apache · hubgeter · Jul 28, 2026 · Jul 30, 2026 · github-actions · Jul 30, 2026
diff --git a/.github/workflows/be-ut-mac.yml b/.github/workflows/be-ut-mac.yml
@@ -48,17 +48,25 @@ jobs:
               - 'be/**'
               - 'gensrc/proto/**'
               - 'gensrc/thrift/**'
+            thirdparty_changes:
+              - 'thirdparty/**'
 
       - name: Ccache ${{ github.ref }}
-        if: ${{ github.event_name == 'schedule' || steps.filter.outputs.be_changes == 'true' }}
+        if: |
+          github.event_name == 'schedule' ||
+          steps.filter.outputs.be_changes == 'true' ||
+          steps.filter.outputs.thirdparty_changes == 'true'
         uses: ./.github/actions/ccache-action
         with:
           key: BE-UT-macOS
           max-size: "5G"
           restore-keys: BE-UT-macOS-
 
       - name: Build BE ${{ github.ref }}
-        if: ${{ github.event_name == 'schedule' || steps.filter.outputs.be_changes == 'true' }}
+        if: |
+          github.event_name == 'schedule' ||
+          steps.filter.outputs.be_changes == 'true' ||
+          steps.filter.outputs.thirdparty_changes == 'true'
         run: |
           cellars=(
             'm4'
@@ -101,6 +109,17 @@ jobs:
               -o doris-thirdparty-prebuilt-darwin-arm64.tar.xz
           fi
           tar -xvf doris-thirdparty-prebuilt-darwin-arm64.tar.xz
+
+          # A thirdparty upgrade can make the shared prebuilt archive stale.
+          # Rebuild the changed Arrow/Paimon stack before compiling BE so this
+          # job validates the versions selected by the PR.
+          if [[ "${{ steps.filter.outputs.thirdparty_changes }}" == "true" ]]; then
+            curl -L https://github.com/apache/doris-thirdparty/releases/download/automation/doris-thirdparty-source.tgz \
+              -o doris-thirdparty-source.tgz
+            tar -zxvf doris-thirdparty-source.tgz
+            export MACOSX_DEPLOYMENT_TARGET=12.0
+            ./build-thirdparty.sh -j "$(nproc)" arrow paimon_cpp
+          fi
           popd
 
           # macos-15 runners are Apple Silicon (arm64), so the JDK env var is

diff --git a/.github/workflows/build-thirdparty.yml b/.github/workflows/build-thirdparty.yml
@@ -195,9 +195,9 @@ jobs:
 
           # Install specific version of cmake
           brew unlink cmake || true
-          wget https://github.com/Kitware/CMake/releases/download/v3.22.1/cmake-3.22.1-macos-universal.tar.gz
-          tar -xzf cmake-3.22.1-macos-universal.tar.gz
-          sudo cp -r cmake-3.22.1-macos-universal/CMake.app/Contents/* /usr/local/
+          wget https://github.com/Kitware/CMake/releases/download/v3.25.3/cmake-3.25.3-macos-universal.tar.gz
+          tar -xzf cmake-3.25.3-macos-universal.tar.gz
+          sudo cp -r cmake-3.25.3-macos-universal/CMake.app/Contents/* /usr/local/
           cmake --version
 
       - name: Build
@@ -260,9 +260,9 @@ jobs:
 
           # Install specific version of cmake
           brew unlink cmake || true
-          wget https://github.com/Kitware/CMake/releases/download/v3.22.1/cmake-3.22.1-macos-universal.tar.gz
-          tar -xzf cmake-3.22.1-macos-universal.tar.gz
-          sudo cp -r cmake-3.22.1-macos-universal/CMake.app/Contents/* /usr/local/
+          wget https://github.com/Kitware/CMake/releases/download/v3.25.3/cmake-3.25.3-macos-universal.tar.gz
+          tar -xzf cmake-3.25.3-macos-universal.tar.gz
+          sudo cp -r cmake-3.25.3-macos-universal/CMake.app/Contents/* /usr/local/
           cmake --version
 
