clang_sanitizers.yml

name: 🧴 clang sanitizers

on: [push, pull_request]

concurrency:
  group: ${{ github.ref }}-${{ github.head_ref }}-clangsanitizers
  cancel-in-progress: true

jobs:
  build_UB_sanitizer:
    name: Clang UB sanitizer
    runs-on: ubuntu-22.04
    container: ubuntu:23.10
    if: github.event.pull_request.draft == false
    env:
      CC: clang
      CXX: clang++
      # On CI for this test, Ninja is slower than the default:
      #CMAKE_GENERATOR: Ninja
    steps:
    - uses: actions/checkout@v4
    - name: install dependencies
      run: |
        .github/workflows/dependencies/clang17.sh
    - name: CCache Cache
      uses: actions/cache@v4
      with:
        path: ~/.cache/ccache
        key: ccache-${{ github.workflow }}-${{ github.job }}-git-${{ github.sha }}
        restore-keys: |
             ccache-${{ github.workflow }}-${{ github.job }}-git-
    - name: build WarpX
      run: |
        export CCACHE_COMPRESS=1
        export CCACHE_COMPRESSLEVEL=10
        export CCACHE_MAXSIZE=100M
        ccache -z

        export CXX=$(which clang++-17)
        export CC=$(which clang-17)
        export CXXFLAGS="-fsanitize=undefined,address,pointer-compare -fno-sanitize-recover=all"

        cmake -S . -B build                 \
          -GNinja                           \
          -DCMAKE_VERBOSE_MAKEFILE=ON       \
          -DCMAKE_BUILD_TYPE=RelWithDebInfo \
          -DWarpX_DIMS="RZ;1;2;3"           \
          -DWarpX_FFT=ON                    \
          -DWarpX_QED=ON                    \
          -DWarpX_QED_TABLE_GEN=ON          \
          -DWarpX_OPENPMD=ON                \
          -DWarpX_PRECISION=SINGLE          \
          -DWarpX_PARTICLE_PRECISION=SINGLE
        cmake --build build -j 4

        ccache -s
        du -hs ~/.cache/ccache

    - name: run with UB sanitizer
      run: |

        # We need these two lines because these tests run inside a docker container
        export OMPI_ALLOW_RUN_AS_ROOT=1
        export OMPI_ALLOW_RUN_AS_ROOT_CONFIRM=1

        export OMP_NUM_THREADS=2

        #MPI implementations often leak memory
        export "ASAN_OPTIONS=detect_leaks=0"

        mpirun -n 2 ./build/bin/warpx.rz Examples/Physics_applications/laser_acceleration/inputs_base_rz
        mpirun -n 2 ./build/bin/warpx.1d Examples/Physics_applications/laser_acceleration/inputs_test_1d_laser_acceleration
        mpirun -n 2 ./build/bin/warpx.2d Examples/Physics_applications/laser_acceleration/inputs_base_2d
        mpirun -n 2 ./build/bin/warpx.3d Examples/Physics_applications/laser_acceleration/inputs_base_3d

  build_thread_sanitizer:
    name: Clang thread sanitizer
    runs-on: ubuntu-22.04
    container: ubuntu:23.10
    if: github.event.pull_request.draft == false
    env:
      CC: clang
      CXX: clang++
      # On CI for this test, Ninja is slower than the default:
      #CMAKE_GENERATOR: Ninja
    steps:
    - uses: actions/checkout@v4
    - name: install dependencies
      run: |
        .github/workflows/dependencies/clang17.sh
    - name: CCache Cache
      uses: actions/cache@v4
      with:
        path: ~/.cache/ccache
        key: ccache-${{ github.workflow }}-${{ github.job }}-git-${{ github.sha }}
        restore-keys: |
             ccache-${{ github.workflow }}-${{ github.job }}-git-
    - name: build WarpX
      run: |
        export CCACHE_COMPRESS=1
        export CCACHE_COMPRESSLEVEL=10
        export CCACHE_MAXSIZE=100M
        ccache -z

        export CXX=$(which clang++-17)
        export CC=$(which clang-17)
        export CXXFLAGS="-fsanitize=thread"

        cmake -S . -B build                 \
          -GNinja                           \
          -DCMAKE_VERBOSE_MAKEFILE=ON       \
          -DCMAKE_BUILD_TYPE=RelWithDebInfo \
          -DWarpX_DIMS="RZ;1;2;3"           \
          -DWarpX_FFT=ON                    \
          -DWarpX_QED=ON                    \
          -DWarpX_QED_TABLE_GEN=ON          \
          -DWarpX_OPENPMD=ON                \
          -DWarpX_EB=OFF                    \
          -DWarpX_PRECISION=DOUBLE          \
          -DWarpX_PARTICLE_PRECISION=DOUBLE
        cmake --build build -j 4

        cmake -S . -B build_EB              \
          -GNinja                           \
          -DCMAKE_VERBOSE_MAKEFILE=ON       \
          -DCMAKE_BUILD_TYPE=RelWithDebInfo \
          -DWarpX_DIMS="2"                  \
          -DWarpX_FFT=ON                    \
          -DWarpX_QED=ON                    \
          -DWarpX_QED_TABLE_GEN=ON          \
          -DWarpX_OPENPMD=ON                \
          -DWarpX_EB=ON                     \
          -DWarpX_PRECISION=DOUBLE          \
          -DWarpX_PARTICLE_PRECISION=DOUBLE
        cmake --build build_EB -j 4

        ccache -s
        du -hs ~/.cache/ccache

    - name: run with thread sanitizer
      run: |
        export PMIX_MCA_gds=hash
        export TSAN_OPTIONS='ignore_noninstrumented_modules=1'
        export ARCHER_OPTIONS="verbose=1"

        # We need these two lines because these tests run inside a docker container
        export OMPI_ALLOW_RUN_AS_ROOT=1
        export OMPI_ALLOW_RUN_AS_ROOT_CONFIRM=1

        export OMP_NUM_THREADS=2

        mpirun -n 2 ./build/bin/warpx.rz Examples/Physics_applications/laser_acceleration/inputs_base_rz warpx.serialize_initial_conditions = 0
        mpirun -n 2 ./build/bin/warpx.1d Examples/Physics_applications/laser_acceleration/inputs_test_1d_laser_acceleration warpx.serialize_initial_conditions = 0
        mpirun -n 2 ./build/bin/warpx.2d Examples/Physics_applications/laser_acceleration/inputs_base_2d warpx.serialize_initial_conditions = 0
        mpirun -n 2 ./build/bin/warpx.3d Examples/Physics_applications/laser_acceleration/inputs_base_3d warpx.serialize_initial_conditions = 0

        git clone https://github.com/ECP-WarpX/warpx-data ../warpx-data
        cd Examples/Tests/embedded_circle

        ulimit -c unlimited

        mpirun -n 2 ../../../build_EB/bin/warpx.2d inputs_test_2d_embedded_circle warpx.serialize_initial_conditions = 0