merge main into amd-staging

Author: ranapratap55

.github/workflows/libcxx-run-benchmarks.yml

@@ -64,17 +64,21 @@ jobs:
           path: repo # Avoid nuking the workspace, where we have the Python virtualenv
 
       - name: Run baseline
+        env:
+          BENCHMARKS: ${{ steps.vars.outputs.benchmarks }}
         run: |
           source .venv/bin/activate && cd repo
           python -m pip install -r libcxx/utils/requirements.txt
           baseline_commit=$(git merge-base ${{ steps.vars.outputs.pr_base }} ${{ steps.vars.outputs.pr_head }})
-          ./libcxx/utils/test-at-commit --commit ${baseline_commit} -B build/baseline -- -sv -j1 --param optimization=speed ${{ steps.vars.outputs.benchmarks }}
+          ./libcxx/utils/test-at-commit --commit ${baseline_commit} -B build/baseline -- -sv -j1 --param optimization=speed "$BENCHMARKS"
           ./libcxx/utils/consolidate-benchmarks build/baseline | tee baseline.lnt
 
       - name: Run candidate
+        env:
+          BENCHMARKS: ${{ steps.vars.outputs.benchmarks }}
         run: |
           source .venv/bin/activate && cd repo
-          ./libcxx/utils/test-at-commit --commit ${{ steps.vars.outputs.pr_head }} -B build/candidate -- -sv -j1 --param optimization=speed ${{ steps.vars.outputs.benchmarks }}
+          ./libcxx/utils/test-at-commit --commit ${{ steps.vars.outputs.pr_head }} -B build/candidate -- -sv -j1 --param optimization=speed "$BENCHMARKS"
           ./libcxx/utils/consolidate-benchmarks build/candidate | tee candidate.lnt
 
       - name: Compare baseline and candidate runs

.github/workflows/release-binaries.yml

@@ -58,7 +58,6 @@ jobs:
       ref: ${{ steps.vars.outputs.ref }}
       upload: ${{ steps.vars.outputs.upload }}
       target-cmake-flags: ${{ steps.vars.outputs.target-cmake-flags }}
-      ccache: ${{ steps.vars.outputs.ccache }}
       build-flang: ${{ steps.vars.outputs.build-flang }}
       release-binary-basename: ${{ steps.vars.outputs.release-binary-basename }}
       release-binary-filename: ${{ steps.vars.outputs.release-binary-filename }}
@@ -123,13 +122,6 @@ jobs:
         echo "release-binary-filename=$release_binary_basename.tar.xz" >> $GITHUB_OUTPUT
 
         target="$RUNNER_OS-$RUNNER_ARCH"
-        # The hendrikmuhs/ccache-action action does not support installing sccache
-        # on arm64 Linux.
-        if [ "$target" = "Linux-ARM64" ]; then
-          echo ccache=ccache >> $GITHUB_OUTPUT
-        else
-          echo ccache=sccache >> $GITHUB_OUTPUT
-        fi
 
         # The macOS builds try to cross compile some libraries so we need to
         # add extra CMake args to disable them.
@@ -222,8 +214,6 @@ jobs:
     - name: Configure
       id: build
       shell: bash
-      env:
-        CCACHE_BIN: ${{ needs.prepare.outputs.ccache }}
       run: |
         # There were some issues on the ARM64 MacOS runners with trying to build x86 object,
         # so we need to set some extra cmake flags to disable this.

