Merge branch 'main' into sdag/arm/glue/vfp-cmp

Author: s-barannikov

bolt/include/bolt/Core/BinarySection.h

@@ -87,6 +87,7 @@ class BinarySection {
                                    // been renamed)
   uint64_t OutputAddress{0};       // Section address for the rewritten binary.
   uint64_t OutputSize{0};          // Section size in the rewritten binary.
+                                   // Can exceed OutputContents with padding.
   uint64_t OutputFileOffset{0};    // File offset in the rewritten binary file.
   StringRef OutputContents;        // Rewritten section contents.
   const uint64_t SectionNumber;    // Order in which the section was created.
@@ -474,6 +475,11 @@ class BinarySection {
   /// Use name \p SectionName for the section during the emission.
   void emitAsData(MCStreamer &Streamer, const Twine &SectionName) const;
 
+  /// Write finalized contents of the section. If OutputSize exceeds the size of
+  /// the OutputContents, append zero padding to the stream and return the
+  /// number of byte written which should match the OutputSize.
+  uint64_t write(raw_ostream &OS) const;
+
   using SymbolResolverFuncTy = llvm::function_ref<uint64_t(const MCSymbol *)>;
 
   /// Flush all pending relocations to patch original contents of sections
@@ -497,6 +503,9 @@ class BinarySection {
     IsFinalized = true;
   }
 
+  /// When writing section contents, add \p PaddingSize zero bytes at the end.
+  void addPadding(uint64_t PaddingSize) { OutputSize += PaddingSize; }
+
   /// Reorder the contents of this section according to /p Order.  If
   /// /p Inplace is true, the entire contents of the section is reordered,
   /// otherwise the new contents contain only the reordered data.

bolt/lib/Core/BinaryEmitter.cpp

@@ -416,17 +416,6 @@ void BinaryEmitter::emitFunctionBody(BinaryFunction &BF, FunctionFragment &FF,
     BF.duplicateConstantIslands();
   }
 
-  if (!FF.empty() && FF.front()->isLandingPad()) {
-    assert(!FF.front()->isEntryPoint() &&
-           "Landing pad cannot be entry point of function");
-    // If the first block of the fragment is a landing pad, it's offset from the
-    // start of the area that the corresponding LSDA describes is zero. In this
-    // case, the call site entries in that LSDA have 0 as offset to the landing
-    // pad, which the runtime interprets as "no handler". To prevent this,
-    // insert some padding.
-    Streamer.emitBytes(BC.MIB->getTrapFillValue());
-  }
-
   // Track the first emitted instruction with debug info.
   bool FirstInstr = true;
   for (BinaryBasicBlock *const BB : FF) {
@@ -926,39 +915,54 @@ void BinaryEmitter::emitLSDA(BinaryFunction &BF, const FunctionFragment &FF) {
   // Emit the LSDA header.
 