       - name: Build
@@ -278,4 +278,3 @@ jobs:
           cd thirdparty
           #./build-thirdparty.sh -j "$(nproc)"
           ./build-thirdparty.sh -j 2
-
diff --git a/be/CMakeLists.txt b/be/CMakeLists.txt
@@ -666,7 +666,7 @@ endif()
 set(PAIMON_FACTORY_REGISTRY_LIBS)
 if (ENABLE_PAIMON_CPP)
     # Plan B: Doris Arrow is now built with COMPUTE/DATASET/ACERO/FILESYSTEM,
-    # so arrow, arrow_dataset, arrow_acero are all in COMMON_THIRDPARTY via
+    # so arrow, arrow_compute, arrow_dataset, arrow_acero are all in COMMON_THIRDPARTY via
     # thirdparty.cmake.  paimon-cpp reuses the same Arrow (no paimon_deps).
     # No dual-stack selection needed — single Arrow for everything.
 

diff --git a/be/cmake/thirdparty.cmake b/be/cmake/thirdparty.cmake
@@ -105,6 +105,7 @@ add_thirdparty(brotlidec LIB64)
 add_thirdparty(brotlienc LIB64)
 add_thirdparty(zstd LIB64)
 add_thirdparty(arrow LIB64)
+add_thirdparty(arrow_compute LIB64)
 add_thirdparty(arrow_flight LIB64)
 add_thirdparty(arrow_flight_sql LIB64)
 add_thirdparty(arrow_dataset LIB64)

diff --git a/be/src/format/table/iceberg/arrow_schema_util.cpp b/be/src/format/table/iceberg/arrow_schema_util.cpp
@@ -97,7 +97,7 @@ Status ArrowSchemaUtil::convert_to(const iceberg::NestedField& field,
 
     case iceberg::TypeID::DECIMAL: {
         auto* dt = dynamic_cast<DecimalType*>(field.field_type());
-        arrow_type = arrow::decimal(dt->get_precision(), dt->get_scale());
+        arrow_type = arrow::decimal128(dt->get_precision(), dt->get_scale());
         break;
     }
 

diff --git a/be/test/core/data_type_serde/data_type_serde_decimal_test.cpp b/be/test/core/data_type_serde/data_type_serde_decimal_test.cpp
@@ -309,7 +309,7 @@ TEST_F(DataTypeDecimalSerDeTest, serdes) {
 // Run with UBSan enabled to catch misalignment errors.
 TEST_F(DataTypeDecimalSerDeTest, ArrowMemNotAligned) {
     // 1.Prepare the data.
-    arrow::Decimal128Builder builder(arrow::decimal(38, 30));
+    arrow::Decimal128Builder builder(arrow::decimal128(38, 30));
     std::vector<std::string> decimal_strings = {"12345.67", "89.10", "1112.13", "1415.16",
                                                 "1718.19"};
 
@@ -335,7 +335,7 @@ TEST_F(DataTypeDecimalSerDeTest, ArrowMemNotAligned) {
     // 4. Create Arrow array with unaligned memory
     auto unaligned_buffer = arrow::Buffer::Wrap(unaligned_data, num_elements * element_size);
 
-    auto arr = std::make_shared<arrow::DecimalArray>(arrow::decimal(38, 30), num_elements,
+    auto arr = std::make_shared<arrow::DecimalArray>(arrow::decimal128(38, 30), num_elements,
                                                      unaligned_buffer);
 
     const auto* raw_values_ptr = arr->raw_values();
@@ -432,7 +432,7 @@ uint8_t* find_misaligned_address(std::vector<uint8_t>& storage) {
 }
 
 TEST_F(DataTypeDecimalSerDeTest, ArrowMemNotAlignedDecimalV2TypeAlignment) {
-    arrow::Decimal128Builder builder(arrow::decimal(27, 9));
+    arrow::Decimal128Builder builder(arrow::decimal128(27, 9));
     std::vector<std::string> decimal_strings = {"1.000000000", "-123456789012345678.123456789"};
 
     for (const auto& str : decimal_strings) {
@@ -453,7 +453,7 @@ TEST_F(DataTypeDecimalSerDeTest, ArrowMemNotAlignedDecimalV2TypeAlignment) {
     memcpy(unaligned_data, original_data, num_elements * element_size);
 
