtual bool isReversibleBranch(const MCInst &Inst) const { return true; }
 
   /// Return true of the instruction is of pseudo kind.
   virtual bool isPseudo(const MCInst &Inst) const {

diff --git a/bolt/include/bolt/Passes/FrameAnalysis.h b/bolt/include/bolt/Passes/FrameAnalysis.h
@@ -170,10 +170,6 @@ class FrameAnalysis {
                      std::unique_ptr<StackPointerTracking>>
       SPTMap;
 
-  /// A vector that stores ids of the allocators that are used in SPT
-  /// computation
-  std::vector<MCPlusBuilder::AllocatorIdTy> SPTAllocatorsId;
-
 public:
   explicit FrameAnalysis(BinaryContext &BC, BinaryFunctionCallGraph &CG);
 

diff --git a/bolt/include/bolt/Passes/IndirectCallPromotion.h b/bolt/include/bolt/Passes/IndirectCallPromotion.h
@@ -104,7 +104,7 @@ class IndirectCallPromotion : public BinaryFunctionPass {
   struct Location {
     MCSymbol *Sym{nullptr};
     uint64_t Addr{0};
-    bool isValid() const { return Sym || (!Sym && Addr != 0); }
+    bool isValid() const { return Sym || Addr != 0; }
     Location() {}
     explicit Location(MCSymbol *Sym) : Sym(Sym) {}
     explicit Location(uint64_t Addr) : Addr(Addr) {}

diff --git a/bolt/include/bolt/Profile/DataAggregator.h b/bolt/include/bolt/Profile/DataAggregator.h
@@ -122,14 +122,14 @@ class DataAggregator : public DataReader {
     uint64_t ExternCount{0};
   };
 
-  struct BranchInfo {
+  struct TakenBranchInfo {
     uint64_t TakenCount{0};
     uint64_t MispredCount{0};
   };
 
   /// Intermediate storage for profile data. We save the results of parsing
   /// and use them later for processing and assigning profile.
-  std::unordered_map<Trace, BranchInfo, TraceHash> BranchLBRs;
+  std::unordered_map<Trace, TakenBranchInfo, TraceHash> BranchLBRs;
   std::unordered_map<Trace, FTInfo, TraceHash> FallthroughLBRs;
   std::vector<AggregatedLBREntry> AggregatedLBRs;
   std::unordered_map<uint64_t, uint64_t> BasicSamples;

diff --git a/bolt/include/bolt/Rewrite/DWARFRewriter.h b/bolt/include/bolt/Rewrite/DWARFRewriter.h
@@ -203,13 +203,16 @@ class DWARFRewriter {
   using OverriddenSectionsMap = std::unordered_map<DWARFSectionKind, StringRef>;
   /// Output .dwo files.
   void writeDWOFiles(DWARFUnit &, const OverriddenSectionsMap &,
-                     const std::string &, DebugLocWriter &);
+                     const std::string &, DebugLocWriter &,
+                     DebugStrOffsetsWriter &, DebugStrWriter &);
   using KnownSectionsEntry = std::pair<MCSection *, DWARFSectionKind>;
   struct DWPState {
     std::unique_ptr<ToolOutputFile> Out;
     std::unique_ptr<BinaryContext> TmpBC;
     std::unique_ptr<MCStreamer> Streamer;
     std::unique_ptr<DWPStringPool> Strings;
+    /// Used to store String sections for .dwo files if they are being modified.
+    std::vector<std::unique_ptr<DebugBufferVector>> StrSections;
     const MCObjectFileInfo *MCOFI = nullptr;
     const DWARFUnitIndex *CUIndex = nullptr;
     std::deque<SmallString<32>> UncompressedSections;
@@ -230,7 +233,8 @@ class DWARFRewriter {
 
   /// add content of dwo to .dwp file.
   void updateDWP(DWARFUnit &, const OverriddenSectionsMap &, const UnitMeta &,
-                 UnitMetaVectorType &, DWPState &, DebugLocWriter &);
+                 UnitMetaVectorType &, DWPState &, DebugLocWriter &,
+                 DebugStrOffsetsWriter &, DebugStrWriter &);
 };
 
 } // namespace bolt

diff --git a/bolt/lib/Core/BinaryBasicBlock.cpp b/bolt/lib/Core/BinaryBasicBlock.cpp
@@ -131,11 +131,10 @@ bool BinaryBasicBlock::validateSuccessorInvariants() {
         break;
       }
       case 2:
-        Valid = (CondBranch &&
-                 (TBB == getConditionalSuccessor(true)->getLabel() &&
-                  ((!UncondBranch && !FBB) ||
-                   (UncondBranch &&
-                    FBB == getConditionalSuccessor(false)->getLabel()))));
+        Valid =
+            CondBranch && TBB == getConditionalSuccessor(true)->getLabel() &&
+            (UncondBranch ? FBB == getConditionalSuccessor(false)->getLabel()
+                          : !FBB);
         break;
       }
     }

diff --git a/bolt/lib/Core/BinaryFunction.cpp b/bolt/lib/Core/BinaryFunction.cpp
@@ -1167,6 +1167,21 @@ void BinaryFunction::handleAArch64IndirectCall(MCInst &Instruction,
   }
 }
 
+std::optional<MCInst>
+BinaryFunction::disassembleInstructionAtOffset(uint64_t Offset) const {
+  assert(CurrentState == State::Empty && "Function should not be disassembled");
+  assert(Offset < MaxSize && "Invalid offset");
+  ErrorOr<ArrayRef<unsigned char>> FunctionData = getData();
+  assert(FunctionData && "Cannot get function as data");
+  MCInst Instr;
+  uint64_t InstrSize = 0;
+  const uint64_t InstrAddress = getAddress() + Offset;
+  if (BC.DisAsm->getInstruction(Instr, InstrSize, FunctionData->slice(Offset),
+                                InstrAddress, nulls()))
+    return Instr;
+  return std::nullopt;
+}
+
 Error BinaryFunction::disassemble() {
   NamedRegionTimer T("disassemble", "Disassemble function", "buildfuncs",
                      "Build Binary Functions", opts::TimeBuild);
@@ -1269,7 +1284,7 @@ Error BinaryFunction::disassemble() {
         const bool IsCondBranch = MIB->isConditionalBranch(Instruction);
         MCSymbol *TargetSymbol = nullptr;
 
-        if (BC.MIB->isUnsupportedBranch(Instruction)) {
+        if (!BC.MIB->isReversibleBranch(Instruction)) {
           setIgnored();
           if (BinaryFunction *TargetFunc =
                   BC.getBinaryFunctionContainingAddress(TargetAddress))
@@ -3237,12 +3252,9 @@ bool BinaryFunction::validateCFG() const {
   if (CurrentState == State::CFG_Finalized)
     return true;
 
-  bool Valid = true;
   for (BinaryBasicBlock *BB : BasicBlocks)
-    Valid &= BB->validateSuccessorInvariants();
-
-  if (!Valid)
-    return Valid;
+    if (!BB->validateSuccessorInvariants())
+      return false;
 
   // Make sure all blocks in CFG are valid.
   auto validateBlock = [this](const BinaryBasicBlock *BB, StringRef Desc) {
@@ -3311,7 +3323,7 @@ bool BinaryFunction::validateCFG() const {
     }
   }
 
-  return Valid;
+  return true;
 }
 
 void BinaryFunction::fixBranches() {
@@ -3369,7 +3381,7 @@ void BinaryFunction::fixBranches() {
 
       // Reverse branch condition and swap successors.
       auto swapSuccessors = [&]() {
-        if (MIB->isUnsupportedBranch(*CondBranch)) {
+        if (!MIB->isReversibleBranch(*CondBranch)) {
           if (opts::Verbosity) {
             BC.outs() << "BOLT-INFO: unable to swap successors in " << *this
                       << '\n';

diff --git a/bolt/lib/Core/DIEBuilder.cpp b/bolt/lib/Core/DIEBuilder.cpp
@@ -22,6 +22,7 @@
 #include "llvm/Support/Casting.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/ErrorHandling.h"
+#include "llvm/Support/FileSystem.h"
 #include "llvm/Support/LEB128.h"
 
 #include <algorithm>
@@ -41,6 +42,90 @@ extern cl::opt<unsigned> Verbosity;
 namespace llvm {
 namespace bolt {
 
+/// Returns DWO Name to be used to update DW_AT_dwo_name/DW_AT_GNU_dwo_name
+/// either in CU or TU unit die. Handles case where user specifies output DWO
+/// directory, and there are duplicate names. Assumes DWO ID is unique.
+static std::string
+getDWOName(llvm::DWARFUnit &CU,
+           std::unordered_map<std::string, uint32_t> &NameToIndexMap,
+           std::optional<StringRef> &DwarfOutputPath) {
+  assert(CU.getDWOId() && "DWO ID not found.");
+  std::string DWOName = dwarf::toString(
+      CU.getUnitDIE().find({dwarf::DW_AT_dwo_name, dwarf::DW_AT_GNU_dwo_name}),
+      "");
+  assert(!DWOName.empty() &&
+         "DW_AT_dwo_name/DW_AT_GNU_dwo_name does not exist.");
+  if (DwarfOutputPath) {
+    DWOName = std::string(sys::path::filename(DWOName));
+    auto Iter = NameToIndexMap.find(DWOName);
+    if (Iter == NameToIndexMap.end())
+      Iter = NameToIndexMap.insert({DWOName, 0}).first;
+    DWOName.append(std::to_string(Iter->second));
+    ++Iter->second;
+  }
+  DWOName.append(".dwo");
+  return DWOName;
+}
+
+/// Adds a \p Str to .debug_str section.
+/// Uses \p AttrInfoVal to either update entry in a DIE for legacy DWARF using
+/// \p DebugInfoPatcher, or for DWARF5 update an index in .debug_str_offsets
+/// for this contribution of \p Unit.
+static void addStringHelper(DebugStrOffsetsWriter &StrOffstsWriter,
+                            DebugStrWriter &StrWriter, DIEBuilder &DIEBldr,
+                            DIE &Die, const DWARFUnit &Unit,
+                            DIEValue &DIEAttrInfo, StringRef Str) {
+  uint32_t NewOffset = StrWriter.addString(Str);
+  if (Unit.getVersion() >= 5) {
+    StrOffstsWriter.updateAddressMap(DIEAttrInfo.getDIEInteger().getValue(),
+                                     NewOffset);
+    return;
+  }
+  DIEBldr.replaceValue(&Die, DIEAttrInfo.getAttribute(), DIEAttrInfo.getForm(),
+                       DIEInteger(NewOffset));
+}
+
+std::string DIEBuilder::updateDWONameCompDir(
+    DebugStrOffsetsWriter &StrOffstsWriter, DebugStrWriter &StrWriter,
+    DWARFUnit &SkeletonCU, std::optional<StringRef> DwarfOutputPath,
+    std::optional<StringRef> DWONameToUse) {
+  DIE &UnitDIE = *getUnitDIEbyUnit(SkeletonCU);
+  DIEValue DWONameAttrInfo = UnitDIE.findAttribute(dwarf::DW_AT_dwo_name);
+  if (!DWONameAttrInfo)
+    DWONameAttrInfo = UnitDIE.findAttribute(dwarf::DW_AT_GNU_dwo_name);
+  if (!DWONameAttrInfo)
+    return "";
+  std::string ObjectName;
+  if (DWONameToUse)
+    ObjectName = *DWONameToUse;
+  else
+    ObjectName = getDWOName(SkeletonCU, NameToIndexMap, DwarfOutputPath);
+  addStringHelper(StrOffstsWriter, StrWriter, *this, UnitDIE, SkeletonCU,
+                  DWONameAttrInfo, ObjectName);
+
+  DIEValue CompDirAttrInfo = UnitDIE.findAttribute(dwarf::DW_AT_comp_dir);
+  assert(CompDirAttrInfo && "DW_AT_comp_dir is not in Skeleton CU.");
+
+  if (DwarfOutputPath) {
+    if (!sys::fs::exists(*DwarfOutputPath))
+      sys::fs::create_directory(*DwarfOutputPath);
+    addStringHelper(StrOffstsWriter, StrWriter, *this, UnitDIE, SkeletonCU,
+                    CompDirAttrInfo, *DwarfOutputPath);
+  }
+  return ObjectName;
+}
+
+void DIEBuilder::updateDWONameCompDirForTypes(
+    DebugStrOffsetsWriter &StrOffstsWriter, DebugStrWriter &StrWriter,
+    DWARFUnit &Unit, std::optional<StringRef> DwarfOutputPath,
+    const StringRef DWOName) {
+  for (DWARFUnit *DU : getState().DWARF5TUVector)
+    updateDWONameCompDir(StrOffstsWriter, StrWriter, *DU, DwarfOutputPath,
+                         DWOName);
+  if (StrOffstsWriter.isStrOffsetsSectionModified())
+    StrOffstsWriter.finalizeSection(Unit, *this);
+}
+
 void DIEBuilder::updateReferences() {
   for (auto &[SrcDIEInfo, ReferenceInfo] : getState().AddrReferences) {
     DIEInfo *DstDIEInfo = ReferenceInfo.Dst;

diff --git a/bolt/lib/Core/DebugData.cpp b/bolt/lib/Core/DebugData.cpp
@@ -31,6 +31,7 @@
 #include <cstdint>
 #include <functional>
 #include <memory>
+#include <optional>
 #include <unordered_map>
 #include <vector>
 
@@ -867,10 +868,17 @@ void DebugStrOffsetsWriter::finalizeSection(DWARFUnit &Unit,
                                             DIEBuilder &DIEBldr) {
   std::optional<AttrInfo> AttrVal =
       findAttributeInfo(Unit.getUnitDIE(), dwarf::DW_AT_str_offsets_base);
-  if (!AttrVal)
+  if (!AttrVal && !Unit.isDWOUnit())
     return;
-  std::optional<uint64_t> Val = AttrVal->V.getAsSectionOffset();
-  assert(Val && "DW_AT_str_offsets_base Value not present.");
+  std::optional<uint64_t> Val = std::nullopt;
+  if (AttrVal) {
+    Val = AttrVal->V.getAsSectionOffset();
+  } else {
+    if (!Unit.isDWOUnit())
+      BC.errs() << "BOLT-WARNING: [internal-dwarf-error]: "
+                   "DW_AT_str_offsets_base Value not present\n";
+    Val = 0;
+  }
   DIE &Die = *DIEBldr.getUnitDIEbyUnit(Unit);
   DIEValue StrListBaseAttrInfo =
       Die.findAttribute(dwarf::DW_AT_str_offsets_base);
@@ -915,7 +923,11 @@ void DebugStrWriter::create() {
 }
 
 void DebugStrWriter::initialize() {
-  auto StrSection = BC.DwCtx->getDWARFObj().getStrSection();
+  StringRef StrSection;
+  if (IsDWO)
+    StrSection = DwCtx.getDWARFObj().getStrDWOSection();
+  else
+    StrSection = DwCtx.getDWARFObj().getStrSection();
   (*StrStream) << StrSection;
 }
 

diff --git a/bolt/lib/Core/ParallelUtilities.cpp b/bolt/lib/Core/ParallelUtilities.cpp
@@ -231,12 +231,7 @@ void runOnEachFunctionWithUniqueAllocId(
     }
   }
 
-  if (!BC.MIB->checkAllocatorExists(AllocId)) {
-    MCPlusBuilder::AllocatorIdTy Id =
-        BC.MIB->initializeNewAnnotationAllocator();
-    (void)Id;
-    assert(AllocId == Id && "unexpected allocator id created");
-  }
+  EnsureAllocatorExists(AllocId);
 
   Pool.async(runBlock, BlockBegin, BC.getBinaryFunctions().end(), AllocId);
   Lock.unlock();

diff --git a/bolt/lib/Passes/BinaryPasses.cpp b/bolt/lib/Passes/BinaryPasses.cpp
@@ -715,6 +715,9 @@ static uint64_t fixDoubleJumps(BinaryFunction &Function, bool MarkInvalid) {
           Pred->removeSuccessor(&BB);
           Pred->eraseInstruction(Pred->findInstruction(Branch));
           Pred->addTailCallInstruction(SuccSym);
+          MCInst *TailCall = Pred->getLastNonPseudoInstr();
+          assert(TailCall);
+          MIB->setOffset(*TailCall, BB.getOffset());
         } else {
           return false;
         }
@@ -910,6 +913,11 @@ uint64_t SimplifyConditionalTailCalls::fixTailCalls(BinaryFunction &BF) {
       auto &CTCAnnotation =
           MIB->getOrCreateAnnotationAs<uint64_t>(*CondBranch, "CTCTakenCount");
       CTCAnnotation = CTCTakenFreq;
+      // Preserve Offset annotation, used in BAT.
+      // Instr is a direct tail call instruction that was created when CTCs are
+      // first expanded, and has the original CTC offset set.
+      if (std::optional<uint32_t> Offset = MIB->getOffset(*Instr))
+        MIB->setOffset(*CondBranch, *Offset);
 
