bytedance · liuq19 · Mar 25, 2026 · Feb 6, 2026 · Mar 17, 2026 · Mar 17, 2026
diff --git a/.github/workflows/clang-format-check.yml b/.github/workflows/clang-format-check.yml
@@ -1,10 +1,44 @@
-name: clang-format Check
-on: [push, pull_request]
+name: CI / clang-format
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
+  cancel-in-progress: true
+
+on:
+  push:
+    branches: [ master ]
+    paths:
+      - '.clang-format'
+      - 'include/sonic/**'
+      - 'benchmark/**'
+      - 'tests/**'
+      - 'example/**'
+      - 'fuzz/**'
+      - '.github/workflows/clang-format-check.yml'
+  pull_request:
+    paths:
+      - '.clang-format'
+      - 'include/sonic/**'
+      - 'benchmark/**'
+      - 'tests/**'
+      - 'example/**'
+      - 'fuzz/**'
+      - '.github/workflows/clang-format-check.yml'
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+defaults:
+  run:
+    shell: bash
+
 jobs:
   formatting-check:
-    name: Formatting Check
+    name: clang-format
     runs-on: ubuntu-24.04
     strategy:
+      fail-fast: false
       matrix:
         path:
           - 'include/sonic'

diff --git a/.github/workflows/test_arm.yml b/.github/workflows/test_arm.yml
@@ -1,4 +1,15 @@
-name: Test ARM
+name: CI / Test (ARM)
+
+permissions:
+  contents: read
+
+defaults:
+  run:
+    shell: bash
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
+  cancel-in-progress: true
 
 on:
   push:
@@ -10,12 +21,13 @@ on:
 jobs:
   test_arm:
     runs-on: ubuntu-24.04-arm
+    timeout-minutes: 60
     steps:
-      - uses: actions/checkout@v3
+      - uses: actions/checkout@v4
       - name: Install dependencies
         run: |
           sudo apt-get update
-          sudo apt-get install -y cmake ninja-build g++
+          sudo apt-get install -y --no-install-recommends cmake ninja-build g++
       - name: Build and test NEON
         run: |
           cmake -S . -B build-neon -G Ninja
@@ -34,6 +46,6 @@ jobs:
       - name: Build and test SVE2
         if: steps.sve_check.outputs.supported == 'true'
         run: |
-          cmake -S . -B build-sve -G Ninja -DENABLE_SVE2_128=ON -DENABLE_ASAN=OFF
+          cmake -S . -B build-sve -G Ninja -DENABLE_SVE2_128=ON
           cmake --build build-sve
-          ./build-sve/tests/unittest
+          ASAN_OPTIONS=detect_stack_use_after_return=0 ./build-sve/tests/unittest
diff --git a/.github/workflows/test_coverage.yml b/.github/workflows/test_coverage.yml
@@ -1,13 +1,29 @@
 # yaml-language-server: $schema=https://json-schema.org/draft-07/schema#
 name: Test Coverage
 
+permissions:
+  contents: read
+  actions: write
+  id-token: write
+
+defaults:
+  run:
+    shell: bash
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
+  cancel-in-progress: true
+
 on:
   push:
+    branches: [ master ]
   pull_request:
 
 jobs:
   test-coverage:
+    name: bazel coverage
     runs-on: ubuntu-24.04
+    timeout-minutes: 60
     env:
       GCC_VERSION: 12
 
@@ -16,21 +32,32 @@ jobs:
         uses: actions/checkout@v4
 
       - name: Read Bazel version
-        shell: bash
         run: |
           echo "BAZEL_VERSION=$(cat .bazelversion)" >> "$GITHUB_ENV"
 
+      - name: Cache Bazel
+        uses: actions/cache@v4
+        with:
+          path: |
+            ~/.cache/bazel
+            ~/.cache/bazelisk
+          key: ${{ runner.os }}-bazel-${{ env.BAZEL_VERSION }}-${{ hashFiles('MODULE.bazel.lock', 'WORKSPACE.bzlmod', 'BUILD.bazel') }}
+          restore-keys: |
+            ${{ runner.os }}-bazel-${{ env.BAZEL_VERSION }}-
+            ${{ runner.os }}-bazel-
+
       - name: Setup bazel
         uses: jwlawson/actions-setup-bazel@v2
         with:
           bazel-version: "${{ env.BAZEL_VERSION }}"
 
       - name: Setup GCC
         run: |
-          sudo apt-get install -y gcc-12 g++-12
+          sudo apt-get update
+          sudo apt-get install -y --no-install-recommends gcc-12 g++-12
       - name: Install lcov
         run: |
-          sudo apt install lcov
+          sudo apt-get install -y --no-install-recommends lcov
 
