Merge pull request #807 from laurynas-biveinis/llvm-21

Add support for LLVM 21

Author: laurynas-biveinis

.clang-tidy

@@ -9,16 +9,17 @@
 #   unfixable signatures
 # - cppcoreguidelines-avoid-c-arrays: duplicated by modernize-avoid-c-arrays
 # - cppcoreguidelines-macro-usage: does not respect __LINE__ in macro definition
-# - cppcoreguidelines-pro-bounds-pointer-arithmetic: because leaf nodes are
-#   std::byte arrays
 # - cppcoreguidelines-pro-type-const-cast: because VALGRIND_MALLOCLIKE_BLOCK
 #   expands a to C-style cast, and we have -Wold-style-cast anyway
 # - hicpp-avoid-c-arrays: duplicated by modernize-avoid-c-arrays
 # - hicpp-explicit-conversions: duplicated by google-explicit-constructor
 # - hicpp-no-assembler: Valgrind client requests
 # - modernize-use-equals-default: until foo() noexcept = default is accepted by
 #   clang
-Checks: '*,-altera-id-dependent-backward-branch,-altera-struct-pack-align,-altera-unroll-loops,-boost-use-ranges,-bugprone-easily-swappable-parameters,-bugprone-use-after-move,-clang-diagnostic-error,-clang-analyzer-core.CallAndMessage,-clang-analyzer-core.NonNullParamChecker,-clang-analyzer-cplusplus.Move,-clang-analyzer-unix.Malloc,-cppcoreguidelines-avoid-c-arrays,-cppcoreguidelines-avoid-magic-numbers,-cppcoreguidelines-init-variables,-cppcoreguidelines-macro-usage,-cppcoreguidelines-non-private-member-variables-in-classes,-cppcoreguidelines-pro-bounds-array-to-pointer-decay,-cppcoreguidelines-pro-bounds-constant-array-index,-cppcoreguidelines-pro-bounds-pointer-arithmetic,-cppcoreguidelines-pro-type-const-cast,-cppcoreguidelines-pro-type-cstyle-cast,-cppcoreguidelines-pro-type-member-init,-cppcoreguidelines-pro-type-reinterpret-cast,-cppcoreguidelines-pro-type-static-cast-downcast,-cppcoreguidelines-pro-type-union-access,-fuchsia-default-arguments-calls,-fuchsia-default-arguments-declarations,-fuchsia-overloaded-operator,-google-readability-braces-around-statements,-google-runtime-references,-hicpp-avoid-c-arrays,-hicpp-braces-around-statements,-hicpp-explicit-conversions,-hicpp-invalid-access-moved,-hicpp-member-init,-hicpp-named-parameter,-hicpp-no-array-decay,-hicpp-no-assembler,-hicpp-use-equals-default,-llvm-include-order,-llvmlibc*,-misc-no-recursion,-misc-non-private-member-variables-in-classes,-modernize-use-equals-default,-modernize-use-trailing-return-type,-portability-simd-intrinsics,-readability-braces-around-statements,-readability-function-cognitive-complexity,-readability-identifier-length,-readability-named-parameter,-readability-magic-numbers'
+# - modernize-use-scoped-lock: std::lock_guard is perfectly fine for its use
+#   cases
+# - llvm-prefer-static-over-anonymous-namespace: I prefer anonymous namespace
+Checks: '*,-altera-id-dependent-backward-branch,-altera-struct-pack-align,-altera-unroll-loops,-boost-use-ranges,-bugprone-easily-swappable-parameters,-bugprone-use-after-move,-clang-diagnostic-error,-clang-analyzer-core.CallAndMessage,-clang-analyzer-core.NonNullParamChecker,-clang-analyzer-cplusplus.Move,-clang-analyzer-unix.Malloc,-cppcoreguidelines-avoid-c-arrays,-cppcoreguidelines-avoid-magic-numbers,-cppcoreguidelines-init-variables,-cppcoreguidelines-macro-usage,-cppcoreguidelines-non-private-member-variables-in-classes,-cppcoreguidelines-pro-bounds-array-to-pointer-decay,-cppcoreguidelines-pro-bounds-constant-array-index,-cppcoreguidelines-pro-type-const-cast,-cppcoreguidelines-pro-type-cstyle-cast,-cppcoreguidelines-pro-type-member-init,-cppcoreguidelines-pro-type-reinterpret-cast,-cppcoreguidelines-pro-type-static-cast-downcast,-cppcoreguidelines-pro-type-union-access,-fuchsia-default-arguments-calls,-fuchsia-default-arguments-declarations,-fuchsia-overloaded-operator,-google-readability-braces-around-statements,-google-runtime-references,-hicpp-avoid-c-arrays,-hicpp-braces-around-statements,-hicpp-explicit-conversions,-hicpp-invalid-access-moved,-hicpp-member-init,-hicpp-named-parameter,-hicpp-no-array-decay,-hicpp-no-assembler,-hicpp-use-equals-default,-llvm-include-order,-llvmlibc*,-llvm-prefer-static-over-anonymous-namespace,-misc-no-recursion,-misc-non-private-member-variables-in-classes,-modernize-use-equals-default,-modernize-use-scoped-lock,-modernize-use-trailing-return-type,-portability-simd-intrinsics,-readability-braces-around-statements,-readability-function-cognitive-complexity,-readability-identifier-length,-readability-named-parameter,-readability-magic-numbers'
 WarningsAsErrors: '*'
 CheckOptions:
   - key: performance-unnecessary-value-param.AllowedTypes