   // If LPStart is omitted, then the start of the FDE is used as a base for
-  // landing pad displacements. Then if a cold fragment starts with
-  // a landing pad, this means that the first landing pad offset will be 0.
-  // As a result, the exception handling runtime will ignore this landing pad
-  // because zero offset denotes the absence of a landing pad.
-  // For this reason, when the binary has fixed starting address we emit LPStart
-  // as 0 and output the absolute value of the landing pad in the table.
+  // landing pad displacements. Then, if a cold fragment starts with a landing
+  // pad, this means that the first landing pad offset will be 0. However, C++
+  // runtime treats 0 as if there is no landing pad present, thus we *must* emit
+  // non-zero offsets for all valid LPs.
   //
-  // If the base address can change, we cannot use absolute addresses for
-  // landing pads (at least not without runtime relocations). Hence, we fall
-  // back to emitting landing pads relative to the FDE start.
-  // As we are emitting label differences, we have to guarantee both labels are
-  // defined in the same section and hence cannot place the landing pad into a
-  // cold fragment when the corresponding call site is in the hot fragment.
-  // Because of this issue and the previously described issue of possible
-  // zero-offset landing pad we have to place landing pads in the same section
-  // as the corresponding invokes for shared objects.
+  // As a solution, for fixed-address binaries we set LPStart to 0, and for
+  // position-independent binaries we set LP start to FDE start minus one byte
+  // for FDEs that start with a landing pad.
+  const bool NeedsLPAdjustment = !FF.empty() && FF.front()->isLandingPad();
   std::function<void(const MCSymbol *)> emitLandingPad;
   if (BC.HasFixedLoadAddress) {
     Streamer.emitIntValue(dwarf::DW_EH_PE_udata4, 1); // LPStart format
     Streamer.emitIntValue(0, 4);                      // LPStart
     emitLandingPad = [&](const MCSymbol *LPSymbol) {
-      if (!LPSymbol)
-        Streamer.emitIntValue(0, 4);
-      else
+      if (LPSymbol)
         Streamer.emitSymbolValue(LPSymbol, 4);
+      else
+        Streamer.emitIntValue(0, 4);
     };
   } else {
-    Streamer.emitIntValue(dwarf::DW_EH_PE_omit, 1); // LPStart format
+    if (NeedsLPAdjustment) {
+      // Use relative LPStart format and emit LPStart as [SymbolStart - 1].
+      Streamer.emitIntValue(dwarf::DW_EH_PE_pcrel | dwarf::DW_EH_PE_sdata4, 1);
+      MCSymbol *DotSymbol = BC.Ctx->createTempSymbol("LPBase");
+      Streamer.emitLabel(DotSymbol);
+
+      const MCExpr *LPStartExpr = MCBinaryExpr::createSub(
+          MCSymbolRefExpr::create(StartSymbol, *BC.Ctx),
+          MCSymbolRefExpr::create(DotSymbol, *BC.Ctx), *BC.Ctx);
+      LPStartExpr = MCBinaryExpr::createSub(
+          LPStartExpr, MCConstantExpr::create(1, *BC.Ctx), *BC.Ctx);
+      Streamer.emitValue(LPStartExpr, 4);
+    } else {
+      // DW_EH_PE_omit means FDE start (StartSymbol) will be used as LPStart.
+      Streamer.emitIntValue(dwarf::DW_EH_PE_omit, 1);
+    }
     emitLandingPad = [&](const MCSymbol *LPSymbol) {
-      if (!LPSymbol)
+      if (LPSymbol) {
+        const MCExpr *LPOffsetExpr = MCBinaryExpr::createSub(
+            MCSymbolRefExpr::create(LPSymbol, *BC.Ctx),
+            MCSymbolRefExpr::create(StartSymbol, *BC.Ctx), *BC.Ctx);
+        if (NeedsLPAdjustment)
+          LPOffsetExpr = MCBinaryExpr::createAdd(
+              LPOffsetExpr, MCConstantExpr::create(1, *BC.Ctx), *BC.Ctx);
+        Streamer.emitValue(LPOffsetExpr, 4);
+      } else {
         Streamer.emitIntValue(0, 4);
-      else
-        Streamer.emitAbsoluteSymbolDiff(LPSymbol, StartSymbol, 4);
+      }
     };
   }
 

bolt/lib/Core/BinarySection.cpp

@@ -142,6 +142,15 @@ void BinarySection::emitAsData(MCStreamer &Streamer,
     Streamer.emitLabel(BC.Ctx->getOrCreateSymbol("__hot_data_end"));
 }
 