       // Remove the unused successor which may be eliminated later
       // if there are no other users.

diff --git a/bolt/lib/Passes/FrameAnalysis.cpp b/bolt/lib/Passes/FrameAnalysis.cpp
@@ -561,11 +561,6 @@ FrameAnalysis::FrameAnalysis(BinaryContext &BC, BinaryFunctionCallGraph &CG)
     NamedRegionTimer T1("clearspt", "clear spt", "FA", "FA breakdown",
                         opts::TimeFA);
     clearSPTMap();
-
-    // Clean up memory allocated for annotation values
-    if (!opts::NoThreads)
-      for (MCPlusBuilder::AllocatorIdTy Id : SPTAllocatorsId)
-        BC.MIB->freeValuesAllocator(Id);
   }
 }
 

diff --git a/bolt/lib/Passes/Instrumentation.cpp b/bolt/lib/Passes/Instrumentation.cpp
@@ -480,7 +480,7 @@ void Instrumentation::instrumentFunction(BinaryFunction &Function,
       else if (BC.MIB->isUnconditionalBranch(Inst))
         HasUnconditionalBranch = true;
       else if ((!BC.MIB->isCall(Inst) && !BC.MIB->isConditionalBranch(Inst)) ||
-               BC.MIB->isUnsupportedBranch(Inst))
+               !BC.MIB->isReversibleBranch(Inst))
         continue;
 
       const uint32_t FromOffset = *BC.MIB->getOffset(Inst);

diff --git a/bolt/lib/Profile/DataAggregator.cpp b/bolt/lib/Profile/DataAggregator.cpp
@@ -23,6 +23,7 @@
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/ScopeExit.h"
 #include "llvm/Support/CommandLine.h"
+#include "llvm/Support/Compiler.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/Errc.h"
 #include "llvm/Support/FileSystem.h"
@@ -773,9 +774,19 @@ bool DataAggregator::doInterBranch(BinaryFunction *FromFunc,
 
 bool DataAggregator::doBranch(uint64_t From, uint64_t To, uint64_t Count,
                               uint64_t Mispreds) {
+  bool IsReturn = false;
   auto handleAddress = [&](uint64_t &Addr, bool IsFrom) -> BinaryFunction * {
     if (BinaryFunction *Func = getBinaryFunctionContainingAddress(Addr)) {
       Addr -= Func->getAddress();
+      if (IsFrom) {
+        auto checkReturn = [&](auto MaybeInst) {
+          IsReturn = MaybeInst && BC->MIB->isReturn(*MaybeInst);
+        };
+        if (Func->hasInstructions())
+          checkReturn(Func->getInstructionAtOffset(Addr));
+        else
+          checkReturn(Func->disassembleInstructionAtOffset(Addr));
+      }
 
       if (BAT)
         Addr = BAT->translate(Func->getAddress(), Addr, IsFrom);
@@ -792,6 +803,9 @@ bool DataAggregator::doBranch(uint64_t From, uint64_t To, uint64_t Count,
   };
 
   BinaryFunction *FromFunc = handleAddress(From, /*IsFrom=*/true);
+  // Ignore returns.
+  if (IsReturn)
+    return true;
   BinaryFunction *ToFunc = handleAddress(To, /*IsFrom=*/false);
   if (!FromFunc && !ToFunc)
     return false;
@@ -1450,7 +1464,7 @@ uint64_t DataAggregator::parseLBRSample(const PerfBranchSample &Sample,
     uint64_t To = getBinaryFunctionContainingAddress(LBR.To) ? LBR.To : 0;
     if (!From && !To)
       continue;
-    BranchInfo &Info = BranchLBRs[Trace(From, To)];
+    TakenBranchInfo &Info = BranchLBRs[Trace(From, To)];
     ++Info.TakenCount;
     Info.MispredCount += LBR.Mispred;
   }
@@ -1595,7 +1609,7 @@ void DataAggregator::processBranchEvents() {
 
   for (const auto &AggrLBR : BranchLBRs) {
     const Trace &Loc = AggrLBR.first;
-    const BranchInfo &Info = AggrLBR.second;
+    const TakenBranchInfo &Info = AggrLBR.second;
     doBranch(Loc.From, Loc.To, Info.TakenCount, Info.MispredCount);
   }
 }
@@ -1986,7 +2000,7 @@ std::error_code DataAggregator::parseMMapEvents() {
     std::pair<StringRef, MMapInfo> FileMMapInfo = FileMMapInfoRes.get();
     if (FileMMapInfo.second.PID == -1)
       continue;
-    if (FileMMapInfo.first.equals("(deleted)"))
+    if (FileMMapInfo.first == "(deleted)")
       continue;
 
     // Consider only the first mapping of the file for any given PID
@@ -2239,13 +2253,13 @@ DataAggregator::writeAggregatedFile(StringRef OutputFilename) const {
   } else {
     for (const auto &KV : NamesToBranches) {
       const FuncBranchData &FBD = KV.second;
-      for (const llvm::bolt::BranchInfo &BI : FBD.Data) {
+      for (const BranchInfo &BI : FBD.Data) {
         writeLocation(BI.From);
         writeLocation(BI.To);
         OutFile << BI.Mispreds << " " << BI.Branches << "\n";
         ++BranchValues;
       }
-      for (const llvm::bolt::BranchInfo &BI : FBD.EntryData) {
+      for (const BranchInfo &BI : FBD.EntryData) {
         // Do not output if source is a known symbol, since this was already
         // accounted for in the source function
         if (BI.From.IsSymbol)
@@ -2326,7 +2340,7 @@ std::error_code DataAggregator::writeBATYAML(BinaryContext &BC,
         continue;
       BinaryFunction *BF = BC.getBinaryFunctionAtAddress(FuncAddress);
       assert(BF);
-      YamlBF.Name = FuncName.str();
+      YamlBF.Name = getLocationName(*BF);
       YamlBF.Id = BF->getFunctionNumber();
       YamlBF.Hash = BAT->getBFHash(FuncAddress);
       YamlBF.ExecCount = BF->getKnownExecutionCount();
@@ -2341,54 +2355,48 @@ std::error_code DataAggregator::writeBATYAML(BinaryContext &BC,
       for (auto BI = BlockMap.begin(), BE = BlockMap.end(); BI != BE; ++BI)
         YamlBF.Blocks[BI->second.getBBIndex()].Hash = BI->second.getBBHash();
 
-      auto getSuccessorInfo = [&](uint32_t SuccOffset, unsigned SuccDataIdx) {
-        const llvm::bolt::BranchInfo &BI = Branches.Data.at(SuccDataIdx);
-        yaml::bolt::SuccessorInfo SI;
-        SI.Index = BlockMap.getBBIndex(SuccOffset);
-        SI.Count = BI.Branches;
-        SI.Mispreds = BI.Mispreds;
-        return SI;
-      };
-
-      auto getCallSiteInfo = [&](Location CallToLoc, unsigned CallToIdx,
-                                 uint32_t Offset) {
-        const llvm::bolt::BranchInfo &BI = Branches.Data.at(CallToIdx);
-        yaml::bolt::CallSiteInfo CSI;
-        CSI.DestId = 0; // designated for unknown functions
-        CSI.EntryDiscriminator = 0;
-        CSI.Count = BI.Branches;
-        CSI.Mispreds = BI.Mispreds;
-        CSI.Offset = Offset;
-        if (BinaryData *BD = BC.getBinaryDataByName(CallToLoc.Name))
-          YAMLProfileWriter::setCSIDestination(BC, CSI, BD->getSymbol(), BAT,
-                                               CallToLoc.Offset);
-        return CSI;
+      // Lookup containing basic block offset and index
+      auto getBlock = [&BlockMap](uint32_t Offset) {
+        auto BlockIt = BlockMap.upper_bound(Offset);
+        if (LLVM_UNLIKELY(BlockIt == BlockMap.begin())) {
+          errs() << "BOLT-ERROR: invalid BAT section\n";
+          exit(1);
+        }
+        --BlockIt;
+        return std::pair(BlockIt->first, BlockIt->second.getBBIndex());
       };
 
-      for (const auto &[FromOffset, SuccKV] : Branches.IntraIndex) {
-        if (!BlockMap.isInputBlock(FromOffset))
-          continue;
-        const unsigned Index = BlockMap.getBBIndex(FromOffset);
-        yaml::bolt::BinaryBasicBlockProfile &YamlBB = YamlBF.Blocks[Index];
-        for (const auto &[SuccOffset, SuccDataIdx] : SuccKV)
-          if (BlockMap.isInputBlock(SuccOffset))
-            YamlBB.Successors.emplace_back(
-                getSuccessorInfo(SuccOffset, SuccDataIdx));
+      for (const BranchInfo &BI : Branches.Data) {
+        using namespace yaml::bolt;
+        const auto &[BlockOffset, BlockIndex] = getBlock(BI.From.Offset);
+        BinaryBasicBlockProfile &YamlBB = YamlBF.Blocks[BlockIndex];
+        if (BI.To.IsSymbol && BI.To.Name == BI.From.Name && BI.To.Offset != 0) {
+          // Internal branch
+          const unsigned SuccIndex = getBlock(BI.To.Offset).second;
+          auto &SI = YamlBB.Successors.emplace_back(SuccessorInfo{SuccIndex});
+          SI.Count = BI.Branches;
+          SI.Mispreds = BI.Mispreds;
+        } else {
+          // Call
+          const uint32_t Offset = BI.From.Offset - BlockOffset;
+          auto &CSI = YamlBB.CallSites.emplace_back(CallSiteInfo{Offset});
+          CSI.Count = BI.Branches;
+          CSI.Mispreds = BI.Mispreds;
+          if (const BinaryData *BD = BC.getBinaryDataByName(BI.To.Name))
+            YAMLProfileWriter::setCSIDestination(BC, CSI, BD->getSymbol(), BAT,
+                                                 BI.To.Offset);
+        }
       }
-      for (const auto &[FromOffset, CallTo] : Branches.InterIndex) {
-        auto BlockIt = BlockMap.upper_bound(FromOffset);
-        --BlockIt;
-        const unsigned BlockOffset = BlockIt->first;
-        const unsigned BlockIndex = BlockIt->second.getBBIndex();
-        yaml::bolt::BinaryBasicBlockProfile &YamlBB = YamlBF.Blocks[BlockIndex];
-        const uint32_t Offset = FromOffset - BlockOffset;
-        for (const auto &[CallToLoc, CallToIdx] : CallTo)
-          YamlBB.CallSites.emplace_back(
-              getCallSiteInfo(CallToLoc, CallToIdx, Offset));
-        llvm::sort(YamlBB.CallSites, [](yaml::bolt::CallSiteInfo &A,
-                                        yaml::bolt::CallSiteInfo &B) {
-          return A.Offset < B.Offset;
-        });
+      // Set entry counts, similar to DataReader::readProfile.
+      for (const BranchInfo &BI : Branches.EntryData) {
+        if (!BlockMap.isInputBlock(BI.To.Offset)) {
+          if (opts::Verbosity >= 1)
+            errs() << "BOLT-WARNING: Unexpected EntryData in " << FuncName
+                   << " at 0x" << Twine::utohexstr(BI.To.Offset) << '\n';
+          continue;
+        }
+        const unsigned BlockIndex = BlockMap.getBBIndex(BI.To.Offset);
+        YamlBF.Blocks[BlockIndex].ExecCount += BI.Branches;
       }
       // Drop blocks without a hash, won't be useful for stale matching.
       llvm::erase_if(YamlBF.Blocks,

diff --git a/bolt/lib/Profile/DataReader.cpp b/bolt/lib/Profile/DataReader.cpp
@@ -775,6 +775,7 @@ bool DataReader::recordBranch(BinaryFunction &BF, uint64_t From, uint64_t To,
     if (collectedInBoltedBinary() && FromBB == ToBB)
       return true;
 
+    // Allow passthrough blocks.
     BinaryBasicBlock *FTSuccessor = FromBB->getConditionalSuccessor(false);
     if (FTSuccessor && FTSuccessor->succ_size() == 1 &&
         FTSuccessor->getSuccessor(ToBB->getLabel())) {
@@ -1205,8 +1206,7 @@ std::error_code DataReader::parse() {
 
     // Add entry data for branches to another function or branches
     // to entry points (including recursive calls)
-    if (BI.To.IsSymbol &&
-        (!BI.From.Name.equals(BI.To.Name) || BI.To.Offset == 0)) {
+    if (BI.To.IsSymbol && (BI.From.Name != BI.To.Name || BI.To.Offset == 0)) {
       I = GetOrCreateFuncEntry(BI.To.Name);
       I->second.EntryData.emplace_back(std::move(BI));
     }

diff --git a/bolt/lib/Profile/YAMLProfileReader.cpp b/bolt/lib/Profile/YAMLProfileReader.cpp
@@ -218,17 +218,28 @@ bool YAMLProfileReader::parseFunctionProfile(
         continue;
       }
 
-      BinaryBasicBlock &SuccessorBB = *Order[YamlSI.Index];
-      if (!BB.getSuccessor(SuccessorBB.getLabel())) {
-        if (opts::Verbosity >= 1)
-          errs() << "BOLT-WARNING: no successor for block " << BB.getName()
-                 << " that matches index " << YamlSI.Index << " or block "
-                 << SuccessorBB.getName() << '\n';
-        ++MismatchedEdges;
-        continue;
+      BinaryBasicBlock *ToBB = Order[YamlSI.Index];
+      if (!BB.getSuccessor(ToBB->getLabel())) {
+        // Allow passthrough blocks.
+        BinaryBasicBlock *FTSuccessor = BB.getConditionalSuccessor(false);
+        if (FTSuccessor && FTSuccessor->succ_size() == 1 &&
+            FTSuccessor->getSuccessor(ToBB->getLabel())) {
+          BinaryBasicBlock::BinaryBranchInfo &FTBI =
+              FTSuccessor->getBranchInfo(*ToBB);
+          FTBI.Count += YamlSI.Count;
+          FTBI.MispredictedCount += YamlSI.Mispreds;
+          ToBB = FTSuccessor;
+        } else {
+          if (opts::Verbosity >= 1)
+            errs() << "BOLT-WARNING: no successor for block " << BB.getName()
+                   << " that matches index " << YamlSI.Index << " or block "
+                   << ToBB->getName() << '\n';
+          ++MismatchedEdges;
+          continue;
+        }
       }
 
-      BinaryBasicBlock::BinaryBranchInfo &BI = BB.getBranchInfo(SuccessorBB);
+      BinaryBasicBlock::BinaryBranchInfo &BI = BB.getBranchInfo(*ToBB);
       BI.Count += YamlSI.Count;
       BI.MispredictedCount += YamlSI.Mispreds;
     }

diff --git a/bolt/lib/Rewrite/CMakeLists.txt b/bolt/lib/Rewrite/CMakeLists.txt
@@ -1,4 +1,5 @@
 set(LLVM_LINK_COMPONENTS
+  Core
   DebugInfoDWARF
   DWP
   JITLink

diff --git a/bolt/lib/Rewrite/DWARFRewriter.cpp b/bolt/lib/Rewrite/DWARFRewriter.cpp
@@ -458,32 +458,6 @@ static std::optional<uint64_t> getAsAddress(const DWARFUnit &DU,
   return std::nullopt;
 }
 
