test-wheel-linux.yml

# SPDX-FileCopyrightText: Copyright (c) 2024-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
#
# SPDX-License-Identifier: Apache-2.0

name: "CI: Test wheels"

on:
  workflow_call:
    inputs:
      build-type:
        type: string
        required: true
      host-platform:
        type: string
        required: true
      build-ctk-ver:
        type: string
        required: true
      matrix_filter:
        type: string
        default: "."
      nruns:
        type: number
        default: 1

defaults:
  run:
    shell: bash --noprofile --norc -xeuo pipefail {0}

jobs:
  compute-matrix:
    runs-on: ubuntu-latest
    env:
      BUILD_TYPE: ${{ inputs.build-type }}
      ARCH: ${{ (inputs.host-platform == 'linux-64' && 'amd64') ||
                (inputs.host-platform == 'linux-aarch64' && 'arm64') }}
    outputs:
      MATRIX: ${{ steps.compute-matrix.outputs.MATRIX }}
      OLD_BRANCH: ${{ steps.compute-matrix.outputs.OLD_BRANCH }}
    steps:
      - name: Checkout ${{ github.event.repository.name }}
        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2

      - name: Validate Test Type
        run: |
          if [[ "$BUILD_TYPE" != "pull-request" ]] && [[ "$BUILD_TYPE" != "nightly" ]] && [[ "$BUILD_TYPE" != "branch" ]]; then
              echo "Invalid build type! Must be one of 'nightly', 'pull-request', or 'branch'."
              exit 1
          fi

      - name: Compute Python Test Matrix
        id: compute-matrix
        run: |
          # Use the nightly matrix for branch tests
          MATRIX_TYPE="${BUILD_TYPE}"
          if [[ "${MATRIX_TYPE}" == "branch" ]]; then
            MATRIX_TYPE="nightly"
          fi

          # Read base matrix from YAML file for the specific architecture
          TEST_MATRIX=$(yq -o json ".linux[\"${MATRIX_TYPE}\"] | map(select(.ARCH == \"${ARCH}\"))" ci/test-matrix.yml)

          # Apply matrix filter and wrap in include structure
          MATRIX=$(echo "$TEST_MATRIX" | jq -c '${{ inputs.matrix_filter }} | if (. | length) > 0 then {include: .} else "Error: Empty matrix\n" | halt_error(1) end')

          echo "MATRIX=${MATRIX}" | tee --append "${GITHUB_OUTPUT}"

          # This job has yq already installed, so let's do it here
          OLD_BRANCH=$(yq '.backport_branch' ci/versions.yml)
          echo "OLD_BRANCH=${OLD_BRANCH}" >> "$GITHUB_OUTPUT"

  test:
    name: py${{ matrix.PY_VER }}, ${{ matrix.CUDA_VER }}, ${{ (matrix.LOCAL_CTK == '1' && 'local') || 'wheels' }}, ${{ matrix.GPU }}${{ matrix.GPU_COUNT != '1' && format('(x{0})', matrix.GPU_COUNT) || '' }}
    needs: compute-matrix
    strategy:
      fail-fast: false
      matrix: ${{ fromJSON(needs.compute-matrix.outputs.MATRIX) }}
    runs-on: "linux-${{ matrix.ARCH }}-gpu-${{ matrix.GPU }}-${{ matrix.DRIVER }}-${{ matrix.GPU_COUNT }}"
    # The build stage could fail but we want the CI to keep moving.
    if: ${{ github.repository_owner == 'nvidia' && !cancelled() }}
    # Our self-hosted runners require a container
    # TODO: use a different (nvidia?) container
    container:
      options: -u root --security-opt seccomp=unconfined --shm-size 16g
      image: ubuntu:22.04
      env:
        NVIDIA_VISIBLE_DEVICES: ${{ env.NVIDIA_VISIBLE_DEVICES }}
    steps:
      - name: Ensure GPU is working
        run: nvidia-smi

      - name: Checkout ${{ github.event.repository.name }}
        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2

      - name: Setup proxy cache
        uses: nv-gha-runners/setup-proxy-cache@main
        continue-on-error: true

      - name: Install dependencies
        uses: ./.github/actions/install_unix_deps
        continue-on-error: false
        with:
          # for artifact fetching, graphics libs
          dependencies: "jq wget libgl1 libegl1"
          dependent_exes: "jq wget"

      - name: Set environment variables
        env:
          BUILD_CUDA_VER: ${{ inputs.build-ctk-ver }}
          CUDA_VER: ${{ matrix.CUDA_VER }}
          HOST_PLATFORM: ${{ inputs.host-platform }}
          LOCAL_CTK: ${{ matrix.LOCAL_CTK }}
          PY_VER: ${{ matrix.PY_VER }}
          SHA: ${{ github.sha }}
        run: ./ci/tools/env-vars test

      - name: Download cuda-pathfinder build artifacts
        uses: actions/download-artifact@70fc10c6e5e1ce46ad2ea6f2b72d43f7d47b13c3  # v8.0.0
        with:
          name: cuda-pathfinder-wheel
          path: ./cuda_pathfinder

      - name: Download cuda-python build artifacts
        if: ${{ env.SKIP_CUDA_BINDINGS_TEST == '0'}}
        uses: actions/download-artifact@70fc10c6e5e1ce46ad2ea6f2b72d43f7d47b13c3  # v8.0.0
        with:
          name: cuda-python-wheel
          path: .