.github/workflows/build.yml

@@ -99,59 +99,59 @@ jobs:
             os: ubuntu-latest
             COMPILER: gcc
 
-          - name: clang 20 Release (Linux Intel)
+          - name: clang 21 Release (Linux Intel)
             os: ubuntu-latest
             BUILD_TYPE: Release
             COMPILER: clang
 
-          - name: clang 20 Release with ASan (Linux Intel)
+          - name: clang 21 Release with ASan (Linux Intel)
             os: ubuntu-latest
             BUILD_TYPE: Release
             SANITIZE_ADDRESS: ON
             COMPILER: clang
 
-          - name: clang 20 Release with TSan (Linux Intel)
+          - name: clang 21 Release with TSan (Linux Intel)
             os: ubuntu-latest
             BUILD_TYPE: Release
             SANITIZE_THREAD: ON
             COMPILER: clang
 
-          - name: clang 20 Release with UBSan (Linux Intel)
+          - name: clang 21 Release with UBSan (Linux Intel)
             os: ubuntu-latest
             BUILD_TYPE: Release
             SANITIZE_UB: ON
             COMPILER: clang
 
-          - name: clang 20 Debug (Linux Intel)
+          - name: clang 21 Debug (Linux Intel)
             os: ubuntu-latest
             BUILD_TYPE: Debug
             COMPILER: clang
 
-          - name: clang 20 Debug with ASan (Linux Intel)
+          - name: clang 21 Debug with ASan (Linux Intel)
             os: ubuntu-latest
             BUILD_TYPE: Debug
             SANITIZE_ADDRESS: ON
             COMPILER: clang
 
-          - name: clang 20 Debug with TSan (Linux Intel)
+          - name: clang 21 Debug with TSan (Linux Intel)
             os: ubuntu-latest
             BUILD_TYPE: Debug
             SANITIZE_THREAD: ON
             COMPILER: clang
 
-          - name: clang 20 Debug with UBSan (Linux Intel)
+          - name: clang 21 Debug with UBSan (Linux Intel)
             os: ubuntu-latest
             BUILD_TYPE: Debug
             SANITIZE_UB: ON
             COMPILER: clang
 
-          - name: clang 20 Release static analysis (Linux Intel)
+          - name: clang 21 Release static analysis (Linux Intel)
             os: ubuntu-latest
             BUILD_TYPE: Release
             COMPILER: clang
             STATIC_ANALYSIS: ON
 