-/// Returns DWO Name to be used. Handles case where user specifies output DWO
-/// directory, and there are duplicate names. Assumes DWO ID is unique.
-static std::string
-getDWOName(llvm::DWARFUnit &CU,
-           std::unordered_map<std::string, uint32_t> &NameToIndexMap) {
-  std::optional<uint64_t> DWOId = CU.getDWOId();
-  assert(DWOId && "DWO ID not found.");
-  (void)DWOId;
-
-  std::string DWOName = dwarf::toString(
-      CU.getUnitDIE().find({dwarf::DW_AT_dwo_name, dwarf::DW_AT_GNU_dwo_name}),
-      "");
-  assert(!DWOName.empty() &&
-         "DW_AT_dwo_name/DW_AT_GNU_dwo_name does not exists.");
-  if (!opts::DwarfOutputPath.empty()) {
-    DWOName = std::string(sys::path::filename(DWOName));
-    auto Iter = NameToIndexMap.find(DWOName);
-    if (Iter == NameToIndexMap.end())
-      Iter = NameToIndexMap.insert({DWOName, 0}).first;
-    DWOName.append(std::to_string(Iter->second));
-    ++Iter->second;
-  }
-  DWOName.append(".dwo");
-  return DWOName;
-}
-
 static std::unique_ptr<DIEStreamer>
 createDIEStreamer(const Triple &TheTriple, raw_pwrite_stream &OutFile,
                   StringRef Swift5ReflectionSegmentName, DIEBuilder &DIEBldr,
@@ -515,7 +489,9 @@ static void emitDWOBuilder(const std::string &DWOName,
                            DIEBuilder &DWODIEBuilder, DWARFRewriter &Rewriter,
                            DWARFUnit &SplitCU, DWARFUnit &CU,
                            DWARFRewriter::DWPState &State,
-                           DebugLocWriter &LocWriter) {
+                           DebugLocWriter &LocWriter,
+                           DebugStrOffsetsWriter &StrOffstsWriter,
+                           DebugStrWriter &StrWriter) {
   // Populate debug_info and debug_abbrev for current dwo into StringRef.
   DWODIEBuilder.generateAbbrevs();
   DWODIEBuilder.finish();
@@ -577,54 +553,10 @@ static void emitDWOBuilder(const std::string &DWOName,
   }
   if (opts::WriteDWP)
     Rewriter.updateDWP(CU, OverriddenSections, CUMI, TUMetaVector, State,
-                       LocWriter);
+                       LocWriter, StrOffstsWriter, StrWriter);
   else
-    Rewriter.writeDWOFiles(CU, OverriddenSections, DWOName, LocWriter);
-}
-
-/// Adds a \p Str to .debug_str section.
-/// Uses \p AttrInfoVal to either update entry in a DIE for legacy DWARF using
-/// \p DebugInfoPatcher, or for DWARF5 update an index in .debug_str_offsets
-/// for this contribution of \p Unit.
-static void addStringHelper(DebugStrOffsetsWriter &StrOffstsWriter,
-                            DebugStrWriter &StrWriter, DIEBuilder &DIEBldr,
-                            DIE &Die, const DWARFUnit &Unit,
-                            DIEValue &DIEAttrInfo, StringRef Str) {
-  uint32_t NewOffset = StrWriter.addString(Str);
-  if (Unit.getVersion() >= 5) {
-    StrOffstsWriter.updateAddressMap(DIEAttrInfo.getDIEInteger().getValue(),
-                                     NewOffset);
-    return;
-  }
-  DIEBldr.replaceValue(&Die, DIEAttrInfo.getAttribute(), DIEAttrInfo.getForm(),
-                       DIEInteger(NewOffset));
-}
-
-static std::string
-updateDWONameCompDir(DebugStrOffsetsWriter &StrOffstsWriter,
-                     DebugStrWriter &StrWriter,
-                     std::unordered_map<std::string, uint32_t> &NameToIndexMap,
-                     DWARFUnit &Unit, DIEBuilder &DIEBldr, DIE &UnitDIE) {
-  DIEValue DWONameAttrInfo = UnitDIE.findAttribute(dwarf::DW_AT_dwo_name);
-  if (!DWONameAttrInfo)
-    DWONameAttrInfo = UnitDIE.findAttribute(dwarf::DW_AT_GNU_dwo_name);
-  assert(DWONameAttrInfo && "DW_AT_dwo_name is not in Skeleton CU.");
-  std::string ObjectName;
-
-  ObjectName = getDWOName(Unit, NameToIndexMap);
-  addStringHelper(StrOffstsWriter, StrWriter, DIEBldr, UnitDIE, Unit,
-                  DWONameAttrInfo, ObjectName.c_str());
-
-  DIEValue CompDirAttrInfo = UnitDIE.findAttribute(dwarf::DW_AT_comp_dir);
-  assert(CompDirAttrInfo && "DW_AT_comp_dir is not in Skeleton CU.");
-
-  if (!opts::DwarfOutputPath.empty()) {
-    if (!sys::fs::exists(opts::DwarfOutputPath))
-      sys::fs::create_directory(opts::DwarfOutputPath);
-    addStringHelper(StrOffstsWriter, StrWriter, DIEBldr, UnitDIE, Unit,
-                    CompDirAttrInfo, opts::DwarfOutputPath.c_str());
-  }
-  return ObjectName;
+    Rewriter.writeDWOFiles(CU, OverriddenSections, DWOName, LocWriter,
+                           StrOffstsWriter, StrWriter);
 }
 
 using DWARFUnitVec = std::vector<DWARFUnit *>;
@@ -673,9 +605,8 @@ void DWARFRewriter::updateDebugInfo() {
     return;
 
   ARangesSectionWriter = std::make_unique<DebugARangesSectionWriter>();
-  StrWriter = std::make_unique<DebugStrWriter>(BC);
-
-  StrOffstsWriter = std::make_unique<DebugStrOffsetsWriter>();
+  StrWriter = std::make_unique<DebugStrWriter>(*BC.DwCtx, false);
+  StrOffstsWriter = std::make_unique<DebugStrOffsetsWriter>(BC);
 
   if (!opts::DeterministicDebugInfo) {
     opts::DeterministicDebugInfo = true;
@@ -720,10 +651,6 @@ void DWARFRewriter::updateDebugInfo() {
     return LocListWritersByCU[CUIndex++].get();
   };
 
-  // Unordered maps to handle name collision if output DWO directory is
-  // specified.
-  std::unordered_map<std::string, uint32_t> NameToIndexMap;
-
   DWARF5AcceleratorTable DebugNamesTable(opts::CreateDebugNames, BC,
                                          *StrWriter);
   DWPState State;
@@ -747,13 +674,20 @@ void DWARFRewriter::updateDebugInfo() {
                                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 = updateDWONameCompDir(*StrOffstsWriter, *StrWriter,
-                                       NameToIndexMap, *Unit, *DIEBlder,
-                                       *DIEBlder->getUnitDIEbyUnit(*Unit));
+        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);
       DebugRangesSectionWriter *TempRangesSectionWriter = RangesSectionWriter;
       if (Unit->getVersion() >= 5) {
@@ -771,7 +705,7 @@ void DWARFRewriter::updateDebugInfo() {
         TempRangesSectionWriter->finalizeSection();
 
       emitDWOBuilder(DWOName, DWODIEBuilder, *this, **SplitCU, *Unit, State,
-                     DebugLocDWoWriter);
+                     DebugLocDWoWriter, DWOStrOffstsWriter, DWOStrWriter);
     }
 
     if (Unit->getVersion() >= 5) {
@@ -1550,7 +1484,7 @@ CUOffsetMap DWARFRewriter::finalizeTypeSections(DIEBuilder &DIEBlder,
   for (const SectionRef &Section : Obj->sections()) {
     StringRef Contents = cantFail(Section.getContents());
     StringRef Name = cantFail(Section.getName());
-    if (Name.equals(".debug_types"))
+    if (Name == ".debug_types")
       BC.registerOrUpdateNoteSection(".debug_types", copyByteArray(Contents),
                                      Contents.size());
   }
@@ -1633,10 +1567,10 @@ void DWARFRewriter::finalizeDebugSections(
   for (const SectionRef &Secs : Obj->sections()) {
     StringRef Contents = cantFail(Secs.getContents());
     StringRef Name = cantFail(Secs.getName());
-    if (Name.equals(".debug_abbrev")) {
+    if (Name == ".debug_abbrev") {
       BC.registerOrUpdateNoteSection(".debug_abbrev", copyByteArray(Contents),
                                      Contents.size());
-    } else if (Name.equals(".debug_info")) {
+    } else if (Name == ".debug_info") {
       BC.registerOrUpdateNoteSection(".debug_info", copyByteArray(Contents),
                                      Contents.size());
     }
@@ -1736,6 +1670,7 @@ std::optional<StringRef> updateDebugData(
     const DWARFUnitIndex::Entry *CUDWOEntry, uint64_t DWOId,
     std::unique_ptr<DebugBufferVector> &OutputBuffer,
     DebugRangeListsSectionWriter *RangeListsWriter, DebugLocWriter &LocWriter,
+    DebugStrOffsetsWriter &StrOffstsWriter, DebugStrWriter &StrWriter,
     const llvm::bolt::DWARFRewriter::OverriddenSectionsMap &OverridenSections) {
 
   using DWOSectionContribution =
@@ -1771,9 +1706,14 @@ std::optional<StringRef> updateDebugData(
   };
   switch (SectionIter->second.second) {
   default: {
-    if (!SectionName.equals("debug_str.dwo"))
+    if (SectionName != "debug_str.dwo")
       errs() << "BOLT-WARNING: unsupported debug section: " << SectionName
              << "\n";
+    if (StrWriter.isInitialized()) {
+      OutputBuffer = StrWriter.releaseBuffer();
+      return StringRef(reinterpret_cast<const char *>(OutputBuffer->data()),
+                       OutputBuffer->size());
+    }
     return SectionContents;
   }
   case DWARFSectionKind::DW_SECT_INFO: {
@@ -1783,6 +1723,11 @@ std::optional<StringRef> updateDebugData(
     return getOverridenSection(DWARFSectionKind::DW_SECT_EXT_TYPES);
   }
   case DWARFSectionKind::DW_SECT_STR_OFFSETS: {
+    if (StrOffstsWriter.isFinalized()) {
+      OutputBuffer = StrOffstsWriter.releaseBuffer();
+      return StringRef(reinterpret_cast<const char *>(OutputBuffer->data()),
+                       OutputBuffer->size());
+    }
     return getSliceData(CUDWOEntry, SectionContents,
                         DWARFSectionKind::DW_SECT_STR_OFFSETS, DWPOffset);
   }
@@ -1884,7 +1829,9 @@ void DWARFRewriter::updateDWP(DWARFUnit &CU,
                               const OverriddenSectionsMap &OverridenSections,
                               const DWARFRewriter::UnitMeta &CUMI,
                               DWARFRewriter::UnitMetaVectorType &TUMetaVector,
-                              DWPState &State, DebugLocWriter &LocWriter) {
+                              DWPState &State, DebugLocWriter &LocWriter,
+                              DebugStrOffsetsWriter &StrOffstsWriter,
+                              DebugStrWriter &StrWriter) {
   const uint64_t DWOId = *CU.getDWOId();
   MCSection *const StrOffsetSection = State.MCOFI->getDwarfStrOffDWOSection();
   assert(StrOffsetSection && "StrOffsetSection does not exist.");
@@ -1941,15 +1888,18 @@ void DWARFRewriter::updateDWP(DWARFUnit &CU,
         TUEntry.Contributions[Index].getLength32();
     State.TypeIndexEntries.insert(std::make_pair(Hash, TUEntry));
   };
+  std::unique_ptr<DebugBufferVector> StrOffsetsOutputData;
+  std::unique_ptr<DebugBufferVector> StrOutputData;
   for (const SectionRef &Section : DWOFile->sections()) {
-    std::unique_ptr<DebugBufferVector> OutputData;
+    std::unique_ptr<DebugBufferVector> OutputData = nullptr;
     StringRef SectionName = getSectionName(Section);
     Expected<StringRef> ContentsExp = Section.getContents();
     assert(ContentsExp && "Invalid contents.");
-    std::optional<StringRef> TOutData = updateDebugData(
-        (*DWOCU)->getContext(), SectionName, *ContentsExp, State.KnownSections,
-        *State.Streamer, *this, CUDWOEntry, DWOId, OutputData,
-        RangeListssWriter, LocWriter, OverridenSections);
+    std::optional<StringRef> TOutData =
+        updateDebugData((*DWOCU)->getContext(), SectionName, *ContentsExp,
+                        State.KnownSections, *State.Streamer, *this, CUDWOEntry,
+                        DWOId, OutputData, RangeListssWriter, LocWriter,
+                        StrOffstsWriter, StrWriter, OverridenSections);
     if (!TOutData)
       continue;
 
@@ -1959,16 +1909,19 @@ void DWARFRewriter::updateDWP(DWARFUnit &CU,
       continue;
     }
 
-    if (SectionName.equals("debug_str.dwo")) {
+    if (SectionName == "debug_str.dwo") {
       CurStrSection = OutData;
+      StrOutputData = std::move(OutputData);
     } else {
       // Since handleDebugDataPatching returned true, we already know this is
       // a known section.
       auto SectionIter = State.KnownSections.find(SectionName);
-      if (SectionIter->second.second == DWARFSectionKind::DW_SECT_STR_OFFSETS)
+      if (SectionIter->second.second == DWARFSectionKind::DW_SECT_STR_OFFSETS) {
         CurStrOffsetSection = OutData;
-      else
+        StrOffsetsOutputData = std::move(OutputData);
+      } else {
         State.Streamer->emitBytes(OutData);
+      }
       unsigned int Index =
           getContributionIndex(SectionIter->second.second, State.IndexVersion);
       uint64_t Offset = State.ContributionOffsets[Index];
@@ -1992,6 +1945,10 @@ void DWARFRewriter::updateDWP(DWARFUnit &CU,
     // based on hash.
     if (!StrSectionWrittenOut && !CurStrOffsetSection.empty() &&
         !CurStrSection.empty()) {
+      // If debug_str.dwo section was modified storing it until dwp is written
+      // out. DWPStringPool stores raw pointers to strings.
+      if (StrOutputData)
+        State.StrSections.push_back(std::move(StrOutputData));
       writeStringsAndOffsets(*State.Streamer.get(), *State.Strings.get(),
                              StrOffsetSection, CurStrSection,
                              CurStrOffsetSection, CU.getVersion());
@@ -2017,7 +1974,8 @@ void DWARFRewriter::updateDWP(DWARFUnit &CU,
 
 void DWARFRewriter::writeDWOFiles(
     DWARFUnit &CU, const OverriddenSectionsMap &OverridenSections,
-    const std::string &DWOName, DebugLocWriter &LocWriter) {
+    const std::string &DWOName, DebugLocWriter &LocWriter,
+    DebugStrOffsetsWriter &StrOffstsWriter, DebugStrWriter &StrWriter) {
   // Setup DWP code once.
   DWARFContext *DWOCtx = BC.getDWOContext();
   const uint64_t DWOId = *CU.getDWOId();
@@ -2072,10 +2030,11 @@ void DWARFRewriter::writeDWOFiles(
     // have .debug_rnglists so won't be part of the loop below.
     if (!RangeListssWriter->empty()) {
       std::unique_ptr<DebugBufferVector> OutputData;
-      if (std::optional<StringRef> OutData = updateDebugData(
-              (*DWOCU)->getContext(), "debug_rnglists.dwo", "", KnownSections,
-              *Streamer, *this, CUDWOEntry, DWOId, OutputData,
-              RangeListssWriter, LocWriter, OverridenSections))
+      if (std::optional<StringRef> OutData =
+              updateDebugData((*DWOCU)->getContext(), "debug_rnglists.dwo", "",
+                              KnownSections, *Streamer, *this, CUDWOEntry,
+                              DWOId, OutputData, RangeListssWriter, LocWriter,
+                              StrOffstsWriter, StrWriter, OverridenSections))
         Streamer->emitBytes(*OutData);
     }
   }
@@ -2090,7 +2049,7 @@ void DWARFRewriter::writeDWOFiles(
     if (std::optional<StringRef> OutData = updateDebugData(
             (*DWOCU)->getContext(), SectionName, *ContentsExp, KnownSections,
             *Streamer, *this, CUDWOEntry, DWOId, OutputData, RangeListssWriter,
-            LocWriter, OverridenSections))
+            LocWriter, StrOffstsWriter, StrWriter, OverridenSections))
       Streamer->emitBytes(*OutData);
   }
   Streamer->finish();

diff --git a/bolt/lib/Rewrite/SDTRewriter.cpp b/bolt/lib/Rewrite/SDTRewriter.cpp
@@ -87,7 +87,7 @@ void SDTRewriter::readSection() {
 
     StringRef Name = DE.getCStr(&Offset);
 
-    if (!Name.equals("stapsdt"))
+    if (Name != "stapsdt")
       errs() << "BOLT-WARNING: SDT note name \"" << Name
              << "\" is not expected\n";
 

diff --git a/bolt/lib/Target/X86/X86MCPlusBuilder.cpp b/bolt/lib/Target/X86/X86MCPlusBuilder.cpp
@@ -328,19 +328,19 @@ class X86MCPlusBuilder : public MCPlusBuilder {
     return false;
   }
 
-  bool isUnsupportedBranch(const MCInst &Inst) const override {
+  bool isReversibleBranch(const MCInst &Inst) const override {
     if (isDynamicBranch(Inst))
-      return true;
+      return false;
 
     switch (Inst.getOpcode()) {
     default:
-      return false;
+      return true;
     case X86::LOOP:
     case X86::LOOPE:
     case X86::LOOPNE:
     case X86::JECXZ:
     case X86::JRCXZ:
-      return true;
+      return false;
     }
   }
 