+uint64_t BinarySection::write(raw_ostream &OS) const {
+  const uint64_t NumValidContentBytes =
+      std::min<uint64_t>(getOutputContents().size(), getOutputSize());
+  OS.write(getOutputContents().data(), NumValidContentBytes);
+  if (getOutputSize() > NumValidContentBytes)
+    OS.write_zeros(getOutputSize() - NumValidContentBytes);
+  return getOutputSize();
+}
+
 void BinarySection::flushPendingRelocations(raw_pwrite_stream &OS,
                                             SymbolResolverFuncTy Resolver) {
   if (PendingRelocations.empty() && Patches.empty())

bolt/lib/Rewrite/RewriteInstance.cpp

@@ -3887,6 +3887,43 @@ void RewriteInstance::mapCodeSections(BOLTLinker::SectionMapper MapSection) {
 
 void RewriteInstance::mapAllocatableSections(
     BOLTLinker::SectionMapper MapSection) {
+
+  if (opts::UseOldText || opts::StrictMode) {
+    auto tryRewriteSection = [&](BinarySection &OldSection,
+                                 BinarySection &NewSection) {
+      if (OldSection.getSize() < NewSection.getOutputSize())
+        return;
+
+      BC->outs() << "BOLT-INFO: rewriting " << OldSection.getName()
+                 << " in-place\n";
+
+      NewSection.setOutputAddress(OldSection.getAddress());
+      NewSection.setOutputFileOffset(OldSection.getInputFileOffset());
+      MapSection(NewSection, OldSection.getAddress());
+
+      // Pad contents with zeros.
+      NewSection.addPadding(OldSection.getSize() - NewSection.getOutputSize());
+
+      // Prevent the original section name from appearing in the section header
+      // table.
+      OldSection.setAnonymous(true);
+    };
+
+    if (EHFrameSection) {
+      BinarySection *NewEHFrameSection =
+          getSection(getNewSecPrefix() + getEHFrameSectionName());
+      assert(NewEHFrameSection && "New contents expected for .eh_frame");
+      tryRewriteSection(*EHFrameSection, *NewEHFrameSection);
+    }
+    BinarySection *EHSection = getSection(".gcc_except_table");
+    BinarySection *NewEHSection =
+        getSection(getNewSecPrefix() + ".gcc_except_table");
+    if (EHSection) {
+      assert(NewEHSection && "New contents expected for .gcc_except_table");
+      tryRewriteSection(*EHSection, *NewEHSection);
+    }
+  }
+
   // Allocate read-only sections first, then writable sections.
   enum : uint8_t { ST_READONLY, ST_READWRITE };
   for (uint8_t SType = ST_READONLY; SType <= ST_READWRITE; ++SType) {
@@ -4164,7 +4201,6 @@ void RewriteInstance::rewriteNoteSections() {
     // New section size.
     uint64_t Size = 0;
     bool DataWritten = false;
-    uint8_t *SectionData = nullptr;
     // Copy over section contents unless it's one of the sections we overwrite.
     if (!willOverwriteSection(SectionName)) {
       Size = Section.sh_size;
@@ -4196,12 +4232,7 @@ void RewriteInstance::rewriteNoteSections() {
     if (BSec->getAllocAddress()) {
       assert(!DataWritten && "Writing section twice.");
       (void)DataWritten;
-      SectionData = BSec->getOutputData();
-
-      LLVM_DEBUG(dbgs() << "BOLT-DEBUG: " << (Size ? "appending" : "writing")
-                        << " contents to section " << SectionName << '\n');
-      OS.write(reinterpret_cast<char *>(SectionData), BSec->getOutputSize());
-      Size += BSec->getOutputSize();
+      Size += BSec->write(OS);
     }
 
     BSec->setOutputFileOffset(NextAvailableOffset);
@@ -4232,8 +4263,7 @@ void RewriteInstance::rewriteNoteSections() {
                << " of size " << Section.getOutputSize() << " at offset 0x"
                << Twine::utohexstr(Section.getOutputFileOffset()) << '\n');
 
-    OS.write(Section.getOutputContents().data(), Section.getOutputSize());
-    NextAvailableOffset += Section.getOutputSize();
+    NextAvailableOffset += Section.write(OS);
   }
 }
 