       - name: Coverage
         run: |

diff --git a/.github/workflows/test_x86.yml b/.github/workflows/test_x86.yml
@@ -1,18 +1,43 @@
-name: Test
-on: [push, pull_request]
+name: CI / Test (x86)
+
+permissions:
+  contents: read
+  actions: write
+
+defaults:
+  run:
+    shell: bash
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
+  cancel-in-progress: true
+
+on:
+  push:
+    branches: [ master ]
+    paths-ignore:
+      - 'docs/**'
+      - 'licenses/**'
+      - '**/*.md'
+      - 'Doxyfile'
+      - '.vscode/**'
+  pull_request:
+  workflow_dispatch:
 
 jobs:
   test-llvm:
+    name: clang ${{ matrix.llvm_version }} / ${{ matrix.tool }} / ${{ matrix.arch }}
     runs-on: ubuntu-24.04
+    timeout-minutes: 45
     strategy:
       fail-fast: false
       matrix:
         llvm_version: ['11', '16.0.4', '17']
         tool: ['cmake', 'bazel']
         arch: [westmere, haswell]
         exclude:
-            - tool: 'cmake'
-              arch: westmere
+          - tool: 'cmake'
+            arch: westmere
     env:
       CC: clang
       CXX: clang++
@@ -22,10 +47,21 @@ jobs:
 
       - name: Read Bazel version
         if: matrix.tool == 'bazel'
-        shell: bash
         run: |
           echo "BAZEL_VERSION=$(cat .bazelversion)" >> "$GITHUB_ENV"
 
+      - name: Cache Bazel
+        if: matrix.tool == 'bazel'
+        uses: actions/cache@v4
+        with:
+          path: |
+            ~/.cache/bazel
+            ~/.cache/bazelisk
+          key: ${{ runner.os }}-bazel-${{ env.BAZEL_VERSION }}-${{ hashFiles('MODULE.bazel.lock', 'WORKSPACE.bzlmod', 'BUILD.bazel') }}
+          restore-keys: |
+            ${{ runner.os }}-bazel-${{ env.BAZEL_VERSION }}-
+            ${{ runner.os }}-bazel-
+
       - name: Install LLVM and Clang
         uses: KyleMayes/install-llvm-action@v2
         with:
@@ -45,7 +81,7 @@ jobs:
 
       - name: Run ${{ matrix.arch }} Test Use Bazel
         if: matrix.tool == 'bazel'
-        run : |
+        run: |
           bash ./scripts/unittest.sh -c --arch=${{ matrix.arch }}
 
       - name: Run Test Use CMake
@@ -54,7 +90,9 @@ jobs:
           bash ./scripts/run_cmake.sh
 
   test-gcc:
+    name: gcc ${{ matrix.gcc_version }} / ${{ matrix.tool }} / ${{ matrix.arch }} / ${{ matrix.dispatch }}
     runs-on: ubuntu-24.04
+    timeout-minutes: 60
     strategy:
       fail-fast: false
       matrix:
@@ -80,13 +118,25 @@ jobs:
 
       - name: Read Bazel version
         if: matrix.tool == 'bazel'
-        shell: bash
         run: |
           echo "BAZEL_VERSION=$(cat .bazelversion)" >> "$GITHUB_ENV"
 
       - name: Install GCC
         run: |
-          sudo apt-get install -y gcc-${{ matrix.gcc_version }} g++-${{ matrix.gcc_version }}
+          sudo apt-get update
+          sudo apt-get install -y --no-install-recommends gcc-${{ matrix.gcc_version }} g++-${{ matrix.gcc_version }}
+
+      - name: Cache Bazel
+        if: matrix.tool == 'bazel'
+        uses: actions/cache@v4
+        with:
+          path: |
+            ~/.cache/bazel
+            ~/.cache/bazelisk
+          key: ${{ runner.os }}-bazel-${{ env.BAZEL_VERSION }}-${{ hashFiles('MODULE.bazel.lock', 'WORKSPACE.bzlmod', 'BUILD.bazel') }}
+          restore-keys: |
+            ${{ runner.os }}-bazel-${{ env.BAZEL_VERSION }}-
+            ${{ runner.os }}-bazel-
 
       - name: Setup bazel
         if: matrix.tool == 'bazel'
@@ -102,7 +152,7 @@ jobs:
 
       - name: Run ${{ matrix.arch }} ${{ matrix.dispatch }} Test Use Bazel
         if: matrix.tool == 'bazel'
-        run : |
+        run: |
           bash ./scripts/unittest.sh -g --arch=${{ matrix.arch }} --dispatch=${{ matrix.dispatch }}
 