clang/cmake/caches/Release.cmake

@@ -44,6 +44,19 @@ set(LLVM_RELEASE_ENABLE_LTO THIN CACHE STRING "")
 set(LLVM_RELEASE_ENABLE_PGO ON CACHE BOOL "")
 set(LLVM_RELEASE_ENABLE_RUNTIMES ${DEFAULT_RUNTIMES} CACHE STRING "")
 set(LLVM_RELEASE_ENABLE_PROJECTS ${DEFAULT_PROJECTS} CACHE STRING "")
+
+# This option enables linking stage2 clang statically with the runtimes
+# (libc++ and compiler-rt) from stage1.  In theory this will give the
+# binaries better performance and make them more portable.  However,
+# this configuration is not well tested and causes build failures with
+# the flang-rt tests cases, since the -stclib=libc++ flag does not
+# get propagated to the runtimes build.  There is also a separate
+# issue on Darwin where clang will use the local libc++ headers, but
+# link with system libc++ which can cause some incompatibilities.
+# See https://github.com/llvm/llvm-project/issues/77653
+# Because of these problems, this option will default to OFF.
+set(LLVM_RELEASE_ENABLE_LINK_LOCAL_RUNTIMES OFF CACHE BOOL "")
+
 # Note we don't need to add install here, since it is one of the pre-defined
 # steps.
 set(LLVM_RELEASE_FINAL_STAGE_TARGETS "clang;package;check-all;check-llvm;check-clang" CACHE STRING "")
@@ -55,8 +68,12 @@ set(CLANG_ENABLE_BOOTSTRAP ON CACHE BOOL "")
 
 set(STAGE1_PROJECTS "clang")
 
+# Need to build compiler-rt in order to use PGO for later stages.
+set(STAGE1_RUNTIMES "compiler-rt")
 # Build all runtimes so we can statically link them into the stage2 compiler.
-set(STAGE1_RUNTIMES "compiler-rt;libcxx;libcxxabi;libunwind")
+if(LLVM_RELEASE_ENABLE_LINK_LOCAL_RUNTIMES)
+  list(APPEND STAGE1_RUNTIMES "libcxx;libcxxabi;libunwind")
+endif()
 
 if (LLVM_RELEASE_ENABLE_PGO)
   list(APPEND STAGE1_PROJECTS "lld")
@@ -118,21 +135,25 @@ set_instrument_and_final_stage_var(LLVM_ENABLE_LTO "${LLVM_RELEASE_ENABLE_LTO}"
 if (LLVM_RELEASE_ENABLE_LTO)
   set_instrument_and_final_stage_var(LLVM_ENABLE_LLD "ON" BOOL)
 endif()
-set_instrument_and_final_stage_var(LLVM_ENABLE_LIBCXX "ON" BOOL)
-set_instrument_and_final_stage_var(LLVM_STATIC_LINK_CXX_STDLIB "ON" BOOL)
-set(RELEASE_LINKER_FLAGS "-rtlib=compiler-rt --unwindlib=libunwind")
-if(NOT ${CMAKE_HOST_SYSTEM_NAME} MATCHES "Darwin")
-  set(RELEASE_LINKER_FLAGS "${RELEASE_LINKER_FLAGS} -static-libgcc")
+if(LLVM_RELEASE_ENABLE_LINK_LOCAL_RUNTIMES)
+  set_instrument_and_final_stage_var(LLVM_ENABLE_LIBCXX "ON" BOOL)
+  set_instrument_and_final_stage_var(LLVM_STATIC_LINK_CXX_STDLIB "ON" BOOL)
+  set(RELEASE_LINKER_FLAGS "-rtlib=compiler-rt --unwindlib=libunwind")
+  if(NOT ${CMAKE_HOST_SYSTEM_NAME} MATCHES "Darwin")
+    set(RELEASE_LINKER_FLAGS "${RELEASE_LINKER_FLAGS} -static-libgcc")
+  endif()
 endif()
 
 # Set flags for bolt
 if (${CMAKE_HOST_SYSTEM_NAME} MATCHES "Linux")
   set(RELEASE_LINKER_FLAGS "${RELEASE_LINKER_FLAGS} -Wl,--emit-relocs,-znow")
 endif()
 