@@ -4347,6 +4377,10 @@ RewriteInstance::getOutputSections(ELFObjectFile<ELFT> *File,
     BinarySection *BinSec = BC->getSectionForSectionRef(SecRef);
     assert(BinSec && "Matching BinarySection should exist.");
 
+    // Exclude anonymous sections.
+    if (BinSec->isAnonymous())
+      continue;
+
     addSection(Section, *BinSec);
   }
 
@@ -5699,8 +5733,8 @@ void RewriteInstance::rewriteFile() {
                  << Twine::utohexstr(Section.getAllocAddress()) << "\n of size "
                  << Section.getOutputSize() << "\n at offset "
                  << Section.getOutputFileOffset() << '\n';
-    OS.pwrite(reinterpret_cast<const char *>(Section.getOutputData()),
-              Section.getOutputSize(), Section.getOutputFileOffset());
+    OS.seek(Section.getOutputFileOffset());
+    Section.write(OS);
   }
 
   for (BinarySection &Section : BC->allocatableSections())

bolt/test/eh-frame-overwrite.test

@@ -0,0 +1,8 @@
+# Check that llvm-bolt can overwrite .eh_frame section in-place.
+
+REQUIRES: system-linux
+
+RUN: %clang %cflags %p/Inputs/hello.c -o %t -Wl,-q
+RUN: llvm-bolt %t -o %t.bolt --strict | FileCheck %s
+
+CHECK: rewriting .eh_frame in-place

clang-tools-extra/clangd/refactor/tweaks/DefineOutline.cpp

@@ -239,6 +239,8 @@ getFunctionSourceCode(const FunctionDecl *FD, const DeclContext *TargetContext,
           return;
 
         for (const NamedDecl *ND : Ref.Targets) {
+          if (ND->getKind() == Decl::TemplateTypeParm)
+            return;
           if (ND->getDeclContext() != Ref.Targets.front()->getDeclContext()) {
             elog("Targets from multiple contexts: {0}, {1}",
                  printQualifiedName(*Ref.Targets.front()),

clang-tools-extra/clangd/unittests/tweaks/DefineOutlineTests.cpp

@@ -411,29 +411,29 @@ inline typename O1<T, U...>::template O2<V, A>::E O1<T, U...>::template O2<V, A>
           R"cpp(
             struct Foo {
               template <typename T, bool B = true>
-              void ^bar() {}
+              T ^bar() { return {}; }
             };)cpp",
           R"cpp(
             struct Foo {
               template <typename T, bool B = true>
-              void bar() ;
+              T bar() ;
             };template <typename T, bool B>
-inline void Foo::bar() {}
+inline T Foo::bar() { return {}; }
 )cpp",
           ""},
 
       // Class template with member template
       {
           R"cpp(
             template <typename T> struct Foo {
-              template <typename U> void ^bar(const T& t, const U& u) {}
+              template <typename U> T ^bar(const T& t, const U& u) { return {}; }
             };)cpp",
           R"cpp(
             template <typename T> struct Foo {
-              template <typename U> void bar(const T& t, const U& u) ;
+              template <typename U> T bar(const T& t, const U& u) ;
             };template <typename T>
 template <typename U>
-inline void Foo<T>::bar(const T& t, const U& u) {}
+inline T Foo<T>::bar(const T& t, const U& u) { return {}; }
 )cpp",
           ""},
   };

clang/cmake/caches/CrossWinToARMLinux.cmake

@@ -119,7 +119,6 @@ if (NOT DEFINED CMAKE_BUILD_TYPE)
   set(CMAKE_BUILD_TYPE "Release" CACHE STRING "")
 endif()
 