@@ -1874,7 +1874,7 @@ class X86MCPlusBuilder : public MCPlusBuilder {
       }
 
       // Handle conditional branches and ignore indirect branches
-      if (!isUnsupportedBranch(*I) && getCondCode(*I) == X86::COND_INVALID) {
+      if (isReversibleBranch(*I) && getCondCode(*I) == X86::COND_INVALID) {
         // Indirect branch
         return false;
       }

diff --git a/bolt/test/X86/Inputs/blarge_new_bat_branchentry.preagg.txt b/bolt/test/X86/Inputs/blarge_new_bat_branchentry.preagg.txt
@@ -0,0 +1 @@
+B 80010c 800194 1 0
diff --git a/bolt/test/X86/Inputs/blarge_new_bat_order.preagg.txt b/bolt/test/X86/Inputs/blarge_new_bat_order.preagg.txt
@@ -0,0 +1,2 @@
+B 800154 401050 20 0
+F 800159 800193 7
diff --git a/bolt/test/X86/Inputs/dwarf5-df-types-debug-names-main.s b/bolt/test/X86/Inputs/dwarf5-df-types-debug-names-main.s
@@ -207,7 +207,7 @@ main:                                   # @main
 .Linfo_string5:
 	.asciz	"f2"                            # string offset=24
 .Linfo_string6:
-	.asciz	"/home/ayermolo/local/tasks/T138552329/typeDedupSplit" # string offset=27
+	.asciz	"." # string offset=27
 .Linfo_string7:
 	.asciz	"main.dwo"                      # string offset=80
 .Linfo_string8:
@@ -234,15 +234,15 @@ main:                                   # @main
 	.long	19
 	.long	24
 	.long	27
-	.long	80
-	.long	89
-	.long	92
-	.long	97
-	.long	100
-	.long	103
-	.long	106
-	.long	112
-	.long	220
+	.long	29
+	.long	38
+	.long	41
+	.long	46
+	.long	49
+	.long	52
+	.long	55
+	.long	61
+	.long	169
 	.section	.debug_info.dwo,"e",@progbits
 	.long	.Ldebug_info_dwo_end2-.Ldebug_info_dwo_start2 # Length of Unit
 .Ldebug_info_dwo_start2:
@@ -474,7 +474,7 @@ main:                                   # @main
 	.byte	1
 	.byte	8
 	.byte	2
-	.ascii	"/home/ayermolo/local/tasks/T138552329/typeDedupSplit"
+	.ascii	"."
 	.byte	0
 	.byte	46
 	.byte	0

diff --git a/bolt/test/X86/Inputs/jump-table-fixed-ref-pic.s b/bolt/test/X86/Inputs/jump-table-fixed-ref-pic.s
@@ -0,0 +1,35 @@
+  .globl main
+  .type main, %function
+main:
+  .cfi_startproc
+  cmpq $0x3, %rdi
+  jae .L4
+  cmpq $0x1, %rdi
+  jne .L4
+  mov .Ljt_pic+8(%rip), %rax
+  lea .Ljt_pic(%rip), %rdx
+  add %rdx, %rax
+  jmpq *%rax
+.L1:
+  movq $0x1, %rax
+  jmp .L5
+.L2:
+  movq $0x0, %rax
+  jmp .L5
+.L3:
+  movq $0x2, %rax
+  jmp .L5
+.L4:
+  mov $0x3, %rax
+.L5:
+  retq
+  .cfi_endproc
+
+  .section .rodata
+  .align 16
+.Ljt_pic:
+  .long .L1 - .Ljt_pic
+  .long .L2 - .Ljt_pic
+  .long .L3 - .Ljt_pic
+  .long .L4 - .Ljt_pic
+
diff --git a/bolt/test/X86/bb-with-two-tail-calls.s b/bolt/test/X86/bb-with-two-tail-calls.s
@@ -9,11 +9,11 @@
 # RUN: llvm-strip --strip-unneeded %t.o
 # RUN: %clang %cflags %t.o -o %t.exe -Wl,-q -nostdlib
 # RUN: llvm-bolt %t.exe -o %t.out --data %t.fdata --lite=0 --dyno-stats \
-# RUN:    --print-sctc --print-only=_start 2>&1 | FileCheck %s
+# RUN:    --print-sctc --print-only=_start -enable-bat 2>&1 | FileCheck %s
 # CHECK-NOT: Assertion `BranchInfo.size() == 2 && "could only be called for blocks with 2 successors"' failed.
 # Two tail calls in the same basic block after SCTC:
-# CHECK:         {{.*}}:   ja      {{.*}} # TAILCALL  # CTCTakenCount: {{.*}}
-# CHECK-NEXT:    {{.*}}:   jmp     {{.*}} # TAILCALL
+# CHECK:         {{.*}}:   ja      {{.*}} # TAILCALL # Offset: 7 # CTCTakenCount: 4
+# CHECK-NEXT:    {{.*}}:   jmp     {{.*}} # TAILCALL # Offset: 12
 
   .globl _start
 _start:

diff --git a/bolt/test/X86/bolt-address-translation-yaml.test b/bolt/test/X86/bolt-address-translation-yaml.test
@@ -5,6 +5,28 @@ RUN: llvm-bolt %t.exe -o %t.out --pa -p %p/Inputs/blarge_new.preagg.txt \
 RUN:   --reorder-blocks=ext-tsp --split-functions --split-strategy=cdsplit \
 RUN:   --reorder-functions=cdsort --enable-bat --dyno-stats --skip-funcs=main \
 RUN:   2>&1 | FileCheck --check-prefix WRITE-BAT-CHECK %s
+# Check that branch with entry in BAT is accounted for.
+RUN: perf2bolt %t.out --pa -p %p/Inputs/blarge_new_bat_branchentry.preagg.txt \
+RUN:   -w %t.yaml -o %t.fdata
+RUN: llvm-bolt %t.exe -data %t.fdata -w %t.yaml-fdata -o %t.null
+RUN: FileCheck --input-file %t.yaml --check-prefix BRANCHENTRY-YAML-CHECK %s
+RUN: FileCheck --input-file %t.yaml-fdata --check-prefix BRANCHENTRY-YAML-CHECK %s
+BRANCHENTRY-YAML-CHECK:    - name: SolveCubic
+BRANCHENTRY-YAML-CHECK:      bid: 0
+BRANCHENTRY-YAML-CHECK:      hash: 0x700F19D24600000
+BRANCHENTRY-YAML-CHECK-NEXT: succ: [ { bid: 7, cnt: 1 }
+# Check that the order is correct between BAT YAML and FDATA->YAML.
+RUN: perf2bolt %t.out --pa -p %p/Inputs/blarge_new_bat_order.preagg.txt \
+RUN:   -w %t.yaml -o %t.fdata
+RUN: llvm-bolt %t.exe -data %t.fdata -w %t.yaml-fdata -o %t.null
+RUN: FileCheck --input-file %t.yaml --check-prefix ORDER-YAML-CHECK %s
+RUN: FileCheck --input-file %t.yaml-fdata --check-prefix ORDER-YAML-CHECK %s
+ORDER-YAML-CHECK:    - name: SolveCubic
+ORDER-YAML-CHECK:      bid: 3
+ORDER-YAML-CHECK:      hash: 0xDDA1DC5F69F900AC
+ORDER-YAML-CHECK-NEXT: calls: [ { off: 0x26, fid: [[#]], cnt: 20 } ]
+ORDER-YAML-CHECK-NEXT: succ: [ { bid: 5, cnt: 7 }
+# Large profile test
 RUN: perf2bolt %t.out --pa -p %p/Inputs/blarge_new_bat.preagg.txt -w %t.yaml -o %t.fdata \
 RUN:   2>&1 | FileCheck --check-prefix READ-BAT-CHECK %s
 RUN: FileCheck --input-file %t.yaml --check-prefix YAML-BAT-CHECK %s
@@ -13,7 +35,7 @@ RUN: llvm-bolt %t.exe -data %t.fdata -w %t.yaml-fdata -o /dev/null
 RUN: FileCheck --input-file %t.yaml-fdata --check-prefix YAML-BAT-CHECK %s
 
 # Test resulting YAML profile with the original binary (no-stale mode)
-RUN: llvm-bolt %t.exe -data %t.yaml -o %t.null -dyno-stats \
+RUN: llvm-bolt %t.exe -data %t.yaml -o %t.null -dyno-stats 2>&1 \
 RUN:   | FileCheck --check-prefix CHECK-BOLT-YAML %s
 
 WRITE-BAT-CHECK: BOLT-INFO: Wrote 5 BAT maps
@@ -48,6 +70,10 @@ YAML-BAT-CHECK-NEXT:   hash:    0x6AF7E61EA3966722
 YAML-BAT-CHECK-NEXT:   exec:    25
 YAML-BAT-CHECK-NEXT:   nblocks: 15
 YAML-BAT-CHECK-NEXT:   blocks:
+YAML-BAT-CHECK-NEXT:   - bid:   0
+YAML-BAT-CHECK-NEXT:     insns: [[#]]
+YAML-BAT-CHECK-NEXT:     hash:  0x700F19D24600000
+YAML-BAT-CHECK-NEXT:     exec:  25
 YAML-BAT-CHECK:        - bid:   3
 YAML-BAT-CHECK-NEXT:     insns: [[#]]
 YAML-BAT-CHECK-NEXT:     hash:  0xDDA1DC5F69F900AC
@@ -63,7 +89,8 @@ YAML-BAT-CHECK-NEXT:   blocks:
 YAML-BAT-CHECK:        - bid:   1
 YAML-BAT-CHECK-NEXT:       insns: [[#]]
 YAML-BAT-CHECK-NEXT:       hash:  0xD70DC695320E0010
-YAML-BAT-CHECK-NEXT:       succ:  {{.*}} { bid: 2, cnt: [[#]] }
+YAML-BAT-CHECK-NEXT:       succ:  {{.*}} { bid: 2, cnt: [[#]]
 