-set_instrument_and_final_stage_var(CMAKE_EXE_LINKER_FLAGS ${RELEASE_LINKER_FLAGS} STRING)
-set_instrument_and_final_stage_var(CMAKE_SHARED_LINKER_FLAGS ${RELEASE_LINKER_FLAGS} STRING)
-set_instrument_and_final_stage_var(CMAKE_MODULE_LINKER_FLAGS ${RELEASE_LINKER_FLAGS} STRING)
+if (RELEASE_LINKER_FLAGS)
+  set_instrument_and_final_stage_var(CMAKE_EXE_LINKER_FLAGS ${RELEASE_LINKER_FLAGS} STRING)
+  set_instrument_and_final_stage_var(CMAKE_SHARED_LINKER_FLAGS ${RELEASE_LINKER_FLAGS} STRING)
+  set_instrument_and_final_stage_var(CMAKE_MODULE_LINKER_FLAGS ${RELEASE_LINKER_FLAGS} STRING)
+endif()
 
 # Final Stage Config (stage2)
 set_final_stage_var(LLVM_ENABLE_RUNTIMES "${LLVM_RELEASE_ENABLE_RUNTIMES}" STRING)

clang/lib/CodeGen/BackendUtil.cpp

@@ -430,12 +430,6 @@ static bool initTargetOptions(const CompilerInstance &CI,
   Options.NoInfsFPMath = LangOpts.NoHonorInfs;
   Options.NoNaNsFPMath = LangOpts.NoHonorNaNs;
   Options.NoZerosInBSS = CodeGenOpts.NoZeroInitializedInBSS;
-  Options.UnsafeFPMath = LangOpts.AllowFPReassoc && LangOpts.AllowRecip &&
-                         LangOpts.NoSignedZero && LangOpts.ApproxFunc &&
-                         (LangOpts.getDefaultFPContractMode() ==
-                              LangOptions::FPModeKind::FPM_Fast ||
-                          LangOpts.getDefaultFPContractMode() ==
-                              LangOptions::FPModeKind::FPM_FastHonorPragmas);
 
   Options.BBAddrMap = CodeGenOpts.BBAddrMap;
   Options.BBSections =

clang/lib/Sema/HeuristicResolver.cpp

@@ -450,7 +450,12 @@ QualType HeuristicResolverImpl::resolveExprToType(const Expr *E) {
   if (const auto *CE = dyn_cast<CallExpr>(E)) {
     if (QualType Resolved = resolveTypeOfCallExpr(CE); !Resolved.isNull())
       return Resolved;
+
+    // Don't proceed to try resolveExprToDecls(), it would just call
+    // resolveTypeOfCallExpr() again.
+    return E->getType();
   }
+
   // Similarly, unwrapping a unary dereference operation does not work via
   // resolveExprToDecls.
   if (const auto *UO = dyn_cast<UnaryOperator>(E->IgnoreParenCasts())) {

clang/unittests/Sema/HeuristicResolverTest.cpp

@@ -524,6 +524,28 @@ TEST(HeuristicResolver, MemberExpr_HangIssue126536) {
       cxxDependentScopeMemberExpr(hasMemberName("foo")).bind("input"));
 }
 
+TEST(HeuristicResolver, MemberExpr_HangOnLongCallChain) {
+  const size_t CallChainLength = 50;
+  std::string Code = R"cpp(
+    template <typename T>
+    void foo(T t) {
+      t
+    )cpp";
+  for (size_t I = 0; I < CallChainLength; ++I)
+    Code.append(".method()\n");
+  Code.append(R"cpp(
+      .lastMethod();
+    }
+  )cpp");
+  // Test that resolution of a name whose base is a long call chain
+  // does not hang. Note that the hang for which this is a regression
+  // test is finite (exponential runtime in the length of the chain),
+  // so a "failure" here manifests as abnormally long runtime.
+  expectResolution(
+      Code, &HeuristicResolver::resolveMemberExpr,
+      cxxDependentScopeMemberExpr(hasMemberName("lastMethod")).bind("input"));
+}
+
 TEST(HeuristicResolver, MemberExpr_DefaultTemplateArgument) {
   std::string Code = R"cpp(
     struct Default {