       - name: Run Test Use CMake

diff --git a/benchmark/main.cpp b/benchmark/main.cpp
@@ -16,11 +16,14 @@
 
 #include <benchmark/benchmark.h>
 
+#include <cstdlib>
 #include <filesystem>
 #include <fstream>
 #include <iostream>
 #include <sstream>
 #include <string_view>
+#include <system_error>
+#include <vector>
 
 #include "cjson.hpp"
 #include "jsoncpp.hpp"
@@ -30,14 +33,50 @@
 #include "sonic.hpp"
 #include "yyjson.hpp"
 
-static std::string get_json(const std::string_view file) {
-  std::ifstream ifs;
+static std::string get_json(const std::filesystem::path &file) {
+  std::ifstream ifs(file, std::ios::in | std::ios::binary);
+  if (!ifs.is_open()) return {};
+
   std::stringstream ss;
-  ifs.open(file.data());
   ss << ifs.rdbuf();
   return ss.str();
 }
 
+static void add_testdata_candidates(std::vector<std::filesystem::path> &out,
+                                    const std::filesystem::path &root) {
+  if (root.empty()) return;
+
+  out.push_back(root / "testdata");
+  out.push_back(root / "_main" / "testdata");
+
+  std::error_code ec;
+  if (!std::filesystem::is_directory(root, ec)) return;
+  for (const auto &e : std::filesystem::directory_iterator(root, ec)) {
+    if (e.is_directory(ec)) out.push_back(e.path() / "testdata");
+  }
+}
+
+static std::filesystem::path find_testdata_dir(const char *argv0) {
+  std::vector<std::filesystem::path> candidates;
+  candidates.emplace_back("testdata");
+
+  if (const char *p = std::getenv("RUNFILES_DIR"))
+    add_testdata_candidates(candidates, std::filesystem::path(p));
+  if (const char *p = std::getenv("TEST_SRCDIR"))
+    add_testdata_candidates(candidates, std::filesystem::path(p));
+
+  if (argv0 && *argv0) {
+    add_testdata_candidates(
+        candidates, std::filesystem::path(std::string(argv0) + ".runfiles"));
+  }
+
+  std::error_code ec;
+  for (const auto &c : candidates) {
+    if (std::filesystem::is_directory(c, ec)) return c;
+  }
+  return {};
+}
+
 template <typename Json, typename PR, typename SR>
 static void BM_Encode(benchmark::State &state, std::string_view filename,
                       std::string_view data) {
@@ -173,7 +212,7 @@ static void BM_Decode(benchmark::State &state, std::string filename,
   state.SetBytesProcessed(int64_t(state.iterations()) * int64_t(data.size()));
 }
 
-static void regitser_OnDemand() {
+static void regitser_OnDemand(const std::filesystem::path &testdata_dir) {
   std::vector<OnDemand> tests = {
       {"twitter", "Normal", {"search_metadata", "count"}, 100, true},
       {"citm_catalog",
@@ -185,7 +224,7 @@ static void regitser_OnDemand() {
   };
 
   for (auto &t : tests) {
-    auto file_path = std::string("testdata/") + t.file + ".json";
+    auto file_path = testdata_dir / (t.file + ".json");
     t.json = get_json(file_path);
 
 #define REG_ONDEMAND(JSON)                                                   \
@@ -203,14 +242,24 @@ static void regitser_OnDemand() {
 int main(int argc, char **argv) {
   benchmark::Initialize(&argc, argv);
 
+  auto testdata_dir = find_testdata_dir(argv[0]);
+  if (testdata_dir.empty()) {
+    std::cerr << "Cannot locate 'testdata' directory. Try running with Bazel "
+                 "runfiles."
+              << std::endl;
+    return 1;
+  }
+
   // Read the data from json files
   std::vector<std::pair<std::filesystem::path, std::string>> jsons;
-  for (const auto &entry : std::filesystem::directory_iterator("testdata"))
+  std::error_code ec;
+  for (const auto &entry :
+       std::filesystem::directory_iterator(testdata_dir, ec)) {
     if (entry.path().extension() == ".json")
-      jsons.push_back(
-          std::make_pair(entry.path(), get_json(entry.path().string())));
+      jsons.push_back(std::make_pair(entry.path(), get_json(entry.path())));
+  }
 
-  regitser_OnDemand();
+  regitser_OnDemand(testdata_dir);
 #define ADD_JSON_BMK(JSON, ACT)                                      \
   do {                                                               \
     benchmark::RegisterBenchmark(                                    \