-          - name: clang 20 Debug static analysis (Linux Intel)
+          - name: clang 21 Debug static analysis (Linux Intel)
             os: ubuntu-latest
             BUILD_TYPE: Debug
             COMPILER: clang
@@ -313,41 +313,41 @@ jobs:
             COMPILER: gcc
             SANITIZE_UB: ON
 
-          - name: clang 20 Release (Linux ARM64)
+          - name: clang 21 Release (Linux ARM64)
             os: ubuntu-24.04-arm
             BUILD_TYPE: Release
             COMPILER: clang
 
-          - name: clang 20 Release with TSan (Linux ARM64)
+          - name: clang 21 Release with TSan (Linux ARM64)
             os: ubuntu-24.04-arm
             BUILD_TYPE: Release
             COMPILER: clang
             SANITIZE_THREAD: ON
 
-          - name: clang 20 Release with UBSan (Linux ARM64)
+          - name: clang 21 Release with UBSan (Linux ARM64)
             os: ubuntu-24.04-arm
             BUILD_TYPE: Release
             COMPILER: clang
             SANITIZE_UB: ON
 
-          - name: clang 20 Debug (Linux ARM64)
+          - name: clang 21 Debug (Linux ARM64)
             os: ubuntu-24.04-arm
             BUILD_TYPE: Debug
             COMPILER: clang
 
-          - name: clang 20 Debug with ASan (Linux ARM64)
+          - name: clang 21 Debug with ASan (Linux ARM64)
             os: ubuntu-24.04-arm
             BUILD_TYPE: Debug
             COMPILER: clang
             SANITIZE_ADDRESS: ON
 
-          - name: clang 20 Debug with TSan (Linux ARM64)
+          - name: clang 21 Debug with TSan (Linux ARM64)
             os: ubuntu-24.04-arm
             BUILD_TYPE: Debug
             COMPILER: clang
             SANITIZE_THREAD: ON
 
-          - name: clang 20 Debug with UBSan (Linux ARM64)
+          - name: clang 21 Debug with UBSan (Linux ARM64)
             os: ubuntu-24.04-arm
             BUILD_TYPE: Debug
             COMPILER: clang
@@ -389,25 +389,25 @@ jobs:
           curl 'https://apt.llvm.org/llvm-snapshot.gpg.key' \
             | sudo apt-key add -
           sudo add-apt-repository -y \
-            "deb http://apt.llvm.org/noble/ llvm-toolchain-noble-20 main"
+            "deb http://apt.llvm.org/noble/ llvm-toolchain-noble-21 main"
           sudo apt-get update
-          sudo apt-get install -y clang-20 iwyu
+          sudo apt-get install -y clang-21 iwyu
         if: runner.os == 'Linux' && env.COMPILER == 'clang'
 
       - name: Setup dependencies for Linux LLVM (Release)
-        run: sudo apt-get install -y libomp5-20 llvm-20 lld-20
+        run: sudo apt-get install -y libomp5 llvm-21 lld-21
         if: >
           runner.os == 'Linux' && env.COMPILER == 'clang'
           && env.BUILD_TYPE == 'Release'
 
       - name: Setup dependencies for Linux LLVM (static analysis)
-        run: sudo apt-get install -y clang-tools-20
+        run: sudo apt-get install -y clang-tools-21
         if: >
           runner.os == 'Linux' && env.COMPILER == 'clang'
           && env.STATIC_ANALYSIS == 'ON'
 
       - name: Setup dependencies for Linux LLVM (not static analysis)
-        run: sudo apt-get install -y clang-tidy-20
+        run: sudo apt-get install -y clang-tidy-21
         if: >
           runner.os == 'Linux' && env.COMPILER == 'clang'
           && env.STATIC_ANALYSIS != 'ON'
@@ -444,7 +444,7 @@ jobs:
             export CC=gcc-$V
             export CXX=g++-$V
           elif [[ $COMPILER == "clang" ]]; then
-            V=20
+            V=21
             export CC=clang-$V
             export CXX=clang++-$V
             if [[ $BUILD_TYPE == "Release" ]]; then
@@ -497,7 +497,7 @@ jobs:
       - name: clang static analysis
         working-directory: ${{github.workspace}}/build
         run: |
-          /usr/bin/scan-build-20 --status-bugs -stats -analyze-headers \
+          /usr/bin/scan-build-21 --status-bugs -stats -analyze-headers \
             --force-analyze-debug-code make -j3 -k;
         if: env.STATIC_ANALYSIS == 'ON' && env.COMPILER == 'clang'
 