 CHECK-BOLT-YAML:      pre-processing profile using YAML profile reader
 CHECK-BOLT-YAML-NEXT: 5 out of 16 functions in the binary (31.2%) have non-empty execution profile
+CHECK-BOLT-YAML-NOT: invalid (possibly stale) profile
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
@@ -0,0 +1,198 @@
+; RUN: rm -rf %t
+; RUN: mkdir %t
+; RUN: cd %t
+; RUN: llvm-mc -dwarf-version=5 -filetype=obj -triple x86_64-unknown-linux %p/Inputs/dwarf5-df-types-debug-names-main.s \
+; RUN: -split-dwarf-file=main.dwo -o main.o
+; RUN: llvm-mc -dwarf-version=5 -filetype=obj -triple x86_64-unknown-linux %p/Inputs/dwarf5-df-types-dup-helper.s \
+; RUN: -split-dwarf-file=helper.dwo -o helper.o
+; RUN: %clang %cflags -gdwarf-5 -gsplit-dwarf=split main.o helper.o -o main.exe
+; RUN: llvm-bolt main.exe -o main.exe.bolt --update-debug-sections
+; RUN: llvm-dwarfdump --debug-info -r 0 main.exe.bolt > log.txt
+; RUN: llvm-dwarfdump --debug-info -r 0 main.dwo.dwo >> log.txt
+; RUN: llvm-dwarfdump --debug-info -r 0 helper.dwo.dwo >> log.txt
+; RUN: llvm-dwarfdump --debug-str-offsets main.dwo.dwo >> log.txt
+; RUN: llvm-dwarfdump --debug-str-offsets helper.dwo.dwo >> log.txt
+; RUN: cat log.txt | FileCheck -check-prefix=BOLT %s
+
+;; Test is a mix of DWARF5 TUs where one has DW_AT_comp_dir/DW_AT_dwo_name, and another one doesn't.
+;; Tests that BOLT correctly updates DW_AT_dwo_name for TUs.
+
+; BOLT: DW_TAG_skeleton_unit
+; BOLT: DW_AT_comp_dir  (".")
+; BOLT: DW_AT_dwo_name  ("main.dwo.dwo")
+; BOLT: DW_TAG_skeleton_unit
+; BOLT: DW_AT_comp_dir  (".")
+; BOLT: DW_AT_dwo_name  ("helper.dwo.dwo")
+; BOLT: DW_TAG_type_unit
+; BOLT: DW_AT_comp_dir  (".")
+; BOLT: DW_AT_dwo_name  ("main.dwo.dwo")
+; BOLT: DW_TAG_type_unit
+; BOLT: DW_AT_comp_dir  (".")
+; BOLT: DW_AT_dwo_name  ("main.dwo.dwo")
+; BOLT: DW_TAG_type_unit
+; BOLT-NOT: DW_AT_dwo_name
+; BOLT: DW_TAG_type_unit
+; BOLT-NOT: DW_AT_dwo_name
+; BOLT: DW_TAG_compile_unit
+; BOLT:      .debug_str_offsets.dwo contents:
+; BOLT-NEXT: 0x00000000: Contribution size = 68, Format = DWARF32, Version = 5
+; BOLT-NEXT: "main"
+; BOLT-NEXT: "int"
+; BOLT-NEXT: "argc"
+; BOLT-NEXT: "argv"
+; BOLT-NEXT: "char"
+; BOLT-NEXT: "f2"
+; BOLT-NEXT: "."
+; BOLT-NEXT: "main.dwo.dwo"
+; BOLT-NEXT: "c1"
+; BOLT-NEXT: "Foo2"
+; BOLT-NEXT: "f3"
+; BOLT-NEXT: "c2"
+; BOLT-NEXT: "c3"
+; BOLT-NEXT: "Foo2a"
+; BOLT-NEXT: "clang version 18.0.0git (git@github.com:ayermolo/llvm-project.git db35fa8fc524127079662802c4735dbf397f86d0)"
+; BOLT-NEXT: "main.cpp"
+; BOLT-NEXT: helper.dwo.dwo: file format elf64-x86-64
+
+; BOLT:      .debug_str_offsets.dwo contents:
+; BOLT-NEXT: 0x00000000: Contribution size = 64, Format = DWARF32, Version = 5
+; BOLT-NEXT: "fooint"
+; BOLT-NEXT: "int"
+; BOLT-NEXT: "_Z3foov"
+; BOLT-NEXT: "foo"
+; BOLT-NEXT: "fint"
+; BOLT-NEXT: "c1"
+; BOLT-NEXT: "c2"
+; BOLT-NEXT: "Foo2Int"
+; BOLT-NEXT: "f"
+; BOLT-NEXT: "char"
+; BOLT-NEXT: "c3"
+; BOLT-NEXT: "Foo2a"
+; BOLT-NEXT: "clang version 18.0.0"
+; BOLT-NEXT: "helper.cpp"
+; BOLT-NEXT: "helper.dwo"
+
+
+;; Tests that BOLT correctly handles updating DW_AT_dwo_name when it outputs a DWP file.
+;; Currently skipping one of Type units because it is not being de-dupped.
+;; In the tu-index this TU is not present.
+; RUN: rm main.exe.bolt
+; RUN: llvm-bolt main.exe -o main.exe.bolt --update-debug-sections --write-dwp
+; RUN: llvm-dwarfdump --debug-info -r 0 main.exe.bolt.dwp > logDWP.txt
+; RUN: llvm-dwarfdump --debug-str-offsets main.exe.bolt.dwp >> logDWP.txt
+; RUN: cat logDWP.txt | FileCheck -check-prefix=BOLT-DWP %s
+; BOLT-DWP: DW_TAG_type_unit
+; BOLT-DWP: DW_AT_comp_dir  (".")
+; BOLT-DWP: DW_AT_dwo_name  ("main.dwo.dwo")
+; BOLT-DWP: DW_TAG_type_unit
+; BOLT-DWP: DW_AT_comp_dir  (".")
+; BOLT-DWP: DW_AT_dwo_name  ("main.dwo.dwo")
+; 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:       Contribution size = 68, Format = DWARF32, Version = 5
+; BOLT-DWP-NEXT: "main"
+; BOLT-DWP-NEXT: "int"
+; BOLT-DWP-NEXT: "argc"
+; BOLT-DWP-NEXT: "argv"
+; BOLT-DWP-NEXT: "char"
+; BOLT-DWP-NEXT: "f2"
+; BOLT-DWP-NEXT: "."
+; BOLT-DWP-NEXT: "main.dwo.dwo"
+; BOLT-DWP-NEXT: "c1"
+; BOLT-DWP-NEXT: "Foo2"
+; BOLT-DWP-NEXT: "f3"
+; BOLT-DWP-NEXT: "c2"
+; BOLT-DWP-NEXT: "c3"
+; BOLT-DWP-NEXT: "Foo2a"
+; BOLT-DWP-NEXT: "clang version 18.0.0git (git@github.com:ayermolo/llvm-project.git db35fa8fc524127079662802c4735dbf397f86d0)"
+; BOLT-DWP-NEXT: "main.cpp"
+; BOLT-DWP-NEXT: Contribution size = 64, Format = DWARF32, Version = 5
+; BOLT-DWP-NEXT: "fooint"
+; BOLT-DWP-NEXT: "int"
+; BOLT-DWP-NEXT: "_Z3foov"
+; BOLT-DWP-NEXT: "foo"
+; BOLT-DWP-NEXT: "fint"
+; BOLT-DWP-NEXT: "c1"
+; BOLT-DWP-NEXT: "c2"
+; BOLT-DWP-NEXT: "Foo2Int"
+; BOLT-DWP-NEXT: "f"
+; BOLT-DWP-NEXT: "char"
+; BOLT-DWP-NEXT: "c3"
+; BOLT-DWP-NEXT: "Foo2a"
+; BOLT-DWP-NEXT: "clang version 18.0.0"
+; BOLT-DWP-NEXT: "helper.cpp"
+; BOLT-DWP-NEXT: "helper.dwo
+
+;; Tests that BOLT correctly handles updating DW_AT_comp_dir/DW_AT_dwo_name when outptut directory is specified.
+
+; RUN: mkdir DWOOut
+; RUN: rm main.exe.bolt
+; RUN: llvm-bolt main.exe -o main.exe.bolt --update-debug-sections --dwarf-output-path=%t/DWOOut
+; RUN: cd DWOOut
+; RUN: llvm-dwarfdump --debug-info -r 0 ../main.exe.bolt > log.txt
+; RUN: llvm-dwarfdump --debug-info -r 0 main.dwo0.dwo >> log.txt
+; RUN: llvm-dwarfdump --debug-info -r 0 helper.dwo0.dwo >> log.txt
+; RUN: llvm-dwarfdump --debug-str-offsets main.dwo0.dwo >> log.txt
+; RUN: llvm-dwarfdump --debug-str-offsets helper.dwo0.dwo >> log.txt
+; RUN: cat log.txt | FileCheck -check-prefix=BOLT-PATH %s
+
+; BOLT-PATH: DW_TAG_skeleton_unit
+; BOLT-PATH: DW_AT_comp_dir  ("
+; BOLT-PATH-SAME: dwarf5-df-types-modify-dwo-name-mixed.test.tmp/DWOOut
+; BOLT-PATH: DW_AT_dwo_name  ("main.dwo0.dwo")
+; BOLT-PATH: DW_TAG_skeleton_unit
+; BOLT-PATH: DW_AT_comp_dir  ("
+; BOLT-PATH-SAME: dwarf5-df-types-modify-dwo-name-mixed.test.tmp/DWOOut
+; BOLT-PATH: DW_AT_dwo_name  ("helper.dwo0.dwo")
+; BOLT-PATH: DW_TAG_type_unit
+; BOLT-PATH: DW_AT_comp_dir  ("
+; BOLT-PATH-SAME: dwarf5-df-types-modify-dwo-name-mixed.test.tmp/DWOOut
+; BOLT-PATH: DW_AT_dwo_name  ("main.dwo0.dwo")
+; BOLT-PATH: DW_TAG_type_unit
+; BOLT-PATH: DW_AT_comp_dir  ("
+; BOLT-PATH-SAME: dwarf5-df-types-modify-dwo-name-mixed.test.tmp/DWOOut
+; BOLT-PATH: DW_AT_dwo_name  ("main.dwo0.dwo")
+; BOLT-PATH: DW_TAG_type_unit
+; BOLT-PATH-NOT: DW_AT_comp_dir
+; BOLT-PATH: DW_TAG_type_unit
+; BOLT-PATH-NOT: DW_AT_comp_dir
+; BOLT-PATH: DW_TAG_compile_unit
+; BOLT-PATH:      .debug_str_offsets.dwo contents:
+; BOLT-PATH-NEXT: 0x00000000: Contribution size = 68, Format = DWARF32, Version = 5
+; BOLT-PATH-NEXT: "main"
+; BOLT-PATH-NEXT: "int"
+; BOLT-PATH-NEXT: "argc"
+; BOLT-PATH-NEXT: "argv"
+; BOLT-PATH-NEXT: "char"
+; BOLT-PATH-NEXT: "f2"
+; BOLT-PATH-NEXT: dwarf5-df-types-modify-dwo-name-mixed.test.tmp/DWOOut"
+; BOLT-PATH-NEXT: "main.dwo0.dwo"
+; BOLT-PATH-NEXT: "c1"
+; BOLT-PATH-NEXT: "Foo2"
+; BOLT-PATH-NEXT: "f3"
+; BOLT-PATH-NEXT: "c2"
+; BOLT-PATH-NEXT: "c3"
+; BOLT-PATH-NEXT: "Foo2a"
+; BOLT-PATH-NEXT: "clang version 18.0.0git (git@github.com:ayermolo/llvm-project.git db35fa8fc524127079662802c4735dbf397f86d0)"
+; BOLT-PATH-NEXT: "main.cpp"
+; BOLT-PATH-NEXT: helper.dwo0.dwo: file format elf64-x86-64
+
+; BOLT-PATH:      .debug_str_offsets.dwo contents:
+; BOLT-PATH-NEXT: Contribution size = 64, Format = DWARF32, Version = 5
+; BOLT-PATH-NEXT: "fooint"
+; BOLT-PATH-NEXT: "int"
+; BOLT-PATH-NEXT: "_Z3foov"
+; BOLT-PATH-NEXT: "foo"
+; BOLT-PATH-NEXT: "fint"
+; BOLT-PATH-NEXT: "c1"
+; BOLT-PATH-NEXT: "c2"
+; BOLT-PATH-NEXT: "Foo2Int"
+; BOLT-PATH-NEXT: "f"
+; BOLT-PATH-NEXT: "char"
+; BOLT-PATH-NEXT: "c3"
+; BOLT-PATH-NEXT: "Foo2a"
+; BOLT-PATH-NEXT: "clang version 18.0.0"
+; BOLT-PATH-NEXT: "helper.cpp"
+; BOLT-PATH-NEXT: "helper.dwo"
diff --git a/bolt/test/X86/dwarf5-df-types-modify-dwo-name.test b/bolt/test/X86/dwarf5-df-types-modify-dwo-name.test
@@ -0,0 +1,175 @@
+; RUN: rm -rf %t
+; RUN: mkdir %t
+; RUN: cd %t
+; RUN: llvm-mc -dwarf-version=5 -filetype=obj -triple x86_64-unknown-linux %p/Inputs/dwarf5-df-types-debug-names-main.s \
+; RUN: -split-dwarf-file=main.dwo -o main.o
+; RUN: llvm-mc -dwarf-version=5 -filetype=obj -triple x86_64-unknown-linux %p/Inputs/dwarf5-df-types-debug-names-helper.s \
+; RUN: -split-dwarf-file=helper.dwo -o helper.o
+; RUN: %clang %cflags -gdwarf-5 -gsplit-dwarf=split main.o helper.o -o main.exe
+; RUN: llvm-bolt main.exe -o main.exe.bolt --update-debug-sections
+; RUN: llvm-dwarfdump --debug-info -r 0 main.exe.bolt > log.txt
+; RUN: llvm-dwarfdump --debug-info -r 0 main.dwo.dwo >> log.txt
+; RUN: llvm-dwarfdump --debug-info -r 0 helper.dwo.dwo >> log.txt
+; RUN: llvm-dwarfdump --debug-str-offsets main.dwo.dwo >> log.txt
+; RUN: llvm-dwarfdump --debug-str-offsets helper.dwo.dwo >> log.txt
+; RUN: cat log.txt | FileCheck -check-prefix=BOLT %s
+
+;; Tests that BOLT correctly updates DW_AT_dwo_name for TU Untis.
+
+; BOLT: DW_TAG_skeleton_unit
+; BOLT: DW_AT_comp_dir  (".")
+; BOLT: DW_AT_dwo_name  ("main.dwo.dwo")
+; BOLT: DW_TAG_skeleton_unit
+; BOLT: DW_AT_comp_dir  (".")
+; BOLT: DW_AT_dwo_name  ("helper.dwo.dwo")
+; BOLT: DW_TAG_type_unit
+; BOLT: DW_AT_comp_dir  (".")
+; BOLT: DW_AT_dwo_name  ("main.dwo.dwo")
+; BOLT: DW_TAG_type_unit
+; BOLT: DW_AT_comp_dir  (".")
+; BOLT: DW_AT_dwo_name  ("main.dwo.dwo")
+; BOLT: DW_TAG_type_unit
+; BOLT: DW_AT_comp_dir  (".")
+; BOLT: DW_AT_dwo_name  ("helper.dwo.dwo")
+; BOLT: DW_TAG_type_unit
+; BOLT: DW_AT_comp_dir  (".")
+; BOLT: DW_AT_dwo_name  ("helper.dwo.dwo")
+; BOLT:      .debug_str_offsets.dwo contents:
+; BOLT-NEXT: 0x00000000: Contribution size = 68, Format = DWARF32, Version = 5
+; BOLT-NEXT: "main"
+; BOLT-NEXT: "int"
+; BOLT-NEXT: "argc"
+; BOLT-NEXT: "argv"
+; BOLT-NEXT: "char"
+; BOLT-NEXT: "f2"
+; BOLT-NEXT: "."
+; BOLT-NEXT: "main.dwo.dwo"
+; BOLT-NEXT: "c1"
+; BOLT-NEXT: "Foo2"
+; BOLT-NEXT: "f3"
+; BOLT-NEXT: "c2"
+; BOLT-NEXT: "c3"
+; BOLT-NEXT: "Foo2a"
+; BOLT-NEXT: "clang version 18.0.0git (git@github.com:ayermolo/llvm-project.git db35fa8fc524127079662802c4735dbf397f86d0)"
+; BOLT-NEXT: "main.cpp"
+; BOLT-NEXT: helper.dwo.dwo: file format elf64-x86-64
+
+; BOLT:      .debug_str_offsets.dwo contents:
+; BOLT-NEXT: 0x00000000: Contribution size = 68, Format = DWARF32, Version = 5
+; BOLT-NEXT: "fooint"
+; BOLT-NEXT: "int"
+; BOLT-NEXT: "_Z3foov"
+; BOLT-NEXT: "foo"
+; BOLT-NEXT: "fint"
+; BOLT-NEXT: "."
+; BOLT-NEXT: "helper.dwo.dwo"
+; BOLT-NEXT: "c1"
+; BOLT-NEXT: "c2"
+; BOLT-NEXT: "Foo2Int"
+; BOLT-NEXT: "f"
+; BOLT-NEXT: "char"
+; BOLT-NEXT: "c3"
+; BOLT-NEXT: "Foo2a"
+; BOLT-NEXT: "clang version 18.0.0git (git@github.com:ayermolo/llvm-project.git db35fa8fc524127079662802c4735dbf397f86d0)"
+; BOLT-NEXT: "helper.cpp"
+
+
+;; Tests that BOLT correctly handles updating DW_AT_dwo_name when it outputs a DWP file.
+;; Currently skipping one of Type units because it is not being de-dupped.
+;; In the tu-index this TU is not present.
+; RUN: rm main.exe.bolt
+; RUN: llvm-bolt main.exe -o main.exe.bolt --update-debug-sections --write-dwp
+; RUN: llvm-dwarfdump --debug-info -r 0 main.exe.bolt.dwp > logDWP.txt
+; RUN: llvm-dwarfdump --debug-str-offsets main.exe.bolt.dwp >> logDWP.txt
+; RUN: cat logDWP.txt | FileCheck -check-prefix=BOLT-DWP %s
+; BOLT-DWP: DW_TAG_type_unit
+; BOLT-DWP: DW_AT_comp_dir  (".")
+; BOLT-DWP: DW_AT_dwo_name  ("main.dwo.dwo")
+; BOLT-DWP: DW_TAG_type_unit
+; BOLT-DWP: DW_AT_comp_dir  (".")
+; BOLT-DWP: DW_AT_dwo_name  ("main.dwo.dwo")
+; BOLT-DWP: DW_TAG_compile_unit
+; BOLT-DWP: DW_AT_dwo_name  ("main.dwo.dwo")
+; BOLT-DWP: DW_TAG_type_unit
+; BOLT-DWP: DW_AT_comp_dir  (".")
+; BOLT-DWP: DW_AT_dwo_name  ("helper.dwo.dwo")
+; BOLT-DWP: DW_TAG_type_unit
+; BOLT-DWP: DW_TAG_compile_unit
+; BOLT-DWP: DW_AT_name  ("helper.cpp")
+; BOLT-DWP: DW_AT_dwo_name  ("helper.dwo.dwo")
+
+;; Tests that BOLT correctly handles updating DW_AT_comp_dir/DW_AT_dwo_name when outptut directory is specified.
+
+; RUN: mkdir DWOOut
+; RUN: rm main.exe.bolt
+; RUN: llvm-bolt main.exe -o main.exe.bolt --update-debug-sections --dwarf-output-path=%t/DWOOut
+; RUN: cd DWOOut
+; RUN: llvm-dwarfdump --debug-info -r 0 ../main.exe.bolt > log.txt
+; RUN: llvm-dwarfdump --debug-info -r 0 main.dwo0.dwo >> log.txt
+; RUN: llvm-dwarfdump --debug-info -r 0 helper.dwo0.dwo >> log.txt
+; RUN: llvm-dwarfdump --debug-str-offsets main.dwo0.dwo >> log.txt
+; RUN: llvm-dwarfdump --debug-str-offsets helper.dwo0.dwo >> log.txt
+; RUN: cat log.txt | FileCheck -check-prefix=BOLT-PATH %s
+
+; BOLT-PATH: DW_TAG_skeleton_unit
+; BOLT-PATH: DW_AT_comp_dir  ("
+; BOLT-PATH-SAME: dwarf5-df-types-modify-dwo-name.test.tmp/DWOOut
+; BOLT-PATH: DW_AT_dwo_name  ("main.dwo0.dwo")
+; BOLT-PATH: DW_TAG_skeleton_unit
+; BOLT-PATH: DW_AT_comp_dir  ("
+; BOLT-PATH-SAME: dwarf5-df-types-modify-dwo-name.test.tmp/DWOOut
+; BOLT-PATH: DW_AT_dwo_name  ("helper.dwo0.dwo")
+; BOLT-PATH: DW_TAG_type_unit
+; BOLT-PATH: DW_AT_comp_dir  ("
+; BOLT-PATH-SAME: dwarf5-df-types-modify-dwo-name.test.tmp/DWOOut
+; BOLT-PATH: DW_AT_dwo_name  ("main.dwo0.dwo")
+; BOLT-PATH: DW_TAG_type_unit
+; BOLT-PATH: DW_AT_comp_dir  ("
+; BOLT-PATH-SAME: dwarf5-df-types-modify-dwo-name.test.tmp/DWOOut
+; BOLT-PATH: DW_AT_dwo_name  ("main.dwo0.dwo")
+; BOLT-PATH: DW_TAG_type_unit
+; BOLT-PATH: DW_AT_comp_dir  ("
+; BOLT-PATH-SAME: dwarf5-df-types-modify-dwo-name.test.tmp/DWOOut
+; BOLT-PATH: DW_AT_dwo_name  ("helper.dwo0.dwo")
+; BOLT-PATH: DW_TAG_type_unit
+; BOLT-PATH: DW_AT_comp_dir  ("
+; BOLT-PATH-SAME: dwarf5-df-types-modify-dwo-name.test.tmp/DWOOut
+; BOLT-PATH: DW_AT_dwo_name  ("helper.dwo0.dwo")
+; BOLT-PATH:      .debug_str_offsets.dwo contents:
+; BOLT-PATH-NEXT: 0x00000000: Contribution size = 68, Format = DWARF32, Version = 5
+; BOLT-PATH-NEXT: "main"
+; BOLT-PATH-NEXT: "int"
+; BOLT-PATH-NEXT: "argc"
+; BOLT-PATH-NEXT: "argv"
+; BOLT-PATH-NEXT: "char"
+; BOLT-PATH-NEXT: "f2"
+; BOLT-PATH-NEXT: dwarf5-df-types-modify-dwo-name.test.tmp/DWOOut"
+; BOLT-PATH-NEXT: "main.dwo0.dwo"
+; BOLT-PATH-NEXT: "c1"
+; BOLT-PATH-NEXT: "Foo2"
+; BOLT-PATH-NEXT: "f3"
+; BOLT-PATH-NEXT: "c2"
+; BOLT-PATH-NEXT: "c3"
+; BOLT-PATH-NEXT: "Foo2a"
+; BOLT-PATH-NEXT: "clang version 18.0.0git (git@github.com:ayermolo/llvm-project.git db35fa8fc524127079662802c4735dbf397f86d0)"
+; BOLT-PATH-NEXT: "main.cpp"
+; BOLT-PATH-NEXT: helper.dwo0.dwo: file format elf64-x86-64
+
+; BOLT-PATH:      .debug_str_offsets.dwo contents:
+; BOLT-PATH-NEXT: 0x00000000: Contribution size = 68, Format = DWARF32, Version = 5
+; BOLT-PATH-NEXT: "fooint"
+; BOLT-PATH-NEXT: "int"
+; BOLT-PATH-NEXT: "_Z3foov"
+; BOLT-PATH-NEXT: "foo"
+; BOLT-PATH-NEXT: "fint"
+; BOLT-PATH-NEXT: dwarf5-df-types-modify-dwo-name.test.tmp/DWOOut"
+; BOLT-PATH-NEXT: "helper.dwo0.dwo"
+; BOLT-PATH-NEXT: "c1"
+; BOLT-PATH-NEXT: "c2"
+; BOLT-PATH-NEXT: "Foo2Int"
+; BOLT-PATH-NEXT: "f"
+; BOLT-PATH-NEXT: "char"
+; BOLT-PATH-NEXT: "c3"
+; BOLT-PATH-NEXT: "Foo2a"
+; BOLT-PATH-NEXT: "clang version 18.0.0git (git@github.com:ayermolo/llvm-project.git db35fa8fc524127079662802c4735dbf397f86d0)"
+; BOLT-PATH-NEXT: "helper.cpp"
diff --git a/bolt/test/X86/jump-table-fixed-ref-pic.test b/bolt/test/X86/jump-table-fixed-ref-pic.test
@@ -0,0 +1,9 @@
+# Verify that BOLT detects fixed destination of indirect jump for PIC
+# case.
+
+XFAIL: *
+
+RUN: %clang %cflags -no-pie %S/Inputs/jump-table-fixed-ref-pic.s -Wl,-q -o %t
+RUN: llvm-bolt %t --relocs -o %t.null 2>&1 | FileCheck %s
+
+CHECK: BOLT-INFO: fixed indirect branch detected in main
diff --git a/bolt/test/X86/profile-passthrough-block.test b/bolt/test/X86/profile-passthrough-block.test
@@ -0,0 +1,67 @@
+## Test YAMLProfileReader support for pass-through blocks in non-matching edges:
+## match the profile edge A -> C to the CFG with blocks A -> B -> C.
+
+# REQUIRES: system-linux
+# RUN: split-file %s %t
+# RUN: llvm-mc -filetype=obj -triple x86_64-unknown-unknown %t/main.s -o %t.o
+# RUN: %clang %cflags %t.o -o %t.exe -Wl,-q -nostdlib
+# RUN: llvm-bolt %t.exe -o %t.out --data %t/yaml --profile-ignore-hash -v=1 \
+# RUN:   --print-cfg 2>&1 | FileCheck %s
+
+# CHECK: Binary Function "main" after building cfg
+# CHECK: Profile Acc : 100.0%
+# CHECK-NOT: BOLT-WARNING: no successor for block .LFT0 that matches index 3 or block .Ltmp0
+
+#--- main.s
+.globl main
+.type	main, @function
+main:
+  .cfi_startproc
+.LBB00:
+  pushq   %rbp
+  movq    %rsp, %rbp
+  subq    $16, %rsp
+  testq   %rax, %rax
+  js      .LBB03
+.LBB01:
+  jne     .LBB04
+.LBB02:
+  nop
+.LBB03:
+  xorl    %eax, %eax
+  addq    $16, %rsp
+  popq    %rbp
+  retq
+.LBB04:
+  xorl    %eax, %eax
+  addq    $16, %rsp
+  popq    %rbp
+  retq
+## For relocations against .text
+.LBB05:
+  call exit
+  .cfi_endproc
+  .size	main, .-main
+
+#--- yaml
+---
+header:
+  profile-version: 1
+  binary-name:     'profile-passthrough-block.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:            0x0000000000000000
+    exec:            1
+    nblocks:         6
+    blocks:
+      - bid:             1
+        insns:           1
+        succ:            [ { bid: 3, cnt: 1} ]
+...
diff --git a/bolt/test/X86/register-fragments-bolt-symbols.s b/bolt/test/X86/register-fragments-bolt-symbols.s
@@ -15,6 +15,8 @@
 # PREAGG: B X:0 #chain.cold.0# 1 0
 # RUN: perf2bolt %t.bolt -p %t.preagg --pa -o %t.bat.fdata -w %t.bat.yaml -v=1 \
 # RUN:   | FileCheck %s --check-prefix=CHECK-REGISTER
+# RUN: FileCheck --input-file %t.bat.fdata --check-prefix=CHECK-FDATA %s
+# RUN: FileCheck --input-file %t.bat.yaml --check-prefix=CHECK-YAML %s
 