      - name: Download cuda.bindings build artifacts
        if: ${{ env.SKIP_CUDA_BINDINGS_TEST == '0'}}
        uses: actions/download-artifact@70fc10c6e5e1ce46ad2ea6f2b72d43f7d47b13c3  # v8.0.0
        with:
          name: ${{ env.CUDA_BINDINGS_ARTIFACT_NAME }}
          path: ${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}

      - name: Download cuda-python & cuda.bindings build artifacts from the prior branch
        if: ${{ env.SKIP_CUDA_BINDINGS_TEST == '1'}}
        env:
          GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
        run: |
          # See https://github.com/cli/cli/blob/trunk/docs/install_linux.md#debian-ubuntu-linux-raspberry-pi-os-apt.
          # gh is needed for artifact fetching.
          mkdir -p -m 755 /etc/apt/keyrings \
                && out=$(mktemp) && wget -nv -O$out https://cli.github.com/packages/githubcli-archive-keyring.gpg \
                && cat $out | tee /etc/apt/keyrings/githubcli-archive-keyring.gpg > /dev/null \
          && chmod go+r /etc/apt/keyrings/githubcli-archive-keyring.gpg \
          && echo "deb [arch=$(dpkg --print-architecture) signed-by=/etc/apt/keyrings/githubcli-archive-keyring.gpg] https://cli.github.com/packages stable main" | tee /etc/apt/sources.list.d/github-cli.list > /dev/null \
          && apt update \
          && apt install gh -y

          OLD_BRANCH=${{ needs.compute-matrix.outputs.OLD_BRANCH }}
          OLD_BASENAME="cuda-bindings-python${PYTHON_VERSION_FORMATTED}-cuda*-${{ inputs.host-platform }}*"
          LATEST_PRIOR_RUN_ID=$(gh run list -b ${OLD_BRANCH} -L 1 -w "ci.yml" -s success -R NVIDIA/cuda-python --json databaseId | jq '.[]| .databaseId')
          if [[ "$LATEST_PRIOR_RUN_ID" == "" ]]; then
            echo "LATEST_PRIOR_RUN_ID not found!"
            exit 1
          fi

          gh run download $LATEST_PRIOR_RUN_ID -p ${OLD_BASENAME} -R NVIDIA/cuda-python
          rm -rf ${OLD_BASENAME}-tests  # exclude cython test artifacts
          ls -al $OLD_BASENAME
          mkdir -p "${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}"
          mv $OLD_BASENAME/*.whl "${{ env.CUDA_BINDINGS_ARTIFACTS_DIR }}"/
          rmdir $OLD_BASENAME

          gh run download $LATEST_PRIOR_RUN_ID -p cuda-python-wheel -R NVIDIA/cuda-python
          ls -al cuda-python-wheel
          mv cuda-python-wheel/*.whl .
          rmdir cuda-python-wheel

      - name: Display structure of downloaded cuda-python artifacts
        run: |
          pwd
          ls -lahR .

      - name: Display structure of downloaded cuda.bindings artifacts
        run: |
          pwd
          ls -lahR $CUDA_BINDINGS_ARTIFACTS_DIR

      - name: Download cuda.bindings Cython tests
        if: ${{ env.SKIP_CYTHON_TEST == '0' }}
        uses: actions/download-artifact@70fc10c6e5e1ce46ad2ea6f2b72d43f7d47b13c3  # v8.0.0
        with:
          name: ${{ env.CUDA_BINDINGS_ARTIFACT_NAME }}-tests
          path: ${{ env.CUDA_BINDINGS_CYTHON_TESTS_DIR }}

      - name: Display structure of downloaded cuda.bindings Cython tests
        if: ${{ env.SKIP_CYTHON_TEST == '0' }}
        run: |
          pwd
          ls -lahR $CUDA_BINDINGS_CYTHON_TESTS_DIR

      - name: Download cuda.core build artifacts
        uses: actions/download-artifact@70fc10c6e5e1ce46ad2ea6f2b72d43f7d47b13c3  # v8.0.0
        with:
          name: ${{ env.CUDA_CORE_ARTIFACT_NAME }}
          path: ${{ env.CUDA_CORE_ARTIFACTS_DIR }}

