[C++] Initialize oversized PageArena allocation metadata

.github/workflows/unit-test-cpp.yml

@@ -80,6 +80,12 @@ jobs:
       - name: Checkout repository
         uses: actions/checkout@v7
 
+      - name: Set up Python for clang-format
+        if: runner.os == 'Linux'
+        uses: actions/setup-python@v7
+        with:
+          python-version: "3.11"
+
       - name: Set up JDK 17
         uses: actions/setup-java@v6.0.1
         with:
@@ -121,10 +127,10 @@ jobs:
             elif command -v yum >/dev/null 2>&1; then
               sudo yum install -y libuuid-devel
             fi
-            sudo update-alternatives --install /usr/bin/clang-format clang-format /usr/bin/clang-format-17 100
-            sudo update-alternatives --set clang-format /usr/bin/clang-format-17
-            sudo apt-get update
-            sudo apt-get install -y uuid-dev
+            python -m venv "$RUNNER_TEMP/clang-format"
+            "$RUNNER_TEMP/clang-format/bin/python" -m pip install --only-binary=:all: clang-format==17.0.6
+            "$RUNNER_TEMP/clang-format/bin/clang-format" --version
+            echo "$RUNNER_TEMP/clang-format/bin" >> "$GITHUB_PATH"
           elif [[ "$RUNNER_OS" == "Windows" ]]; then
             choco install llvm --version 17.0.6 --force
           else

.github/workflows/unit-test-python.yml

@@ -45,6 +45,13 @@ jobs:
       matrix:
         os: [ ubuntu-latest, macos-latest, windows-latest ]
         python-version: ["3.9", "3.11", "3.14"]
+        exclude:
+          - os: ubuntu-latest
+            python-version: "3.9"
+        include:
+          # Python 3.9 has no setup-python build for Ubuntu 26.04.
+          - os: ubuntu-22.04
+            python-version: "3.9"
     runs-on: ${{ matrix.os }}
 
     steps:
@@ -92,8 +99,10 @@ jobs:
         shell: bash
         run: |
           if [[ "$RUNNER_OS" == "Linux" ]]; then
-            sudo update-alternatives --install /usr/bin/clang-format clang-format /usr/bin/clang-format-17 100
-            sudo update-alternatives --set clang-format /usr/bin/clang-format-17
+            python -m venv "$RUNNER_TEMP/clang-format"
+            "$RUNNER_TEMP/clang-format/bin/python" -m pip install --only-binary=:all: clang-format==17.0.6
+            "$RUNNER_TEMP/clang-format/bin/clang-format" --version
+            echo "$RUNNER_TEMP/clang-format/bin" >> "$GITHUB_PATH"
             sudo apt-get update
             sudo apt-get install -y uuid-dev
           elif [[ "$RUNNER_OS" == "Windows" ]]; then

cpp/src/common/allocator/page_arena.cc

@@ -45,7 +45,12 @@ char* PageArena::alloc(uint32_t size) {
         return new_page->alloc(size);
     } else {
         void* ptr = base_allocator_.alloc(size + sizeof(Page), mid_);
-        Page* new_page = new (ptr) Page(dummy_head_.next_);
+        if (UNLIKELY(ptr == nullptr)) {
+            return nullptr;
+        }
+        Page* new_page = new (ptr) Page(size, dummy_head_.next_);
+        // A dedicated oversized page is fully consumed and cannot be reused.
+        new_page->cur_alloc_ = new_page->page_end_;
         dummy_head_.next_ = new_page;
         return reinterpret_cast<char*>(new_page) + sizeof(Page);
     }

cpp/src/common/allocator/page_arena.h

@@ -76,8 +76,6 @@ class PageArena {
     class Page {
        public:
         Page() : next_(nullptr), page_end_(nullptr), cur_alloc_(nullptr) {}
-        explicit Page(Page* next_page)
-            : next_(next_page), page_end_(nullptr), cur_alloc_(nullptr) {}
         Page(uint32_t page_size, Page* next_page) {
             next_ = next_page;
             cur_alloc_ =

cpp/test/common/allocator/page_arena_test.cc

@@ -55,6 +55,37 @@ TEST(PageArenaTest, AllocMoreThanPageSize) {
     page_arena.reset();
 }
 
+TEST(PageArenaTest, OversizedAllocationTracksUsedBytesAndPreservesData) {
+    PageArena arena;
+    arena.init(64, MOD_DEFAULT);
+    char* large = arena.alloc(256);
+    ASSERT_NE(large, nullptr);
+    memset(large, 'x', 256);
+    ASSERT_EQ(arena.get_total_used_bytes(), 256);
+
+    char* small = arena.alloc(8);
+    ASSERT_NE(small, nullptr);
+    memset(small, 'y', 8);
+    EXPECT_EQ(arena.get_total_used_bytes(), 264);
+    EXPECT_EQ(arena.TEST_get_page_count(), 2);
+    for (int i = 0; i < 256; i++) {
+        ASSERT_EQ(large[i], 'x');
+    }
+    arena.reset();
+    EXPECT_EQ(arena.get_total_used_bytes(), 0);
+}
+
+TEST(PageArenaTest, ConsecutiveOversizedAllocationsTrackUsedBytes) {
+    PageArena arena;
+    arena.init(64, MOD_DEFAULT);
+    ASSERT_NE(arena.alloc(256), nullptr);
+    ASSERT_NE(arena.alloc(512), nullptr);
+    ASSERT_EQ(arena.get_total_used_bytes(), 768);
+    EXPECT_EQ(arena.TEST_get_page_count(), 2);
+    ASSERT_NE(arena.alloc(8), nullptr);
+    EXPECT_EQ(arena.get_total_used_bytes(), 776);
+}
+
 struct MyStruct {
     double x;
     int64_t y;