 # CHECK-SYMS: l df *ABS*          [[#]] chain.s
 # CHECK-SYMS: l  F .bolt.org.text [[#]] chain
@@ -24,6 +26,9 @@
 
 # CHECK-REGISTER: BOLT-INFO: marking chain.cold.0/1(*2) as a fragment of chain/2(*2)
 
+# CHECK-FDATA: 0 [unknown] 0 1 chain/chain.s/2 10 0 1
+# CHECK-YAML: - name: 'chain/chain.s/2'
+
 .file "chain.s"
         .text
         .type   chain, @function

diff --git a/bolt/test/X86/sctc-bug4.test b/bolt/test/X86/sctc-bug4.test
@@ -1,25 +1,29 @@
-# Check that fallthrough blocks are handled properly.
+# Check that fallthrough blocks are handled properly and Offset annotation is
+# set for conditional tail calls.
 
 RUN: %clang %cflags %S/Inputs/sctc_bug4.s -o %t
-RUN: llvm-bolt %t -o %t.null \
+RUN: llvm-bolt %t -o %t.null --enable-bat \
 RUN:   -funcs=test_func -print-sctc -sequential-disassembly 2>&1 | FileCheck %s
 
 CHECK:      .Ltmp2 (3 instructions, align : 1)
 CHECK-NEXT:   CFI State : 0
+CHECK-NEXT:   Input offset: 0x24
 CHECK-NEXT:   Predecessors: .LFT1
 CHECK-NEXT:     00000024: 	cmpq	$0x20, %rsi
-CHECK-NEXT:     00000028: 	ja	dummy # TAILCALL {{.*}}# CTCTakenCount: 0
+CHECK-NEXT:     00000028: 	ja	dummy # TAILCALL # Offset: 53 # CTCTakenCount: 0
 CHECK-NEXT:     0000002a: 	jmp .Ltmp4
 CHECK-NEXT:   Successors: .Ltmp4
 CHECK-NEXT:   CFI State: 0
 
 CHECK:      .Ltmp1 (2 instructions, align : 1)
 CHECK-NEXT:   CFI State : 0
+CHECK-NEXT:   Input offset: 0x2c
 CHECK-NEXT:   Predecessors: .LFT0
 CHECK-NEXT:     0000002c: 	xorq	%r11, %rax
 CHECK-NEXT:     0000002f: 	retq
 CHECK-NEXT:   CFI State: 0
 
 CHECK:      .Ltmp4 (4 instructions, align : 1)
 CHECK-NEXT:  CFI State : 0
+CHECK-NEXT:  Input offset: 0x3a
 CHECK-NEXT:  Predecessors: .Ltmp2
diff --git a/bolt/test/runtime/bolt-reserved.cpp b/bolt/test/runtime/bolt-reserved.cpp
@@ -0,0 +1,40 @@
+// REQUIRES: system-linux
+
+/*
+ * Check that llvm-bolt uses reserved space in a binary for allocating
+ * new sections.
+ */
+
+// RUN: %clang %s -o %t.exe -Wl,-q
+// RUN: llvm-bolt %t.exe -o %t.bolt.exe 2>&1 | FileCheck %s
+// RUN: %t.bolt.exe
+
+// CHECK: BOLT-INFO: using reserved space
+
+/*
+ * Check that llvm-bolt detects a condition when the reserved space is
+ * not enough for allocating new sections.
+ */
+
+// RUN: %clang %s -o %t.tiny.exe -Wl,--no-eh-frame-hdr -Wl,-q -DTINY
+// RUN: not llvm-bolt %t.tiny.exe -o %t.tiny.bolt.exe 2>&1 | \
+// RUN:   FileCheck %s --check-prefix=CHECK-TINY
+
+// CHECK-TINY: BOLT-ERROR: reserved space (1 byte) is smaller than required
+
+#ifdef TINY
+#define RSIZE "1"
+#else
+#define RSIZE "8192 * 1024"
+#endif
+
+asm(".pushsection .text \n\
+       .globl __bolt_reserved_start \n\
+       .type __bolt_reserved_start, @object \n\
+       __bolt_reserved_start: \n\
+       .space " RSIZE " \n\
+       .globl __bolt_reserved_end \n\
+       __bolt_reserved_end: \n\
+     .popsection");
+
+int main() { return 0; }
diff --git a/clang-tools-extra/clang-query/CMakeLists.txt b/clang-tools-extra/clang-query/CMakeLists.txt
@@ -20,7 +20,6 @@ clang_target_link_libraries(clangQuery
   clangBasic
   clangDynamicASTMatchers
   clangFrontend
-  clangTooling
   clangSerialization
   )
 

diff --git a/clang-tools-extra/clang-query/Query.cpp b/clang-tools-extra/clang-query/Query.cpp
@@ -7,12 +7,12 @@
 //===----------------------------------------------------------------------===//
 
 #include "Query.h"
+#include "QueryParser.h"
 #include "QuerySession.h"
 #include "clang/AST/ASTDumper.h"
 #include "clang/ASTMatchers/ASTMatchFinder.h"
 #include "clang/Frontend/ASTUnit.h"
 #include "clang/Frontend/TextDiagnostic.h"
-#include "clang/Tooling/NodeIntrospection.h"
 #include "llvm/Support/raw_ostream.h"
 #include <optional>
 
@@ -68,8 +68,6 @@ bool HelpQuery::run(llvm::raw_ostream &OS, QuerySession &QS) const {
         "Diagnostic location for bound nodes.\n"
         "  detailed-ast                      "
         "Detailed AST output for bound nodes.\n"
-        "  srcloc                            "
-        "Source locations and ranges for bound nodes.\n"
         "  dump                              "
         "Detailed AST output for bound nodes (alias of detailed-ast).\n\n";
   return true;
@@ -90,90 +88,6 @@ struct CollectBoundNodes : MatchFinder::MatchCallback {
   }
 };
 
-void dumpLocations(llvm::raw_ostream &OS, DynTypedNode Node, ASTContext &Ctx,
-                   const DiagnosticsEngine &Diags, SourceManager const &SM) {
-  auto Locs = clang::tooling::NodeIntrospection::GetLocations(Node);
-
-  auto PrintLocations = [](llvm::raw_ostream &OS, auto Iter, auto End) {
-    auto CommonEntry = Iter->first;
-    auto Scout = Iter;
-    SmallVector<std::string> LocationStrings;
-    while (Scout->first == CommonEntry) {
-      LocationStrings.push_back(
-          tooling::LocationCallFormatterCpp::format(*Iter->second));
-      if (Scout == End)
-        break;
-      ++Scout;
-      if (Scout->first == CommonEntry)
-        ++Iter;
-    }
-    llvm::sort(LocationStrings);
-    for (auto &LS : LocationStrings) {
-      OS << " * \"" << LS << "\"\n";
-    }
-    return Iter;
-  };
-
-  TextDiagnostic TD(OS, Ctx.getLangOpts(), &Diags.getDiagnosticOptions());
-
-  for (auto Iter = Locs.LocationAccessors.begin();
-       Iter != Locs.LocationAccessors.end(); ++Iter) {
-    if (!Iter->first.isValid())
-      continue;
-
-    TD.emitDiagnostic(FullSourceLoc(Iter->first, SM), DiagnosticsEngine::Note,
-                      "source locations here", std::nullopt, std::nullopt);
-
-    Iter = PrintLocations(OS, Iter, Locs.LocationAccessors.end());
-    OS << '\n';
-  }
-
-  for (auto Iter = Locs.RangeAccessors.begin();
-       Iter != Locs.RangeAccessors.end(); ++Iter) {
-
-    if (!Iter->first.getBegin().isValid())
-      continue;
-
-    if (SM.getPresumedLineNumber(Iter->first.getBegin()) !=
-        SM.getPresumedLineNumber(Iter->first.getEnd()))
-      continue;
-
-    TD.emitDiagnostic(
-        FullSourceLoc(Iter->first.getBegin(), SM), DiagnosticsEngine::Note,
-        "source ranges here " + Iter->first.printToString(SM),
-        CharSourceRange::getTokenRange(Iter->first), std::nullopt);
-
-    Iter = PrintLocations(OS, Iter, Locs.RangeAccessors.end());
-  }
-  for (auto Iter = Locs.RangeAccessors.begin();
-       Iter != Locs.RangeAccessors.end(); ++Iter) {
-
-    if (!Iter->first.getBegin().isValid())
-      continue;
-
-    if (SM.getPresumedLineNumber(Iter->first.getBegin()) ==
-        SM.getPresumedLineNumber(Iter->first.getEnd()))
-      continue;
-
-    TD.emitDiagnostic(
-        FullSourceLoc(Iter->first.getBegin(), SM), DiagnosticsEngine::Note,
-        "source range " + Iter->first.printToString(SM) + " starting here...",
-        CharSourceRange::getTokenRange(Iter->first), std::nullopt);
-
-    auto ColNum = SM.getPresumedColumnNumber(Iter->first.getEnd());
-    auto LastLineLoc = Iter->first.getEnd().getLocWithOffset(-(ColNum - 1));
-
-    TD.emitDiagnostic(FullSourceLoc(Iter->first.getEnd(), SM),
-                      DiagnosticsEngine::Note, "... ending here",
-                      CharSourceRange::getTokenRange(
-                          SourceRange(LastLineLoc, Iter->first.getEnd())),
-                      std::nullopt);
-
-    Iter = PrintLocations(OS, Iter, Locs.RangeAccessors.end());
-  }
-  OS << "\n";
-}
-
 } // namespace
 
 bool MatchQuery::run(llvm::raw_ostream &OS, QuerySession &QS) const {
@@ -194,8 +108,7 @@ bool MatchQuery::run(llvm::raw_ostream &OS, QuerySession &QS) const {
       return false;
     }
 
-    auto &Ctx = AST->getASTContext();
-    const auto &SM = Ctx.getSourceManager();
+    ASTContext &Ctx = AST->getASTContext();
     Ctx.getParentMapContext().setTraversalKind(QS.TK);
     Finder.matchAST(Ctx);
 
@@ -243,19 +156,11 @@ bool MatchQuery::run(llvm::raw_ostream &OS, QuerySession &QS) const {
         }
         if (QS.DetailedASTOutput) {
           OS << "Binding for \"" << BI->first << "\":\n";
-          const ASTContext &Ctx = AST->getASTContext();
           ASTDumper Dumper(OS, Ctx, AST->getDiagnostics().getShowColors());
           Dumper.SetTraversalKind(QS.TK);
           Dumper.Visit(BI->second);
           OS << "\n";
         }
-        if (QS.SrcLocOutput) {
-          OS << "\n  \"" << BI->first << "\" Source locations\n";
-          OS << "  " << std::string(19 + BI->first.size(), '-') << '\n';
-
-          dumpLocations(OS, BI->second, Ctx, AST->getDiagnostics(), SM);
-          OS << "\n";
-        }
       }
 
       if (MI->getMap().empty())
@@ -281,5 +186,26 @@ const QueryKind SetQueryKind<bool>::value;
 const QueryKind SetQueryKind<OutputKind>::value;
 #endif
 
+bool FileQuery::run(llvm::raw_ostream &OS, QuerySession &QS) const {
+  auto Buffer = llvm::MemoryBuffer::getFile(StringRef{File}.trim());
+  if (!Buffer) {
+    if (Prefix.has_value())
+      llvm::errs() << *Prefix << ": ";
+    llvm::errs() << "cannot open " << File << ": "
+                 << Buffer.getError().message() << "\n";
+    return false;
+  }
+
+  StringRef FileContentRef(Buffer.get()->getBuffer());
+
+  while (!FileContentRef.empty()) {
+    QueryRef Q = QueryParser::parse(FileContentRef, QS);
+    if (!Q->run(llvm::outs(), QS))
+      return false;
+    FileContentRef = Q->RemainingContent;
+  }
+  return true;
+}
+
 } // namespace query
 } // namespace clang
diff --git a/clang-tools-extra/clang-query/Query.h b/clang-tools-extra/clang-query/Query.h
@@ -17,7 +17,7 @@
 namespace clang {
 namespace query {
 
