_binary-test-linux.yml

name: linux-binary-test

# Reusable test workflow for linux binary (manywheel/libtorch).
#
# Two execution modes selected by `use_container_directive`:
# - true (default): launch the test container via GHA's native `container:`
#   directive — checkout, env-setup, and the .ci/* test scripts all run
#   directly inside it. Used by x86 / aarch64 / cuda runners.
# - false: run on the bare runner and spawn the test container with
#   `docker run` from a step. Used by s390x self-hosted runners, which
#   register only with the bare `linux.s390x` label and lack a host-level
#   docker CLI compatible with GHA's container directive.
#
# /run.sh is generated by binary_linux_test.sh; symlinks bridge the legacy
# /pytorch and /final_pkgs paths it hardcodes to the GHA-managed workspace
# and artifact dir.

on:
  workflow_call:
    inputs:
      build_name:
        required: true
        type: string
      build_environment:
        required: true
        type: string
      runner_prefix:
        required: false
        default: ""
        type: string
      runs_on:
        required: true
        type: string
      timeout-minutes:
        required: false
        default: 240
        type: number
      use_container_directive:
        required: false
        default: true
        type: boolean
        description: |
          When true, use GHA's native `container:` directive. When false,
          run on the bare runner and spawn the test container via
          `docker run`. Set to false on runners that lack a docker CLI
          compatible with the container directive (e.g. s390x).
      ALPINE_IMAGE:
        required: false
        type: string
        default: "308535385114.dkr.ecr.us-east-1.amazonaws.com/tool/alpine"
      PYTORCH_ROOT:
        required: true
        type: string
      PACKAGE_TYPE:
        required: true
        type: string
      DESIRED_CUDA:
        required: true
        type: string
      GPU_ARCH_VERSION:
        required: false
        type: string
      GPU_ARCH_TYPE:
        required: true
        type: string
      DOCKER_IMAGE:
        required: true
        type: string
      DOCKER_IMAGE_TAG_PREFIX:
        required: true
        type: string
      DESIRED_PYTHON:
        required: false
        type: string
      LIBTORCH_CONFIG:
        required: false
        type: string
      LIBTORCH_VARIANT:
        required: false
        type: string
    secrets:
      github-token:
        required: true

permissions:
  id-token: write
  contents: read

jobs:
  test:
    runs-on: ${{ inputs.runner_prefix }}${{ inputs.runs_on }}
    timeout-minutes: ${{ inputs.timeout-minutes }}
    container:
      image: ${{ inputs.use_container_directive && format('pytorch/{0}:{1}', inputs.DOCKER_IMAGE, inputs.DOCKER_IMAGE_TAG_PREFIX) || '' }}
      options: ${{ inputs.use_container_directive && inputs.GPU_ARCH_TYPE == 'cuda' && '--gpus all -e NVIDIA_DRIVER_CAPABILITIES=compute,utility,video' || ' ' }}
    env:
      PYTORCH_ROOT: ${{ github.workspace }}
      PACKAGE_TYPE: ${{ inputs.PACKAGE_TYPE }}
      DESIRED_CUDA: ${{ inputs.DESIRED_CUDA }}
      GPU_ARCH_VERSION: ${{ inputs.GPU_ARCH_VERSION }}
      GPU_ARCH_TYPE: ${{ inputs.GPU_ARCH_TYPE }}
      DOCKER_IMAGE: pytorch/${{ inputs.DOCKER_IMAGE }}:${{ inputs.DOCKER_IMAGE_TAG_PREFIX }}
      SKIP_ALL_TESTS: 1
      LIBTORCH_CONFIG: ${{ inputs.LIBTORCH_CONFIG }}
      LIBTORCH_VARIANT: ${{ inputs.LIBTORCH_VARIANT }}
      DESIRED_PYTHON: ${{ inputs.DESIRED_PYTHON }}
      ALPINE_IMAGE: ${{ inputs.ALPINE_IMAGE }}
      AWS_DEFAULT_REGION: us-east-1
      BINARY_ENV_FILE: /tmp/env
      BUILD_ENVIRONMENT: ${{ inputs.build_environment }}
      GITHUB_TOKEN: ${{ secrets.github-token }}
      PR_NUMBER: ${{ github.event.pull_request.number }}
      SHA1: ${{ github.event.pull_request.head.sha || github.sha }}
    steps:
      - name: Configure git safe directory
        run: git config --global --add safe.directory "$GITHUB_WORKSPACE"