     auto unaligned_buffer = arrow::Buffer::Wrap(unaligned_data, num_elements * element_size);
-    auto arr = std::make_shared<arrow::DecimalArray>(arrow::decimal(27, 9), num_elements,
+    auto arr = std::make_shared<arrow::DecimalArray>(arrow::decimal128(27, 9), num_elements,
                                                      unaligned_buffer);
 
     const auto* raw_values_ptr = arr->raw_values();

diff --git a/be/test/format_v2/table/remote_doris_reader_test.cpp b/be/test/format_v2/table/remote_doris_reader_test.cpp
@@ -213,7 +213,7 @@ TFileRangeDesc remote_doris_range(const BlockingFlightServer& server) {
     auto range = remote_doris_range();
     auto& params = range.table_format_params.remote_doris_params;
     params.__set_location_uri("grpc://localhost:" + std::to_string(server.port()));
-    arrow::flight::Ticket ticket {.ticket = "ticket"};
+    arrow::flight::Ticket ticket {"ticket"};
     params.__set_ticket(ticket.SerializeToString().ValueOrDie());
     return range;
 }

diff --git a/docker/compilation/Dockerfile.gcc10 b/docker/compilation/Dockerfile.gcc10
@@ -26,7 +26,7 @@ RUN yum makecache && yum -y update && yum -y groupinstall 'Development Tools' &&
     devtoolset-10-libasan-devel
 
 # build cmake
-ARG CMAKE_VERSION=3.22.1
+ARG CMAKE_VERSION=3.25.3
 ARG CMAKE_BASE_URL=https://github.com/Kitware/CMake/releases/download/v${CMAKE_VERSION}
 RUN wget ${CMAKE_BASE_URL}/cmake-${CMAKE_VERSION}-Linux-x86_64.sh -q -O /tmp/cmake-install.sh && \
     chmod u+x /tmp/cmake-install.sh && \
@@ -121,4 +121,3 @@ ENV JAVA_HOME="/usr/lib/jvm/java-11" \
 WORKDIR /root
 
 CMD ["/bin/bash"]
-
diff --git a/thirdparty/build-thirdparty.sh b/thirdparty/build-thirdparty.sh
@@ -1088,9 +1088,7 @@ build_arrow() {
         ldflags="-L${TP_LIB_DIR}"
     fi
 
-    CPPFLAGS="-I${TP_INCLUDE_DIR}" \
-        CXXFLAGS="-I${TP_INCLUDE_DIR}" \
-        LDFLAGS="${ldflags}" \
+    LDFLAGS="${ldflags}" \
         "${CMAKE_CMD}" -DCMAKE_POLICY_VERSION_MINIMUM=3.5 \
         -DCMAKE_CXX_STANDARD="${TP_CXX_STANDARD}" \
         -G "${GENERATOR}" -DARROW_PARQUET=ON -DARROW_IPC=ON -DARROW_BUILD_SHARED=OFF \
@@ -1125,6 +1123,7 @@ build_arrow() {
         -Dxsimd_SOURCE=BUNDLED \
         -DBrotli_SOURCE=BUNDLED \
         -DARROW_LZ4_USE_SHARED=OFF \
+        -DLZ4_ROOT="${TP_INSTALL_DIR};${TP_INSTALL_DIR}/include/lz4" \
         -DLZ4_LIB="${TP_INSTALL_DIR}/lib/liblz4.a" -DLZ4_INCLUDE_DIR="${TP_INSTALL_DIR}/include/lz4" \
         -DLz4_SOURCE=SYSTEM \
         -DARROW_ZSTD_USE_SHARED=OFF \
@@ -1147,6 +1146,7 @@ build_arrow() {
     cp -rf ./brotli_ep/src/brotli_ep-install/lib/libbrotlidec-static.a "${TP_INSTALL_DIR}/lib64/libbrotlidec.a"
     cp -rf ./brotli_ep/src/brotli_ep-install/lib/libbrotlicommon-static.a "${TP_INSTALL_DIR}/lib64/libbrotlicommon.a"
     strip_lib libarrow.a
+    strip_lib libarrow_compute.a
     strip_lib libparquet.a
     strip_lib libarrow_dataset.a
     strip_lib libarrow_acero.a
@@ -2091,6 +2091,7 @@ build_paimon_cpp() {
         mkdir -p "${paimon_deps_dir}"
         for paimon_arrow_dep in \
             libarrow.a \
+            libarrow_compute.a \
             libarrow_filesystem.a \
             libarrow_dataset.a \
             libarrow_acero.a \