.github/workflows/old-compilers.yml

@@ -469,6 +469,52 @@ jobs:
             COMPILER: clang
             VERSION: 19
 
+          - name: clang 20 Release
+            BUILD_TYPE: Release
+            COMPILER: clang
+            VERSION: 20
+
+          - name: clang 20 Release with ASan
+            BUILD_TYPE: Release
+            SANITIZE_ADDRESS: ON
+            COMPILER: clang
+            VERSION: 20
+
+          - name: clang 20 Release with TSan
+            BUILD_TYPE: Release
+            SANITIZE_THREAD: ON
+            COMPILER: clang
+            VERSION: 20
+
+          - name: clang 20 Release with UBSan
+            BUILD_TYPE: Release
+            SANITIZE_UB: ON
+            COMPILER: clang
+            VERSION: 20
+
+          - name: clang 20 Debug
+            BUILD_TYPE: Debug
+            COMPILER: clang
+            VERSION: 20
+
+          - name: clang 20 Debug with ASan
+            BUILD_TYPE: Debug
+            SANITIZE_ADDRESS: ON
+            COMPILER: clang
+            VERSION: 20
+
+          - name: clang 20 Debug with TSan
+            BUILD_TYPE: Debug
+            SANITIZE_THREAD: ON
+            COMPILER: clang
+            VERSION: 20
+
+          - name: clang 20 Debug with UBSan
+            BUILD_TYPE: Debug
+            SANITIZE_UB: ON
+            COMPILER: clang
+            VERSION: 20
+
           - name: GCC 10 Release
             BUILD_TYPE: Release
             COMPILER: gcc

CMakeLists.txt

@@ -88,6 +88,8 @@ set(CLANG_LT_13_CXX_WARNING_FLAGS "-Wreserved-id-macro")
 set(CLANG_GE_13_CXX_WARNING_FLAGS "-Wreserved-identifier")
 # Do not enable clang 16+ -Wunsafe-buffer-usage because it warns on all memcpy
 # calls and our pointer- and span-like class implementations.
+set(CLANG_GE_21_CXX_WARNING_FLAGS "-Wnrvo" "-Wunique-object-duplication"
+  "-Wshift-bool" "-Wc99-designator" "-Wms-bitfield-padding")
 
 set(GCC_CXX_WARNING_FLAGS
   # Warning groups
@@ -296,10 +298,17 @@ if(SANITIZE_UB)
     if(is_clang AND CMAKE_CXX_COMPILER_VERSION VERSION_GREATER_EQUAL 19.0)
       list(APPEND SANITIZER_CXX_FLAGS "-fsanitize=implicit-bitfield-conversion")
     endif()
+    if(is_clang AND CMAKE_CXX_COMPILER_VERSION VERSION_GREATER_EQUAL 21.0)
+      list(APPEND SANITIZER_CXX_FLAGS "-fsanitize=vptr")
+    endif()
   endif()
 
   set(SANITIZER_LD_FLAGS "-fsanitize=undefined")
 
+  if(is_clang AND CMAKE_CXX_COMPILER_VERSION VERSION_GREATER_EQUAL 21.0)
+    list(APPEND SANITIZER_LD_FLAGS "-fsanitize=vptr")
+  endif()
+
   string(CONCAT UBSAN_ENV "UBSAN_OPTIONS="
     "print_stacktrace=1:halt_on_error=1:abort_on_error=1")
 