-enum OutputKind { OK_Diag, OK_Print, OK_DetailedAST, OK_SrcLoc };
+enum OutputKind { OK_Diag, OK_Print, OK_DetailedAST };
 
 enum QueryKind {
   QK_Invalid,
@@ -30,7 +30,8 @@ enum QueryKind {
   QK_SetTraversalKind,
   QK_EnableOutputKind,
   QK_DisableOutputKind,
-  QK_Quit
+  QK_Quit,
+  QK_File
 };
 
 class QuerySession;
@@ -148,7 +149,6 @@ struct SetExclusiveOutputQuery : Query {
     QS.DiagOutput = false;
     QS.DetailedASTOutput = false;
     QS.PrintOutput = false;
-    QS.SrcLocOutput = false;
     QS.*Var = true;
     return true;
   }
@@ -188,6 +188,21 @@ struct DisableOutputQuery : SetNonExclusiveOutputQuery {
   }
 };
 
+struct FileQuery : Query {
+  FileQuery(StringRef File, StringRef Prefix = StringRef())
+      : Query(QK_File), File(File),
+        Prefix(!Prefix.empty() ? std::optional<std::string>(Prefix)
+                               : std::nullopt) {}
+
+  bool run(llvm::raw_ostream &OS, QuerySession &QS) const override;
+
+  static bool classof(const Query *Q) { return Q->Kind == QK_File; }
+
+private:
+  std::string File;
+  std::optional<std::string> Prefix;
+};
+
 } // namespace query
 } // namespace clang
 

diff --git a/clang-tools-extra/clang-query/QueryParser.cpp b/clang-tools-extra/clang-query/QueryParser.cpp
@@ -11,7 +11,6 @@
 #include "QuerySession.h"
 #include "clang/ASTMatchers/Dynamic/Parser.h"
 #include "clang/Basic/CharInfo.h"
-#include "clang/Tooling/NodeIntrospection.h"
 #include "llvm/ADT/StringRef.h"
 #include "llvm/ADT/StringSwitch.h"
 #include <optional>
@@ -104,19 +103,16 @@ QueryRef QueryParser::parseSetBool(bool QuerySession::*Var) {
 
 template <typename QueryType> QueryRef QueryParser::parseSetOutputKind() {
   StringRef ValStr;
-  bool HasIntrospection = tooling::NodeIntrospection::hasIntrospectionSupport();
-  unsigned OutKind =
-      LexOrCompleteWord<unsigned>(this, ValStr)
-          .Case("diag", OK_Diag)
-          .Case("print", OK_Print)
-          .Case("detailed-ast", OK_DetailedAST)
-          .Case("srcloc", OK_SrcLoc, /*IsCompletion=*/HasIntrospection)
-          .Case("dump", OK_DetailedAST)
-          .Default(~0u);
+  unsigned OutKind = LexOrCompleteWord<unsigned>(this, ValStr)
+                         .Case("diag", OK_Diag)
+                         .Case("print", OK_Print)
+                         .Case("detailed-ast", OK_DetailedAST)
+                         .Case("dump", OK_DetailedAST)
+                         .Default(~0u);
   if (OutKind == ~0u) {
-    return new InvalidQuery("expected 'diag', 'print', 'detailed-ast'" +
-                            StringRef(HasIntrospection ? ", 'srcloc'" : "") +
-                            " or 'dump', got '" + ValStr + "'");
+    return new InvalidQuery("expected 'diag', 'print', 'detailed-ast' or "
+                            "'dump', got '" +
+                            ValStr + "'");
   }
 
   switch (OutKind) {
@@ -126,10 +122,6 @@ template <typename QueryType> QueryRef QueryParser::parseSetOutputKind() {
     return new QueryType(&QuerySession::DiagOutput);
   case OK_Print:
     return new QueryType(&QuerySession::PrintOutput);
-  case OK_SrcLoc:
-    if (HasIntrospection)
-      return new QueryType(&QuerySession::SrcLocOutput);
-    return new InvalidQuery("'srcloc' output support is not available.");
   }
 
   llvm_unreachable("Invalid output kind");
@@ -183,7 +175,8 @@ enum ParsedQueryKind {
   PQK_Unlet,
   PQK_Quit,
   PQK_Enable,
-  PQK_Disable
+  PQK_Disable,
+  PQK_File
 };
 
 enum ParsedQueryVariable {
@@ -222,12 +215,14 @@ QueryRef QueryParser::doParse() {
                               .Case("let", PQK_Let)
                               .Case("m", PQK_Match, /*IsCompletion=*/false)
                               .Case("match", PQK_Match)
-                              .Case("q", PQK_Quit,  /*IsCompletion=*/false)
+                              .Case("q", PQK_Quit, /*IsCompletion=*/false)
                               .Case("quit", PQK_Quit)
                               .Case("set", PQK_Set)
                               .Case("enable", PQK_Enable)
                               .Case("disable", PQK_Disable)
                               .Case("unlet", PQK_Unlet)
+                              .Case("f", PQK_File, /*IsCompletion=*/false)
+                              .Case("file", PQK_File)
                               .Default(PQK_Invalid);
 
   switch (QKind) {
@@ -351,6 +346,9 @@ QueryRef QueryParser::doParse() {
     return endQuery(new LetQuery(Name, VariantValue()));
   }
 
+  case PQK_File:
+    return new FileQuery(Line);
+
   case PQK_Invalid:
     return new InvalidQuery("unknown command: " + CommandStr);
   }

diff --git a/clang-tools-extra/clang-query/QuerySession.h b/clang-tools-extra/clang-query/QuerySession.h
@@ -25,15 +25,14 @@ class QuerySession {
 public:
   QuerySession(llvm::ArrayRef<std::unique_ptr<ASTUnit>> ASTs)
       : ASTs(ASTs), PrintOutput(false), DiagOutput(true),
-        DetailedASTOutput(false), SrcLocOutput(false), BindRoot(true),
-        PrintMatcher(false), Terminate(false), TK(TK_AsIs) {}
+        DetailedASTOutput(false), BindRoot(true), PrintMatcher(false),
+        Terminate(false), TK(TK_AsIs) {}
 
   llvm::ArrayRef<std::unique_ptr<ASTUnit>> ASTs;
 
   bool PrintOutput;
   bool DiagOutput;
   bool DetailedASTOutput;
-  bool SrcLocOutput;
 
   bool BindRoot;
   bool PrintMatcher;

diff --git a/clang-tools-extra/clang-query/tool/ClangQuery.cpp b/clang-tools-extra/clang-query/tool/ClangQuery.cpp
@@ -74,22 +74,8 @@ static cl::opt<std::string> PreloadFile(
 
 bool runCommandsInFile(const char *ExeName, std::string const &FileName,
                        QuerySession &QS) {
-  auto Buffer = llvm::MemoryBuffer::getFile(FileName);
-  if (!Buffer) {
-    llvm::errs() << ExeName << ": cannot open " << FileName << ": "
-                 << Buffer.getError().message() << "\n";
-    return true;
-  }
-
-  StringRef FileContentRef(Buffer.get()->getBuffer());
-
-  while (!FileContentRef.empty()) {
-    QueryRef Q = QueryParser::parse(FileContentRef, QS);
-    if (!Q->run(llvm::outs(), QS))
-      return true;
-    FileContentRef = Q->RemainingContent;
-  }
-  return false;
+  FileQuery Query(FileName, ExeName);
+  return !Query.run(llvm::errs(), QS);
 }
 
 int main(int argc, const char **argv) {

diff --git a/clang-tools-extra/clang-tidy/ClangTidy.cpp b/clang-tools-extra/clang-tidy/ClangTidy.cpp
@@ -373,11 +373,11 @@ static CheckersList getAnalyzerCheckersAndPackages(ClangTidyContext &Context,
 
   const auto &RegisteredCheckers =
       AnalyzerOptions::getRegisteredCheckers(IncludeExperimental);
-  bool AnalyzerChecksEnabled = false;
-  for (StringRef CheckName : RegisteredCheckers) {
-    std::string ClangTidyCheckName((AnalyzerCheckNamePrefix + CheckName).str());
-    AnalyzerChecksEnabled |= Context.isCheckEnabled(ClangTidyCheckName);
-  }
+  const bool AnalyzerChecksEnabled =
+      llvm::any_of(RegisteredCheckers, [&](StringRef CheckName) -> bool {
+        return Context.isCheckEnabled(
+            (AnalyzerCheckNamePrefix + CheckName).str());
+      });
 
   if (!AnalyzerChecksEnabled)
     return List;

diff --git a/clang-tools-extra/clang-tidy/ClangTidyCheck.cpp b/clang-tools-extra/clang-tidy/ClangTidyCheck.cpp
@@ -171,7 +171,7 @@ std::optional<int64_t> ClangTidyCheck::OptionsView::getEnumInt(
     if (IgnoreCase) {
       if (Value.equals_insensitive(NameAndEnum.second))
         return NameAndEnum.first;
-    } else if (Value.equals(NameAndEnum.second)) {
+    } else if (Value == NameAndEnum.second) {
       return NameAndEnum.first;
     } else if (Value.equals_insensitive(NameAndEnum.second)) {
       Closest = NameAndEnum.second;

diff --git a/clang-tools-extra/clang-tidy/ClangTidyDiagnosticConsumer.cpp b/clang-tools-extra/clang-tidy/ClangTidyDiagnosticConsumer.cpp
@@ -311,7 +311,18 @@ ClangTidyDiagnosticConsumer::ClangTidyDiagnosticConsumer(
     : Context(Ctx), ExternalDiagEngine(ExternalDiagEngine),
       RemoveIncompatibleErrors(RemoveIncompatibleErrors),
       GetFixesFromNotes(GetFixesFromNotes),
-      EnableNolintBlocks(EnableNolintBlocks) {}
+      EnableNolintBlocks(EnableNolintBlocks) {
+
+  if (Context.getOptions().HeaderFilterRegex &&
+      !Context.getOptions().HeaderFilterRegex->empty())
+    HeaderFilter =
+        std::make_unique<llvm::Regex>(*Context.getOptions().HeaderFilterRegex);
+
+  if (Context.getOptions().ExcludeHeaderFilterRegex &&
+      !Context.getOptions().ExcludeHeaderFilterRegex->empty())
+    ExcludeHeaderFilter = std::make_unique<llvm::Regex>(
+        *Context.getOptions().ExcludeHeaderFilterRegex);
+}
 
 void ClangTidyDiagnosticConsumer::finalizeLastError() {
   if (!Errors.empty()) {
@@ -562,22 +573,17 @@ void ClangTidyDiagnosticConsumer::checkFilters(SourceLocation Location,
   }
 
   StringRef FileName(File->getName());
-  LastErrorRelatesToUserCode = LastErrorRelatesToUserCode ||
-                               Sources.isInMainFile(Location) ||
-                               getHeaderFilter()->match(FileName);
+  LastErrorRelatesToUserCode =
+      LastErrorRelatesToUserCode || Sources.isInMainFile(Location) ||
+      (HeaderFilter &&
+       (HeaderFilter->match(FileName) &&
+        !(ExcludeHeaderFilter && ExcludeHeaderFilter->match(FileName))));
 
   unsigned LineNumber = Sources.getExpansionLineNumber(Location);
   LastErrorPassesLineFilter =
       LastErrorPassesLineFilter || passesLineFilter(FileName, LineNumber);
 }
 
-llvm::Regex *ClangTidyDiagnosticConsumer::getHeaderFilter() {
-  if (!HeaderFilter)
-    HeaderFilter =
-        std::make_unique<llvm::Regex>(*Context.getOptions().HeaderFilterRegex);
-  return HeaderFilter.get();
-}
-
 void ClangTidyDiagnosticConsumer::removeIncompatibleErrors() {
   // Each error is modelled as the set of intervals in which it applies
   // replacements. To detect overlapping replacements, we use a sweep line

diff --git a/clang-tools-extra/clang-tidy/ClangTidyDiagnosticConsumer.h b/clang-tools-extra/clang-tidy/ClangTidyDiagnosticConsumer.h
@@ -313,6 +313,7 @@ class ClangTidyDiagnosticConsumer : public DiagnosticConsumer {
   bool EnableNolintBlocks;
   std::vector<ClangTidyError> Errors;
   std::unique_ptr<llvm::Regex> HeaderFilter;
+  std::unique_ptr<llvm::Regex> ExcludeHeaderFilter;
   bool LastErrorRelatesToUserCode = false;
   bool LastErrorPassesLineFilter = false;
   bool LastErrorWasIgnored = false;

diff --git a/clang-tools-extra/clang-tidy/ClangTidyOptions.cpp b/clang-tools-extra/clang-tidy/ClangTidyOptions.cpp
@@ -170,6 +170,8 @@ template <> struct MappingTraits<ClangTidyOptions> {
     IO.mapOptional("ImplementationFileExtensions",
                    Options.ImplementationFileExtensions);
     IO.mapOptional("HeaderFilterRegex", Options.HeaderFilterRegex);
+    IO.mapOptional("ExcludeHeaderFilterRegex",
+                   Options.ExcludeHeaderFilterRegex);
     IO.mapOptional("FormatStyle", Options.FormatStyle);
     IO.mapOptional("User", Options.User);
     IO.mapOptional("CheckOptions", Options.CheckOptions);
@@ -191,7 +193,8 @@ ClangTidyOptions ClangTidyOptions::getDefaults() {
   Options.WarningsAsErrors = "";
   Options.HeaderFileExtensions = {"", "h", "hh", "hpp", "hxx"};
   Options.ImplementationFileExtensions = {"c", "cc", "cpp", "cxx"};
-  Options.HeaderFilterRegex = "";
+  Options.HeaderFilterRegex = std::nullopt;
+  Options.ExcludeHeaderFilterRegex = std::nullopt;
   Options.SystemHeaders = false;
   Options.FormatStyle = "none";
   Options.User = std::nullopt;
@@ -231,6 +234,7 @@ ClangTidyOptions &ClangTidyOptions::mergeWith(const ClangTidyOptions &Other,
   overrideValue(ImplementationFileExtensions,
                 Other.ImplementationFileExtensions);
   overrideValue(HeaderFilterRegex, Other.HeaderFilterRegex);
+  overrideValue(ExcludeHeaderFilterRegex, Other.ExcludeHeaderFilterRegex);
   overrideValue(SystemHeaders, Other.SystemHeaders);
   overrideValue(FormatStyle, Other.FormatStyle);
   overrideValue(User, Other.User);

diff --git a/clang-tools-extra/clang-tidy/ClangTidyOptions.h b/clang-tools-extra/clang-tidy/ClangTidyOptions.h
@@ -83,6 +83,10 @@ struct ClangTidyOptions {
   /// main files will always be displayed.
   std::optional<std::string> HeaderFilterRegex;
 
+  /// \brief Exclude warnings from headers matching this filter, even if they
+  /// match \c HeaderFilterRegex.
+  std::optional<std::string> ExcludeHeaderFilterRegex;
+
   /// Output warnings from system headers matching \c HeaderFilterRegex.
   std::optional<bool> SystemHeaders;
 

diff --git a/clang-tools-extra/clang-tidy/bugprone/ForwardingReferenceOverloadCheck.cpp b/clang-tools-extra/clang-tidy/bugprone/ForwardingReferenceOverloadCheck.cpp
@@ -25,8 +25,8 @@ AST_MATCHER(QualType, isEnableIf) {
     const NamedDecl *TypeDecl =
         Spec->getTemplateName().getAsTemplateDecl()->getTemplatedDecl();
     return TypeDecl->isInStdNamespace() &&
-           (TypeDecl->getName().equals("enable_if") ||
-            TypeDecl->getName().equals("enable_if_t"));
+           (TypeDecl->getName() == "enable_if" ||
+            TypeDecl->getName() == "enable_if_t");
   };
   const Type *BaseType = Node.getTypePtr();
   // Case: pointer or reference to enable_if.

diff --git a/clang-tools-extra/clang-tidy/bugprone/ImplicitWideningOfMultiplicationResultCheck.cpp b/clang-tools-extra/clang-tidy/bugprone/ImplicitWideningOfMultiplicationResultCheck.cpp
@@ -9,20 +9,20 @@
 #include "ImplicitWideningOfMultiplicationResultCheck.h"
 #include "clang/AST/ASTContext.h"
 #include "clang/ASTMatchers/ASTMatchFinder.h"
+#include "clang/ASTMatchers/ASTMatchersMacros.h"
 #include "clang/Lex/Lexer.h"
 #include <optional>
 
 using namespace clang::ast_matchers;
 