diff --git a/thirdparty/download-thirdparty.sh b/thirdparty/download-thirdparty.sh
@@ -438,19 +438,19 @@ if [[ " ${TP_ARCHIVES[*]} " =~ " ARROW " ]]; then
         fi
         cd -
     fi
-    if [[ "${ARROW_SOURCE}" == "arrow-apache-arrow-17.0.0" ]]; then
+    if [[ "${ARROW_SOURCE}" == "arrow-apache-arrow-24.0.0" ]]; then
         cd "${TP_SOURCE_DIR}/${ARROW_SOURCE}"
         if [[ ! -f "${PATCHED_MARK}" ]]; then
             # Paimon-cpp parquet patches: row-group-aware batch reader, max_row_group_size,
             # GetBufferedSize(), int96 NANO guard, and Thrift_VERSION empty fix.
-            patch -p1 <"${TP_PATCH_DIR}/apache-arrow-17.0.0-paimon.patch"
+            patch -p1 <"${TP_PATCH_DIR}/apache-arrow-24.0.0-paimon.patch"
 
-            # apache-arrow-17.0.0-force-write-int96-timestamps.patch : 
-            # Introducing the parameter that forces writing int96 timestampes for compatibility with Paimon cpp. 
-            patch -p1 <"${TP_PATCH_DIR}/apache-arrow-17.0.0-force-write-int96-timestamps.patch"
+            # Introducing the parameter that forces writing INT96 timestamps for
+            # compatibility with the Doris Parquet writer.
+            patch -p1 <"${TP_PATCH_DIR}/apache-arrow-24.0.0-force-write-int96-timestamps.patch"
 
             # Add Parquet LZO page decompression support used by file scanner v2.
-            patch -p1 <"${TP_PATCH_DIR}/apache-arrow-17.0.0-lzo.patch"
+            patch -p1 <"${TP_PATCH_DIR}/apache-arrow-24.0.0-lzo.patch"
             touch "${PATCHED_MARK}"
         fi
         cd -
@@ -732,13 +732,14 @@ fi
 # patch paimon-cpp
 if [[ " ${TP_ARCHIVES[*]} " =~ " PAIMON_CPP " ]]; then
     cd "${TP_SOURCE_DIR}/${PAIMON_CPP_SOURCE}"
-    if [[ ! -f "${PATCHED_MARK}" ]]; then
-        if patch -p1 -N --batch --dry-run <"${TP_PATCH_DIR}/paimon-cpp-buildutils-static-deps.patch" >/dev/null 2>&1; then
-            patch -p1 -N --batch <"${TP_PATCH_DIR}/paimon-cpp-buildutils-static-deps.patch"
-        else
-            echo "Skip paimon-cpp patch: already applied or not applicable for current source"
+    PAIMON_CPP_ARROW_24_PATCHED_MARK="patched_mark_arrow_24"
+    if [[ ! -f "${PAIMON_CPP_ARROW_24_PATCHED_MARK}" ]]; then
+        if [[ ! -f "${PATCHED_MARK}" ]]; then
+            patch -p1 <"${TP_PATCH_DIR}/paimon-cpp-buildutils-static-deps.patch"
+            touch "${PATCHED_MARK}"
         fi
-        touch "${PATCHED_MARK}"
+        patch -p1 <"${TP_PATCH_DIR}/paimon-cpp-arrow-24-compatibility.patch"
+        touch "${PAIMON_CPP_ARROW_24_PATCHED_MARK}"
     fi
     cd -
     echo "Finished patching ${PAIMON_CPP_SOURCE}"

diff --git a/thirdparty/paimon-cpp-cache.cmake b/thirdparty/paimon-cpp-cache.cmake
@@ -65,7 +65,7 @@ set(LZ4_INCLUDE_DIR "${DORIS_INCLUDE_DIR}" CACHE PATH "LZ4 include directory")
 