@@ -402,9 +411,13 @@ set(cxx_ge_12 "$<VERSION_GREATER_EQUAL:$<CXX_COMPILER_VERSION>,12.0>")
 set(cxx_lt_13 "$<VERSION_LESS:$<CXX_COMPILER_VERSION>,13.0>")
 set(cxx_ge_13 "$<VERSION_GREATER_EQUAL:$<CXX_COMPILER_VERSION>,13.0>")
 set(cxx_ge_14 "$<VERSION_GREATER_EQUAL:$<CXX_COMPILER_VERSION>,14.0>")
+set(cxx_ge_21 "$<VERSION_GREATER_EQUAL:$<CXX_COMPILER_VERSION>,21.0>")
 set(is_clang_lt_13_not_windows "$<AND:${is_clang_not_windows},${cxx_lt_13}>")
 set(is_clang_ge_13_not_windows "$<AND:${is_clang_not_windows},${cxx_ge_13}>")
 set(is_clang_ge_14_not_windows "$<AND:${is_clang_not_windows},${cxx_ge_14}>")
+set(is_clang_ge_21_not_windows "$<AND:${is_clang_not_windows},${cxx_ge_21}>")
+set(is_darwin_clang_ge_21_x86_64 "$<AND:$<PLATFORM_ID:Darwin>,${is_clang_ge_21_not_windows},${is_not_windows_x86_64}>")
+set(is_darwin_clang_ge_21_arm64 "$<AND:$<PLATFORM_ID:Darwin>,${is_clang_ge_21_not_windows},$<NOT:${is_not_windows_x86_64}>>")
 set(is_gxx_ge_11 "$<AND:${is_gxx_genex},${cxx_ge_11}>")
 set(is_gxx_ge_12 "$<AND:${is_gxx_genex},${cxx_ge_12}>")
 set(is_gxx_ge_14 "$<AND:${is_gxx_genex},${cxx_ge_14}>")
@@ -616,6 +629,10 @@ else()
   else()
     message(STATUS "clang-tidy found: ${CLANG_TIDY_EXE}")
     set(DO_CLANG_TIDY "${CLANG_TIDY_EXE}" "-p=${CMAKE_BINARY_DIR}")
+    # Add suppression for unknown warning options when clang-tidy might not
+    # recognize all compiler warning flags (e.g., when scan-build from LLVM 21
+    # wraps the build)
+    list(APPEND DO_CLANG_TIDY "--extra-arg=-Wno-unknown-warning-option")
   endif()
 endif()
 
@@ -674,7 +691,9 @@ if(IWYU)
         IWYU_VERSION_OUTPUT)
       message(STATUS
         "include-what-you-use found: ${IWYU_EXE}, --version: ${IWYU_VERSION_OUTPUT}")
-      set(DO_IWYU "${IWYU_EXE}")
+      # IWYU is not necessarily based on the same version clang is and might not
+      # understand newer options
+      set(DO_IWYU "${IWYU_EXE};-Wno-unknown-warning-option")
     endif()
   endif()
 endif()
@@ -709,6 +728,7 @@ function(COMMON_TARGET_PROPERTIES TARGET)
     "$<$<AND:${is_standalone},${is_any_clang_genex},${is_not_windows}>:${CLANG_CXX_WARNING_FLAGS}>"
     "$<$<AND:${is_standalone},${is_clang_lt_13_not_windows}>:${CLANG_LT_13_CXX_WARNING_FLAGS}>"
     "$<$<AND:${is_standalone},${is_clang_ge_13_not_windows}>:${CLANG_GE_13_CXX_WARNING_FLAGS}>"
+    "$<$<AND:${is_standalone},${is_clang_ge_21_not_windows}>:${CLANG_GE_21_CXX_WARNING_FLAGS}>"
     "$<$<AND:${is_standalone},${is_gxx_genex}>:${GCC_CXX_WARNING_FLAGS}>"
     "$<$<AND:${is_standalone},${is_gxx_ge_11}>:${GCC_GE_11_CXX_WARNING_FLAGS}>"
     "$<$<AND:${is_standalone},${is_gxx_ge_12}>:${GCC_GE_12_CXX_WARNING_FLAGS}>"
@@ -721,6 +741,10 @@ function(COMMON_TARGET_PROPERTIES TARGET)
   target_link_options(${TARGET} INTERFACE "$<${coverage_on}:--coverage>")
   target_link_options(${TARGET} PRIVATE
     "$<${is_apple_clang_genex}:-Wl,-no_warn_duplicate_libraries>"
+    "$<${is_darwin_clang_ge_21_x86_64}:-L/usr/local/opt/llvm/lib/c++>"
+    "$<${is_darwin_clang_ge_21_x86_64}:-Wl,-rpath,/usr/local/opt/llvm/lib/c++>"
+    "$<${is_darwin_clang_ge_21_arm64}:-L/opt/homebrew/opt/llvm/lib/c++>"
+    "$<${is_darwin_clang_ge_21_arm64}:-Wl,-rpath,/opt/homebrew/opt/llvm/lib/c++>"
     "${SANITIZER_LD_FLAGS}"
   )
   target_link_libraries(${TARGET} PRIVATE