libclc/cmake/modules/AddLibclc.cmake

@@ -392,7 +392,7 @@ function(add_libclc_builtin_set)
   list( PREPEND bytecode_files ${bytecode_ir_files} )
 
   if( NOT bytecode_files )
-    message(FATAL_ERROR "Cannot create an empty builtins library")
+    message(FATAL_ERROR "Cannot create an empty builtins library for ${ARG_ARCH_SUFFIX}")
   endif()
 
   set( builtins_link_lib_tgt builtins.link.${ARG_ARCH_SUFFIX} )

libcxx/include/__cxx03/__algorithm/for_each.h

@@ -14,15 +14,11 @@
 #include <__cxx03/__config>
 #include <__cxx03/__iterator/segmented_iterator.h>
 #include <__cxx03/__type_traits/enable_if.h>
-#include <__cxx03/__utility/move.h>
 
 #if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER)
 #  pragma GCC system_header
 #endif
 
-_LIBCPP_PUSH_MACROS
-#include <__cxx03/__undef_macros>
-
 _LIBCPP_BEGIN_NAMESPACE_STD
 
 template <class _InputIterator, class _Function>
@@ -34,6 +30,4 @@ _LIBCPP_HIDE_FROM_ABI _Function for_each(_InputIterator __first, _InputIterator
 
 _LIBCPP_END_NAMESPACE_STD
 
-_LIBCPP_POP_MACROS
-
 #endif // _LIBCPP___CXX03___ALGORITHM_FOR_EACH_H

lldb/test/API/functionalities/thread/finish-from-empty-func/TestEmptyFuncThreadStepOut.py

@@ -13,12 +13,31 @@ class FinishFromEmptyFunctionTestCase(TestBase):
 
     @skipIf(compiler="clang", compiler_version=['<', '17.0'])
     def test_finish_from_empty_function(self):
-        """Test that when stopped at a breakpoint in an empty function, finish leaves it correctly."""
+        """Test that when stopped at a breakpoint located at the last instruction
+        of a function, finish leaves it correctly."""
         self.build()
-        exe = self.getBuildArtifact("a.out")
-        target, process, thread, _ = lldbutil.run_to_name_breakpoint(
-            self, "done", exe_name=exe
+        target, _, thread, _ = lldbutil.run_to_source_breakpoint(
+            self, "// Set breakpoint here", lldb.SBFileSpec("main.c")
         )
+        # Find the address of the last instruction of 'done()' and set a breakpoint there.
+        # Even though 'done()' is empty, it may contain prologue and epilogue code, so
+        # simply setting a breakpoint at the function can place it before 'ret'.
+        error = lldb.SBError()
+        ret_bp_addr = lldb.SBAddress()
+        while True:
+            thread.StepInstruction(False, error)
+            self.assertTrue(error.Success())
+            frame = thread.GetSelectedFrame()
+            if "done" in frame.GetFunctionName():
+                ret_bp_addr = frame.GetPCAddress()
+            elif ret_bp_addr.IsValid():
+                # The entire function 'done()' has been stepped through, so 'ret_bp_addr'
+                # now contains the address of its last instruction, i.e. 'ret'.
+                break
+        ret_bp = target.BreakpointCreateByAddress(ret_bp_addr.GetLoadAddress(target))
+        self.assertTrue(ret_bp.IsValid())
+        # Resume the execution and hit the new breakpoint.
+        self.runCmd("cont")
         if self.TraceOn():
             self.runCmd("bt")
 
@@ -29,7 +48,6 @@ def test_finish_from_empty_function(self):
         )
         self.assertTrue(safety_bp.IsValid())
 
-        error = lldb.SBError()
         thread.StepOut(error)
         self.assertTrue(error.Success())
 

lldb/test/API/functionalities/thread/finish-from-empty-func/main.c

@@ -2,6 +2,7 @@
 void done() {}
 int main() {
   puts("in main");
+  done(); // Set breakpoint here
   done();
   puts("leaving main");
   return 0;