 # ============================================================================
 # Arrow - Reuse from Doris (Doris Arrow now includes COMPUTE/DATASET/ACERO/FILESYSTEM)
-# Doris's Arrow 17.0.0 is built with the full module set that paimon-cpp
+# Doris's Arrow 24.0.0 is built with the full module set that paimon-cpp
 # needs, so we skip paimon-cpp's internal externalproject_add(arrow_ep ...).
 # ============================================================================
 set(PAIMON_USE_EXTERNAL_ARROW ON CACHE BOOL "Use pre-built Arrow from Doris instead of building from source")
@@ -74,6 +74,7 @@ set(DORIS_LIB64_DIR "${DORIS_THIRDPARTY_DIR}/lib64" CACHE PATH "Doris lib64 dire
 
 set(PAIMON_EXTERNAL_ARROW_INCLUDE_DIR "${DORIS_INCLUDE_DIR}" CACHE PATH "Arrow include directory")
 set(PAIMON_EXTERNAL_ARROW_LIB "${DORIS_LIB64_DIR}/libarrow.a" CACHE FILEPATH "Arrow core library")
+set(PAIMON_EXTERNAL_ARROW_COMPUTE_LIB "${DORIS_LIB64_DIR}/libarrow_compute.a" CACHE FILEPATH "Arrow Compute library")
 set(PAIMON_EXTERNAL_ARROW_DATASET_LIB "${DORIS_LIB64_DIR}/libarrow_dataset.a" CACHE FILEPATH "Arrow Dataset library")
 set(PAIMON_EXTERNAL_ARROW_ACERO_LIB "${DORIS_LIB64_DIR}/libarrow_acero.a" CACHE FILEPATH "Arrow Acero library")
 set(PAIMON_EXTERNAL_PARQUET_LIB "${DORIS_LIB64_DIR}/libparquet.a" CACHE FILEPATH "Parquet library")

diff --git a/...17.0.0-force-write-int96-timestamps.patch → ...24.0.0-force-write-int96-timestamps.patch b/...17.0.0-force-write-int96-timestamps.patch → ...24.0.0-force-write-int96-timestamps.patch
@@ -1,28 +1,28 @@
-diff -ruN arrow-apache-arrow-17.0.0-after-paimon/cpp/src/parquet/arrow/schema.cc arrow-apache-arrow-17.0.0/cpp/src/parquet/arrow/schema.cc
---- arrow-apache-arrow-17.0.0-after-paimon/cpp/src/parquet/arrow/schema.cc	2026-03-27 01:23:23.651831424 +0800
-+++ arrow-apache-arrow-17.0.0/cpp/src/parquet/arrow/schema.cc	2026-03-27 01:28:36.855281965 +0800
-@@ -178,7 +178,8 @@
+--- a/cpp/src/parquet/arrow/schema.cc
++++ b/cpp/src/parquet/arrow/schema.cc
+@@ -215,8 +215,9 @@
 
    // The user is explicitly asking for Impala int96 encoding, there is no
    // logical type.
--  if (arrow_properties.support_deprecated_int96_timestamps() && target_unit == ::arrow::TimeUnit::NANO) {
+-  if (arrow_properties.support_deprecated_int96_timestamps() &&
+-      target_unit == ::arrow::TimeUnit::NANO) {
 +  if (arrow_properties.force_write_int96_timestamps() ||
-+      (arrow_properties.support_deprecated_int96_timestamps() && target_unit == ::arrow::TimeUnit::NANO)) {
++      (arrow_properties.support_deprecated_int96_timestamps() &&
++       target_unit == ::arrow::TimeUnit::NANO)) {
      *physical_type = ParquetType::INT96;
      return Status::OK();
    }
-diff -ruN arrow-apache-arrow-17.0.0-after-paimon/cpp/src/parquet/properties.h arrow-apache-arrow-17.0.0/cpp/src/parquet/properties.h
---- arrow-apache-arrow-17.0.0-after-paimon/cpp/src/parquet/properties.h	2026-03-27 01:23:23.643831362 +0800
-+++ arrow-apache-arrow-17.0.0/cpp/src/parquet/properties.h	2026-03-27 01:27:47.717897537 +0800
-@@ -980,6 +980,7 @@
+--- a/cpp/src/parquet/properties.h
++++ b/cpp/src/parquet/properties.h
+@@ -1309,6 +1309,7 @@
     public:
      Builder()
          : write_timestamps_as_int96_(false),
 +          force_write_int96_timestamps_(false),
            coerce_timestamps_enabled_(false),
            coerce_timestamps_unit_(::arrow::TimeUnit::SECOND),
            truncated_timestamps_allowed_(false),
-@@ -1005,6 +1006,21 @@
+@@ -1334,6 +1335,21 @@
        return this;
      }
 