art_internal_impl.hpp

@@ -1444,6 +1444,7 @@ class basic_inode_4 : public basic_inode_4_parent<ArtPolicy> {
         keys.byte_array.cbegin(), keys.byte_array.cbegin() + children_count_));
   }
 
+  UNODB_DETAIL_DISABLE_CLANG_21_WARNING("-Wnrvo")
   [[nodiscard]] constexpr auto leave_last_child(std::uint8_t child_to_delete,
                                                 db_type &db_instance) noexcept {
     UNODB_DETAIL_ASSERT(this->is_min_size());
@@ -1465,6 +1466,7 @@ class basic_inode_4 : public basic_inode_4_parent<ArtPolicy> {
     }
     return child_to_leave_ptr;
   }
+  UNODB_DETAIL_RESTORE_CLANG_21_WARNINGS()
 
   [[nodiscard, gnu::pure]] find_result find_child(std::byte key_byte) noexcept {
 #ifdef UNODB_DETAIL_X86_64

benchmark/micro_benchmark_node_utils.hpp

@@ -229,8 +229,8 @@ number_to_full_node_tree_with_gaps_key(std::uint64_t i) noexcept {
 // Key vectors
 
 template <typename NumberToKeyFn>
-[[nodiscard]] auto generate_keys_to_limit(std::uint64_t key_limit,
-                                          NumberToKeyFn number_to_key_fn) {
+[[nodiscard]] std::vector<std::uint64_t> generate_keys_to_limit(
+    std::uint64_t key_limit, NumberToKeyFn number_to_key_fn) {
   std::vector<std::uint64_t> result;
   std::uint64_t i = 0;
   while (true) {

fuzz_deepstate/test_art_fuzz_deepstate.cpp

@@ -201,6 +201,7 @@ void op_with_oom_test(oracle_type &oracle, std::vector<std::uint64_t> &keys,
 
 UNODB_START_DEEPSTATE_TESTS()
 
+// NOLINTBEGIN(clang-analyzer-security.ArrayBound)
 TEST(ART, DeepStateFuzz) {
   const auto limit_max_key = DeepState_Bool();
   const auto max_key_value =
@@ -308,3 +309,4 @@ TEST(ART, DeepStateFuzz) {
 #endif  // UNODB_DETAIL_WITH_STATS
   ASSERT(test_db.empty());
 }
+// NOLINTEND(clang-analyzer-security.ArrayBound)

fuzz_deepstate/test_qsbr_fuzz_deepstate.cpp

@@ -611,6 +611,7 @@ void do_or_dispatch_op(std::size_t thread_i, thread_operation op) {
 
 UNODB_START_DEEPSTATE_TESTS()
 
+// NOLINTBEGIN(clang-analyzer-security.ArrayBound)
 TEST(QSBR, DeepStateFuzz) {
   const auto test_length = DeepState_ShortInRange(0, 2000);
   LOG(TRACE) << "Test length " << test_length;
@@ -758,5 +759,6 @@ TEST(QSBR, DeepStateFuzz) {
   ASSERT(unodb::this_thread().previous_interval_requests_empty());
   ASSERT(unodb::this_thread().current_interval_requests_empty());
 }
+// NOLINTEND(clang-analyzer-security.ArrayBound)
 
 }  // namespace