TheRock/rocm-systems/.github/workflows/rocprofiler-compute-sanitizers.yml

name: rocprofiler-compute Sanitizers
run-name: rocprofiler-compute-sanitizers

on:
  workflow_dispatch:
  push:
    branches:
      - develop
      - rocprofiler-compute-develop
    paths:
      - ".github/workflows/rocprofiler-compute-sanitizers.yml"
      - "projects/rocprofiler-compute/**"
      - "!**/*.md"
      - "!**/*.rtf"
      - "!**/*.rst"
      - "!**/.markdownlint-ci2.yaml"
      - "!**/.readthedocs.yaml"
      - "!**/.spellcheck.local.yaml"
      - "!**/.wordlist.txt"
      - "!projects/rocprofiler-compute/docs/**"
      - "!projects/rocprofiler-compute/source/docs/**"
      - "!projects/rocprofiler-compute/source/python/gui/**"
      - "!projects/rocprofiler-compute/docker/**"
      - "!projects/rocprofiler-compute/CMakePresets.json"
  pull_request:
    paths:
      - ".github/workflows/rocprofiler-compute-sanitizers.yml"
      - "projects/rocprofiler-compute/**"
      - "!**/*.md"
      - "!**/*.rtf"
      - "!**/*.rst"
      - "!**/.markdownlint-ci2.yaml"
      - "!**/.readthedocs.yaml"
      - "!**/.spellcheck.local.yaml"
      - "!**/.wordlist.txt"
      - "!projects/rocprofiler-compute/docs/**"
      - "!projects/rocprofiler-compute/source/docs/**"
      - "!projects/rocprofiler-compute/source/python/gui/**"
      - "!projects/rocprofiler-compute/docker/**"
      - "!projects/rocprofiler-compute/CMakePresets.json"

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

permissions:
  contents: read
  packages: read

env:
  ROCPROFSYS_CI: ON
  ROCM_PATH: "/opt/rocm"

jobs:
  sanitizer:
    name: ${{ matrix.sanitizer.name }}
    runs-on: linux-gfx950-1gpu-ccs-ossci-rocm
    timeout-minutes: 120
    strategy:
      fail-fast: false
      matrix:
        sanitizer:
          - name: asan
            enable_sanitizer: ASAN
          - name: tsan
            enable_sanitizer: TSAN
          - name: ubsan
            enable_sanitizer: UBSAN
    env:
      HIP_PLATFORM: "amd"
      OMPI_ALLOW_RUN_AS_ROOT: 1
      OMPI_ALLOW_RUN_AS_ROOT_CONFIRM: 1
    container:
      image: ghcr.io/rocm/rocprofiler-ubuntu:24.04-compute-ci-multiarch
      options: >-
        --privileged
        --ipc host
        --group-add video
        --device /dev/kfd
        --device /dev/dri
        --cap-add CAP_SYS_ADMIN
        --env-file /etc/podinfo/gha-gpu-isolation-settings

    steps:
      - uses: actions/checkout@v6
        with:
          sparse-checkout: |
            projects/rocprofiler-compute
            shared/ctest

      - name: Install dependencies
        shell: bash
        run: |
          apt-get update
          apt-get install -y libdw-dev pkg-config
          echo "✅ Dependencies Installed!"

      - name: Install latest nightly ROCm
        shell: bash
        run: |
          set -e
          mkdir "${ROCM_PATH}"
          tarball="$(ls "${ROCM_PATH}"-*-multiarch.tar.gz | head -n1)"
          tar -xf "${tarball}" -C "${ROCM_PATH}"
          echo "ROCm installed to: ${ROCM_PATH}"
          echo "✅ ROCm Installation Complete!"