-set(CMAKE_CROSSCOMPILING                    ON CACHE BOOL "")
 set(CMAKE_CL_SHOWINCLUDES_PREFIX            "Note: including file: " CACHE STRING "")
 # Required if COMPILER_RT_DEFAULT_TARGET_ONLY is ON
 set(CMAKE_C_COMPILER_TARGET                 "${TOOLCHAIN_TARGET_TRIPLE}" CACHE STRING "")
@@ -219,6 +218,11 @@ set(RUNTIMES_${TOOLCHAIN_TARGET_TRIPLE}_LIBCXX_CXX_ABI
 set(RUNTIMES_${TOOLCHAIN_TARGET_TRIPLE}_LIBCXX_ENABLE_NEW_DELETE_DEFINITIONS      ON CACHE BOOL "")
 # Merge libc++ and libc++abi libraries into the single libc++ library file.
 set(RUNTIMES_${TOOLCHAIN_TARGET_TRIPLE}_LIBCXX_ENABLE_STATIC_ABI_LIBRARY          ON CACHE BOOL "")
+# Forcely disable the libc++ benchmarks on Windows build hosts
+# (current benchmark test configuration does not support the cross builds there).
+if (WIN32)
+  set(RUNTIMES_${TOOLCHAIN_TARGET_TRIPLE}_LIBCXX_INCLUDE_BENCHMARKS               OFF CACHE BOOL "")
+endif(WIN32)
 
 # Avoid searching for the python3 interpreter during the runtimes configuration for the cross builds.
 # It starts searching the python3 package using the target's sysroot path, that usually is not compatible with the build host.

clang/docs/ReleaseNotes.rst

@@ -357,6 +357,7 @@ Non-comprehensive list of changes in this release
 
 - ``__builtin_reduce_add`` function can now be used in constant expressions.
 - ``__builtin_reduce_mul`` function can now be used in constant expressions.
+- ``__builtin_reduce_and`` function can now be used in constant expressions.
 
 New Compiler Flags
 ------------------
@@ -556,6 +557,8 @@ Improvements to Clang's diagnostics
       getS(); // Now diagnoses "Reason 2", previously diagnoses "Reason 1"
     }
 
+- Clang now diagnoses ``= delete("reason")`` extension warnings only in pedantic mode rather than on by default. (#GH109311).
+
 Improvements to Clang's time-trace
 ----------------------------------
 

clang/docs/ThreadSafetyAnalysis.rst

@@ -933,11 +933,25 @@ implementation.
     MutexLocker(Mutex *mu, defer_lock_t) EXCLUDES(mu) : mut(mu), locked(false) {}
 
     // Same as constructors, but without tag types. (Requires C++17 copy elision.)
-    static MutexLocker Lock(Mutex *mu) ACQUIRE(mu);
-    static MutexLocker Adopt(Mutex *mu) REQUIRES(mu);
-    static MutexLocker ReaderLock(Mutex *mu) ACQUIRE_SHARED(mu);
-    static MutexLocker AdoptReaderLock(Mutex *mu) REQUIRES_SHARED(mu);
-    static MutexLocker DeferLock(Mutex *mu) EXCLUDES(mu);
+    static MutexLocker Lock(Mutex *mu) ACQUIRE(mu) {
+      return MutexLocker(mu);
+    }
+
+    static MutexLocker Adopt(Mutex *mu) REQUIRES(mu) {
+      return MutexLocker(mu, adopt_lock);
+    }
+
+    static MutexLocker ReaderLock(Mutex *mu) ACQUIRE_SHARED(mu) {
+      return MutexLocker(mu, shared_lock);
+    }
+
+    static MutexLocker AdoptReaderLock(Mutex *mu) REQUIRES_SHARED(mu) {
+      return MutexLocker(mu, adopt_lock, shared_lock);
+    }
+
+    static MutexLocker DeferLock(Mutex *mu) EXCLUDES(mu) {
+      return MutexLocker(mu, defer_lock);
+    }
 
     // Release *this and all associated mutexes, if they are still held.
     // There is no warning if the scope was already unlocked before.