-namespace clang {
+namespace clang::tidy::bugprone {
+
 namespace {
 AST_MATCHER(ImplicitCastExpr, isPartOfExplicitCast) {
   return Node.isPartOfExplicitCast();
 }
+AST_MATCHER(Expr, containsErrors) { return Node.containsErrors(); }
 } // namespace
-} // namespace clang
-
-namespace clang::tidy::bugprone {
 
 static const Expr *getLHSOfMulBinOp(const Expr *E) {
   assert(E == E->IgnoreParens() && "Already skipped all parens!");
@@ -250,7 +250,8 @@ void ImplicitWideningOfMultiplicationResultCheck::handlePointerOffsetting(
 
 void ImplicitWideningOfMultiplicationResultCheck::registerMatchers(
     MatchFinder *Finder) {
-  Finder->addMatcher(implicitCastExpr(unless(anyOf(isInTemplateInstantiation(),
+  Finder->addMatcher(implicitCastExpr(unless(anyOf(containsErrors(),
+                                                   isInTemplateInstantiation(),
                                                    isPartOfExplicitCast())),
                                       hasCastKind(CK_IntegralCast))
                          .bind("x"),

diff --git a/clang-tools-extra/clang-tidy/bugprone/OptionalValueConversionCheck.cpp b/clang-tools-extra/clang-tidy/bugprone/OptionalValueConversionCheck.cpp
@@ -71,7 +71,9 @@ void OptionalValueConversionCheck::registerMatchers(MatchFinder *Finder) {
               ofClass(matchers::matchesAnyListedName(OptionalTypes)))),
           hasType(ConstructTypeMatcher),
           hasArgument(0U, ignoringImpCasts(anyOf(OptionalDereferenceMatcher,
-                                                 StdMoveCallMatcher))))
+                                                 StdMoveCallMatcher))),
+          unless(anyOf(hasAncestor(typeLoc()),
+                       hasAncestor(expr(matchers::hasUnevaluatedContext())))))
           .bind("expr"),
       this);
 }

diff --git a/clang-tools-extra/clang-tidy/bugprone/ReservedIdentifierCheck.cpp b/clang-tools-extra/clang-tidy/bugprone/ReservedIdentifierCheck.cpp
@@ -178,8 +178,11 @@ std::optional<RenamerClangTidyCheck::FailureInfo>
 ReservedIdentifierCheck::getDeclFailureInfo(const NamedDecl *Decl,
                                             const SourceManager &) const {
   assert(Decl && Decl->getIdentifier() && !Decl->getName().empty() &&
-         !Decl->isImplicit() &&
          "Decl must be an explicit identifier with a name.");
+  // Implicit identifiers cannot fail.
+  if (Decl->isImplicit())
+    return std::nullopt;
+
   return getFailureInfoImpl(
       Decl->getName(), isa<TranslationUnitDecl>(Decl->getDeclContext()),
       /*IsMacro = */ false, getLangOpts(), Invert, AllowedIdentifiers);

diff --git a/clang-tools-extra/clang-tidy/bugprone/SuspiciousEnumUsageCheck.cpp b/clang-tools-extra/clang-tidy/bugprone/SuspiciousEnumUsageCheck.cpp
@@ -171,8 +171,7 @@ void SuspiciousEnumUsageCheck::check(const MatchFinder::MatchResult &Result) {
     // Skip when one of the parameters is an empty enum. The
     // hasDisjointValueRange function could not decide the values properly in
     // case of an empty enum.
-    if (EnumDec->enumerator_begin() == EnumDec->enumerator_end() ||
-        OtherEnumDec->enumerator_begin() == OtherEnumDec->enumerator_end())
+    if (EnumDec->enumerators().empty() || OtherEnumDec->enumerators().empty())
       return;
 
     if (!hasDisjointValueRange(EnumDec, OtherEnumDec))

diff --git a/clang-tools-extra/clang-tidy/cppcoreguidelines/MisleadingCaptureDefaultByValueCheck.cpp b/clang-tools-extra/clang-tidy/cppcoreguidelines/MisleadingCaptureDefaultByValueCheck.cpp
@@ -67,8 +67,7 @@ static std::string createReplacementText(const LambdaExpr *Lambda) {
       AppendName("this");
     }
   }
-  if (!Replacement.empty() &&
-      Lambda->explicit_capture_begin() != Lambda->explicit_capture_end()) {
+  if (!Replacement.empty() && !Lambda->explicit_captures().empty()) {
     // Add back separator if we are adding explicit capture variables.
     Stream << ", ";
   }

diff --git a/clang-tools-extra/clang-tidy/cppcoreguidelines/SpecialMemberFunctionsCheck.cpp b/clang-tools-extra/clang-tidy/cppcoreguidelines/SpecialMemberFunctionsCheck.cpp
@@ -25,25 +25,44 @@ SpecialMemberFunctionsCheck::SpecialMemberFunctionsCheck(
                                          "AllowMissingMoveFunctions", false)),
       AllowSoleDefaultDtor(Options.get("AllowSoleDefaultDtor", false)),
       AllowMissingMoveFunctionsWhenCopyIsDeleted(
-          Options.get("AllowMissingMoveFunctionsWhenCopyIsDeleted", false)) {}
+          Options.get("AllowMissingMoveFunctionsWhenCopyIsDeleted", false)),
+      AllowImplicitlyDeletedCopyOrMove(
+          Options.get("AllowImplicitlyDeletedCopyOrMove", false)) {}
 
 void SpecialMemberFunctionsCheck::storeOptions(
     ClangTidyOptions::OptionMap &Opts) {
   Options.store(Opts, "AllowMissingMoveFunctions", AllowMissingMoveFunctions);
   Options.store(Opts, "AllowSoleDefaultDtor", AllowSoleDefaultDtor);
   Options.store(Opts, "AllowMissingMoveFunctionsWhenCopyIsDeleted",
                 AllowMissingMoveFunctionsWhenCopyIsDeleted);
+  Options.store(Opts, "AllowImplicitlyDeletedCopyOrMove",
+                AllowImplicitlyDeletedCopyOrMove);
+}
+
+std::optional<TraversalKind>
+SpecialMemberFunctionsCheck::getCheckTraversalKind() const {
+  return AllowImplicitlyDeletedCopyOrMove ? TK_AsIs
+                                          : TK_IgnoreUnlessSpelledInSource;
 }
 
 void SpecialMemberFunctionsCheck::registerMatchers(MatchFinder *Finder) {
+  auto IsNotImplicitOrDeleted = anyOf(unless(isImplicit()), isDeleted());
+
   Finder->addMatcher(
       cxxRecordDecl(
-          eachOf(has(cxxDestructorDecl().bind("dtor")),
-                 has(cxxConstructorDecl(isCopyConstructor()).bind("copy-ctor")),
-                 has(cxxMethodDecl(isCopyAssignmentOperator())
+          unless(isImplicit()),
+          eachOf(has(cxxDestructorDecl(unless(isImplicit())).bind("dtor")),
+                 has(cxxConstructorDecl(isCopyConstructor(),
+                                        IsNotImplicitOrDeleted)
+                         .bind("copy-ctor")),
+                 has(cxxMethodDecl(isCopyAssignmentOperator(),
+                                   IsNotImplicitOrDeleted)
                          .bind("copy-assign")),
-                 has(cxxConstructorDecl(isMoveConstructor()).bind("move-ctor")),
-                 has(cxxMethodDecl(isMoveAssignmentOperator())
+                 has(cxxConstructorDecl(isMoveConstructor(),
+                                        IsNotImplicitOrDeleted)
+                         .bind("move-ctor")),
+                 has(cxxMethodDecl(isMoveAssignmentOperator(),
+                                   IsNotImplicitOrDeleted)
                          .bind("move-assign"))))
           .bind("class-def"),
       this);
@@ -127,7 +146,8 @@ void SpecialMemberFunctionsCheck::check(
   for (const auto &KV : Matchers)
     if (const auto *MethodDecl =
             Result.Nodes.getNodeAs<CXXMethodDecl>(KV.first)) {
-      StoreMember({KV.second, MethodDecl->isDeleted()});
+      StoreMember(
+          {KV.second, MethodDecl->isDeleted(), MethodDecl->isImplicit()});
     }
 }
 
@@ -144,7 +164,13 @@ void SpecialMemberFunctionsCheck::checkForMissingMembers(
 
   auto HasMember = [&](SpecialMemberFunctionKind Kind) {
     return llvm::any_of(DefinedMembers, [Kind](const auto &Data) {
-      return Data.FunctionKind == Kind;
+      return Data.FunctionKind == Kind && !Data.IsImplicit;
+    });
+  };
+
+  auto HasImplicitDeletedMember = [&](SpecialMemberFunctionKind Kind) {
+    return llvm::any_of(DefinedMembers, [Kind](const auto &Data) {
+      return Data.FunctionKind == Kind && Data.IsImplicit && Data.IsDeleted;
     });
   };
 
@@ -154,9 +180,17 @@ void SpecialMemberFunctionsCheck::checkForMissingMembers(
     });
   };
 
-  auto RequireMember = [&](SpecialMemberFunctionKind Kind) {
-    if (!HasMember(Kind))
-      MissingMembers.push_back(Kind);
+  auto RequireMembers = [&](SpecialMemberFunctionKind Kind1,
+                            SpecialMemberFunctionKind Kind2) {
+    if (AllowImplicitlyDeletedCopyOrMove && HasImplicitDeletedMember(Kind1) &&
+        HasImplicitDeletedMember(Kind2))
+      return;
+
+    if (!HasMember(Kind1))
+      MissingMembers.push_back(Kind1);
+
+    if (!HasMember(Kind2))
+      MissingMembers.push_back(Kind2);
   };
 
   bool RequireThree =
@@ -180,23 +214,25 @@ void SpecialMemberFunctionsCheck::checkForMissingMembers(
         !HasMember(SpecialMemberFunctionKind::NonDefaultDestructor))
       MissingMembers.push_back(SpecialMemberFunctionKind::Destructor);
 
-    RequireMember(SpecialMemberFunctionKind::CopyConstructor);
-    RequireMember(SpecialMemberFunctionKind::CopyAssignment);
+    RequireMembers(SpecialMemberFunctionKind::CopyConstructor,
+                   SpecialMemberFunctionKind::CopyAssignment);
   }
 
   if (RequireFive &&
       !(AllowMissingMoveFunctionsWhenCopyIsDeleted &&
         (IsDeleted(SpecialMemberFunctionKind::CopyConstructor) &&
          IsDeleted(SpecialMemberFunctionKind::CopyAssignment)))) {
     assert(RequireThree);
-    RequireMember(SpecialMemberFunctionKind::MoveConstructor);
-    RequireMember(SpecialMemberFunctionKind::MoveAssignment);
+    RequireMembers(SpecialMemberFunctionKind::MoveConstructor,
+                   SpecialMemberFunctionKind::MoveAssignment);
   }
 
   if (!MissingMembers.empty()) {
     llvm::SmallVector<SpecialMemberFunctionKind, 5> DefinedMemberKinds;
-    llvm::transform(DefinedMembers, std::back_inserter(DefinedMemberKinds),
-                    [](const auto &Data) { return Data.FunctionKind; });
+    for (const auto &Data : DefinedMembers) {
+      if (!Data.IsImplicit)
+        DefinedMemberKinds.push_back(Data.FunctionKind);
+    }
     diag(ID.first, "class '%0' defines %1 but does not define %2")
         << ID.second << cppcoreguidelines::join(DefinedMemberKinds, " and ")
         << cppcoreguidelines::join(MissingMembers, " or ");

diff --git a/clang-tools-extra/clang-tidy/cppcoreguidelines/SpecialMemberFunctionsCheck.h b/clang-tools-extra/clang-tidy/cppcoreguidelines/SpecialMemberFunctionsCheck.h
@@ -30,9 +30,8 @@ class SpecialMemberFunctionsCheck : public ClangTidyCheck {
   void registerMatchers(ast_matchers::MatchFinder *Finder) override;
   void check(const ast_matchers::MatchFinder::MatchResult &Result) override;
   void onEndOfTranslationUnit() override;
-  std::optional<TraversalKind> getCheckTraversalKind() const override {
-    return TK_IgnoreUnlessSpelledInSource;
-  }
+  std::optional<TraversalKind> getCheckTraversalKind() const override;
+
   enum class SpecialMemberFunctionKind : uint8_t {
     Destructor,
     DefaultDestructor,
@@ -46,6 +45,7 @@ class SpecialMemberFunctionsCheck : public ClangTidyCheck {
   struct SpecialMemberFunctionData {
     SpecialMemberFunctionKind FunctionKind;
     bool IsDeleted;
+    bool IsImplicit = false;
 
     bool operator==(const SpecialMemberFunctionData &Other) const {
       return (Other.FunctionKind == FunctionKind) &&
@@ -67,6 +67,7 @@ class SpecialMemberFunctionsCheck : public ClangTidyCheck {
   const bool AllowMissingMoveFunctions;
   const bool AllowSoleDefaultDtor;
   const bool AllowMissingMoveFunctionsWhenCopyIsDeleted;
+  const bool AllowImplicitlyDeletedCopyOrMove;
   ClassDefiningSpecialMembersMap ClassWithSpecialMembers;
 };
 

diff --git a/clang-tools-extra/clang-tidy/hicpp/SignedBitwiseCheck.cpp b/clang-tools-extra/clang-tidy/hicpp/SignedBitwiseCheck.cpp
@@ -9,6 +9,7 @@
 #include "SignedBitwiseCheck.h"
 #include "clang/AST/ASTContext.h"
 #include "clang/ASTMatchers/ASTMatchFinder.h"
+#include "clang/ASTMatchers/ASTMatchers.h"
 
 using namespace clang::ast_matchers;
 using namespace clang::ast_matchers::internal;
@@ -29,8 +30,8 @@ void SignedBitwiseCheck::storeOptions(ClangTidyOptions::OptionMap &Opts) {
 void SignedBitwiseCheck::registerMatchers(MatchFinder *Finder) {
   const auto SignedIntegerOperand =
       (IgnorePositiveIntegerLiterals
-           ? expr(ignoringImpCasts(hasType(isSignedInteger())),
-                  unless(integerLiteral()))
+           ? expr(ignoringImpCasts(
+                 allOf(hasType(isSignedInteger()), unless(integerLiteral()))))
            : expr(ignoringImpCasts(hasType(isSignedInteger()))))
           .bind("signed-operand");
 

diff --git a/clang-tools-extra/clang-tidy/misc/UnusedParametersCheck.cpp b/clang-tools-extra/clang-tidy/misc/UnusedParametersCheck.cpp
@@ -192,9 +192,7 @@ void UnusedParametersCheck::check(const MatchFinder::MatchResult &Result) {
 
     // In non-strict mode ignore function definitions with empty bodies
     // (constructor initializer counts for non-empty body).
-    if (StrictMode ||
-        (Function->getBody()->child_begin() !=
-         Function->getBody()->child_end()) ||
+    if (StrictMode || !Function->getBody()->children().empty() ||
         (isa<CXXConstructorDecl>(Function) &&
          cast<CXXConstructorDecl>(Function)->getNumCtorInitializers() > 0))
       warnOnUnusedParameter(Result, Function, I);

diff --git a/clang-tools-extra/clang-tidy/modernize/CMakeLists.txt b/clang-tools-extra/clang-tidy/modernize/CMakeLists.txt
@@ -41,6 +41,7 @@ add_clang_library(clangTidyModernizeModule
   UseNullptrCheck.cpp
   UseOverrideCheck.cpp
   UseStartsEndsWithCheck.cpp
+  UseStdFormatCheck.cpp
   UseStdNumbersCheck.cpp
   UseStdPrintCheck.cpp
   UseTrailingReturnTypeCheck.cpp

diff --git a/clang-tools-extra/clang-tidy/modernize/LoopConvertCheck.cpp b/clang-tools-extra/clang-tidy/modernize/LoopConvertCheck.cpp
@@ -421,7 +421,7 @@ getContainerFromBeginEndCall(const Expr *Init, bool IsBegin, bool *IsArrow,
     return {};
   if (IsReverse && !Call->Name.consume_back("r"))
     return {};
-  if (!Call->Name.empty() && !Call->Name.equals("c"))
+  if (!Call->Name.empty() && Call->Name != "c")
     return {};
   return std::make_pair(Call->Container, Call->CallKind);
 }