      - name: Display structure of downloaded cuda.core build artifacts
        run: |
          pwd
          ls -lahR $CUDA_CORE_ARTIFACTS_DIR

      - name: Download cuda.core Cython tests
        if: ${{ env.SKIP_CYTHON_TEST == '0' }}
        uses: actions/download-artifact@70fc10c6e5e1ce46ad2ea6f2b72d43f7d47b13c3  # v8.0.0
        with:
          name: ${{ env.CUDA_CORE_ARTIFACT_NAME }}-tests
          path: ${{ env.CUDA_CORE_CYTHON_TESTS_DIR }}

      - name: Display structure of downloaded cuda.core Cython tests
        if: ${{ env.SKIP_CYTHON_TEST == '0' }}
        run: |
          pwd
          ls -lahR $CUDA_CORE_CYTHON_TESTS_DIR

      - name: Set up Python ${{ matrix.PY_VER }}
        uses: actions/setup-python@a309ff8b426b58ec0e2a45f0f869d46889d02405  # v6.2.0
        with:
          python-version: ${{ matrix.PY_VER }}
        env:
          # we use self-hosted runners on which setup-python behaves weirdly (Python include can't be found)...
          AGENT_TOOLSDIRECTORY: "/opt/hostedtoolcache"

      - name: Set up mini CTK
        if: ${{ matrix.LOCAL_CTK == '1' }}
        uses: ./.github/actions/fetch_ctk
        continue-on-error: false
        with:
          host-platform: ${{ inputs.host-platform }}
          cuda-version: ${{ matrix.CUDA_VER }}

      - name: Set up latest cuda_sanitizer_api
        if: ${{ env.SETUP_SANITIZER == '1' }}
        uses: ./.github/actions/fetch_ctk
        continue-on-error: false
        with:
          host-platform: ${{ inputs.host-platform }}
          cuda-version: ${{ env.LATEST_CUDA_VERSION }}
          cuda-components: "cuda_sanitizer_api"

      - name: Set up compute-sanitizer
        run: setup-sanitizer

      - name: Set up test repetition on nightly runs
        run: echo "PYTEST_ADDOPTS=\"--count=${{ inputs.nruns }}\"" >> "$GITHUB_ENV"

      - name: Run cuda.pathfinder tests with see_what_works
        env:
          CUDA_PATHFINDER_TEST_LOAD_NVIDIA_DYNAMIC_LIB_STRICTNESS: see_what_works
          CUDA_PATHFINDER_TEST_FIND_NVIDIA_HEADERS_STRICTNESS: see_what_works
          CUDA_PATHFINDER_TEST_FIND_NVIDIA_BITCODE_LIB_STRICTNESS: see_what_works
        run: run-tests pathfinder

      - name: Run cuda.bindings tests
        if: ${{ env.SKIP_CUDA_BINDINGS_TEST == '0' }}
        env:
          CUDA_VER: ${{ matrix.CUDA_VER }}
          LOCAL_CTK: ${{ matrix.LOCAL_CTK }}
        run: run-tests bindings

      - name: Run cuda.bindings examples
        if: ${{ env.SKIP_CUDA_BINDINGS_TEST == '0' }}
        env:
          CUDA_VER: ${{ matrix.CUDA_VER }}
          LOCAL_CTK: ${{ matrix.LOCAL_CTK }}
        run: |
          pushd cuda_bindings
          ${SANITIZER_CMD} pytest -ra -s -vv examples/
          popd

      - name: Run cuda.core tests
        env:
          CUDA_VER: ${{ matrix.CUDA_VER }}
          LOCAL_CTK: ${{ matrix.LOCAL_CTK }}
        run: run-tests core

      - name: Ensure cuda-python installable
        run: |
          if [[ "${{ matrix.LOCAL_CTK }}" == 1 ]]; then
            pip install --only-binary=:all: cuda_python*.whl
          else
            pip install --only-binary=:all: $(ls cuda_python*.whl)[all]
          fi

      - name: Install cuda.pathfinder extra wheels for testing
        run: |
          set -euo pipefail
          pushd cuda_pathfinder
          pip install --only-binary=:all: -v ./*.whl --group "test-cu${TEST_CUDA_MAJOR}"
          pip list
          popd

      - name: Run cuda.pathfinder tests with all_must_work
        env:
          CUDA_PATHFINDER_TEST_LOAD_NVIDIA_DYNAMIC_LIB_STRICTNESS: all_must_work
          CUDA_PATHFINDER_TEST_FIND_NVIDIA_HEADERS_STRICTNESS: all_must_work
          CUDA_PATHFINDER_TEST_FIND_NVIDIA_BITCODE_LIB_STRICTNESS: all_must_work
        run: run-tests pathfinder