      - name: Output TheRock manifest
        run: cat "${ROCM_PATH}/share/therock/therock_manifest.json"

      - name: Locate clang sanitizer runtime
        id: sanitizer_runtime
        shell: bash
        run: |
          resource_dir="$("${ROCM_PATH}/llvm/bin/clang" -print-resource-dir)"
          sanitizer_lib_dir="${resource_dir}/lib/linux"
          if [ ! -d "${sanitizer_lib_dir}" ]; then
            echo "❌ Missing clang sanitizer runtime: ${sanitizer_lib_dir}"
            exit 1
          fi
          echo "sanitizer_lib_dir=${sanitizer_lib_dir}" >> "${GITHUB_OUTPUT}"
          echo "Clang sanitizer runtime: ${sanitizer_lib_dir}"

      - name: Install Python requirements
        working-directory: projects/rocprofiler-compute
        shell: bash
        run: |
          for attempt in 1 2 3; do
            pip install \
              -r requirements.txt \
              --break-system-packages \
              --ignore-installed \
              --timeout 60 && break
            echo "⚠️ pip install attempt ${attempt} failed, retrying..."
            sleep 10
          done
          for attempt in 1 2 3; do
            pip install \
              -r requirements-test.txt \
              --break-system-packages \
              --ignore-installed \
              --timeout 60 && break
            echo "⚠️ pip install attempt ${attempt} failed, retrying..."
            sleep 10
          done
          echo "✅ pip requirements installed!"

      - name: Configure, build, and test
        id: test
        continue-on-error: true
        working-directory: projects/rocprofiler-compute
        shell: bash
        run: |
          set -e
          git config --global --add safe.directory "${GITHUB_WORKSPACE}"
          git config --global --add safe.directory "${PWD}"

          export PATH="${ROCM_PATH}/bin:${ROCM_PATH}/llvm/bin:${PATH}"
          export LD_LIBRARY_PATH="${ROCM_PATH}/lib:${ROCM_PATH}/llvm/lib:${ROCM_PATH}/lib/rocm_sysdeps/lib:${{ steps.sanitizer_runtime.outputs.sanitizer_lib_dir }}:${LD_LIBRARY_PATH:-}"

          c_compiler="$(command -v amdclang)"
          cxx_compiler="$(command -v amdclang++)"

          python3 ./tools/run-ci.py \
            --name "ROCm/rocprofiler-compute-${{ github.ref_name }}-ubuntu-24.04-${{ matrix.sanitizer.name }}" \
            --actor "${{ github.actor }}" \
            --site linux-gfx950-1gpu-ccs-ossci-rocm \
            --mode Continuous \
            --build-jobs 16 \
            --repeat-until-pass 1 \
            --repeat-after-timeout 1 \
            -- \
            "-DCMAKE_C_COMPILER=${c_compiler}" \
            "-DCMAKE_CXX_COMPILER=${cxx_compiler}" \
            "-DCMAKE_HIP_COMPILER=${cxx_compiler}" \
            -DCMAKE_BUILD_TYPE=Debug \
            "-DCMAKE_PREFIX_PATH=${ROCM_PATH}" \
            -DENABLE_TESTS=ON \
            -DINSTALL_TESTS=ON \
            "-DENABLE_SANITIZER=${{ matrix.sanitizer.enable_sanitizer }}" \
            -- \
            -L sanitizer

      - name: Output logs
        if: steps.test.outcome == 'failure'
        working-directory: projects/rocprofiler-compute
        shell: bash
        run: |
          echo "❌ Run Failed: Outputting available log files..."
          for log in build/Testing/Temporary/LastTest*.log \
            build/Testing/Temporary/LastConfigure*.log; do
            if [ -f "${log}" ]; then
              echo "=== ${log} ==="
              cat "${log}"
            fi
          done

      - name: Check test results
        if: always()
        shell: bash
        run: |
          if [ "${{ steps.test.outcome }}" = "failure" ]; then
            echo "❌ Tests failed - see CDash for details: https://my.cdash.org/index.php?project=rocprofiler-compute"
            exit 1
          fi
          echo "✅ All tests passed!"