@@ -44,17 +44,17 @@ diff -ruN arrow-apache-arrow-17.0.0-after-paimon/cpp/src/parquet/properties.h ar
      /// \brief Coerce all timestamps to the specified time unit.
      /// \param unit time unit to truncate to.
      /// For Parquet versions 1.0 and 2.4, nanoseconds are casted to microseconds.
-@@ -1085,7 +1101,8 @@
+@@ -1423,7 +1439,8 @@
      /// Create the final properties.
      std::shared_ptr<ArrowWriterProperties> build() {
        return std::shared_ptr<ArrowWriterProperties>(new ArrowWriterProperties(
 -          write_timestamps_as_int96_, coerce_timestamps_enabled_, coerce_timestamps_unit_,
 +          write_timestamps_as_int96_, force_write_int96_timestamps_,
 +          coerce_timestamps_enabled_, coerce_timestamps_unit_,
            truncated_timestamps_allowed_, store_schema_, compliant_nested_types_,
-           engine_version_, use_threads_, executor_));
+           engine_version_, use_threads_, executor_, write_time_adjusted_to_utc_));
      }
-@@ -1093,6 +1110,8 @@
+@@ -1431,6 +1448,8 @@
     private:
      bool write_timestamps_as_int96_;
 
@@ -63,7 +63,7 @@ diff -ruN arrow-apache-arrow-17.0.0-after-paimon/cpp/src/parquet/properties.h ar
      bool coerce_timestamps_enabled_;
      ::arrow::TimeUnit::type coerce_timestamps_unit_;
      bool truncated_timestamps_allowed_;
-@@ -1107,6 +1126,8 @@
+@@ -1447,6 +1466,8 @@
 
    bool support_deprecated_int96_timestamps() const { return write_timestamps_as_int96_; }
 
@@ -72,24 +72,24 @@ diff -ruN arrow-apache-arrow-17.0.0-after-paimon/cpp/src/parquet/properties.h ar
    bool coerce_timestamps_enabled() const { return coerce_timestamps_enabled_; }
    ::arrow::TimeUnit::type coerce_timestamps_unit() const {
      return coerce_timestamps_unit_;
-@@ -1138,6 +1159,7 @@
+@@ -1483,6 +1504,7 @@
 
   private:
    explicit ArrowWriterProperties(bool write_nanos_as_int96,
 +                                 bool force_write_int96_timestamps,
                                   bool coerce_timestamps_enabled,
                                   ::arrow::TimeUnit::type coerce_timestamps_unit,
                                   bool truncated_timestamps_allowed, bool store_schema,
-@@ -1145,6 +1167,7 @@
-                                  EngineVersion engine_version, bool use_threads,
-                                  ::arrow::internal::Executor* executor)
+@@ -1491,6 +1513,7 @@
+                                  ::arrow::internal::Executor* executor,
+                                  bool write_time_adjusted_to_utc)
        : write_timestamps_as_int96_(write_nanos_as_int96),
 +        force_write_int96_timestamps_(force_write_int96_timestamps),
          coerce_timestamps_enabled_(coerce_timestamps_enabled),
          coerce_timestamps_unit_(coerce_timestamps_unit),
          truncated_timestamps_allowed_(truncated_timestamps_allowed),
-@@ -1155,6 +1178,7 @@
-         executor_(executor) {}
+@@ -1502,6 +1525,7 @@
+         write_time_adjusted_to_utc_(write_time_adjusted_to_utc) {}
 
    const bool write_timestamps_as_int96_;
 +  const bool force_write_int96_timestamps_;