      - name: Checkout PyTorch
        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
        with:
          ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
          fetch-depth: 1
          submodules: recursive
          show-progress: false

      - name: Make test env permanent
        shell: bash
        run: |
          mkdir -p "$(dirname "$BINARY_ENV_FILE")"
          PYTORCH_FINAL_PACKAGE_DIR="${RUNNER_TEMP}/artifacts"
          {
            echo "PYTORCH_ROOT=${PYTORCH_ROOT}"
            echo "PACKAGE_TYPE=${PACKAGE_TYPE}"
            echo "DESIRED_CUDA=${DESIRED_CUDA}"
            echo "GPU_ARCH_VERSION=${GPU_ARCH_VERSION}"
            echo "GPU_ARCH_TYPE=${GPU_ARCH_TYPE}"
            echo "DOCKER_IMAGE=${DOCKER_IMAGE}"
            echo "SKIP_ALL_TESTS=${SKIP_ALL_TESTS}"
            echo "LIBTORCH_CONFIG=${LIBTORCH_CONFIG}"
            echo "LIBTORCH_VARIANT=${LIBTORCH_VARIANT}"
            echo "DESIRED_PYTHON=${DESIRED_PYTHON}"
            echo "ALPINE_IMAGE=${ALPINE_IMAGE}"
            echo "AWS_DEFAULT_REGION=${AWS_DEFAULT_REGION}"
            echo "BINARY_ENV_FILE=${BINARY_ENV_FILE}"
            echo "BUILD_ENVIRONMENT=${BUILD_ENVIRONMENT}"
            echo "BUILD_NAME=${{ inputs.build_name }}"
            echo "PR_NUMBER=${PR_NUMBER}"
            echo "PYTORCH_FINAL_PACKAGE_DIR=${PYTORCH_FINAL_PACKAGE_DIR}"
            echo "SHA1=${SHA1}"
          } >> "${GITHUB_ENV}"
          mkdir -p "$PYTORCH_FINAL_PACKAGE_DIR"

      - uses: actions/download-artifact@65a9edc5881444af0b9093a5e628f2fe47ea3b2e # v4.1.7
        with:
          name: ${{ inputs.build_name }}
          path: "${{ runner.temp }}/artifacts/"

      # Container-directive path: test runs directly inside the container.
      - name: Populate binary env
        if: inputs.use_container_directive
        shell: bash
        run: bash .ci/pytorch/binary_populate_env.sh

      - name: Test PyTorch binary
        if: inputs.use_container_directive
        shell: bash
        env:
          OUTPUT_SCRIPT: /run.sh
        run: |
          ln -sf "${GITHUB_WORKSPACE}" /pytorch
          ln -sf "${PYTORCH_FINAL_PACKAGE_DIR}" /final_pkgs
          bash .ci/pytorch/binary_linux_test.sh
          source "${BINARY_ENV_FILE}"
          bash -x /run.sh

      # Bare-runner path: spawn the test container manually.
      - name: Test PyTorch binary (docker run)
        if: ${{ !inputs.use_container_directive }}
        shell: bash
        run: |
          set -x
          container_name=$(docker run \
            -e BINARY_ENV_FILE \
            -e BUILD_ENVIRONMENT \
            -e DESIRED_CUDA \
            -e DESIRED_PYTHON \
            -e GITHUB_ACTIONS \
            -e GPU_ARCH_TYPE \
            -e GPU_ARCH_VERSION \
            -e LIBTORCH_CONFIG \
            -e LIBTORCH_VARIANT \
            -e PACKAGE_TYPE \
            -e PYTORCH_FINAL_PACKAGE_DIR \
            -e PYTORCH_ROOT \
            -e SKIP_ALL_TESTS \
            --tty \
            --detach \
            -v "${GITHUB_WORKSPACE}:${GITHUB_WORKSPACE}" \
            -v "${PYTORCH_FINAL_PACKAGE_DIR}:${PYTORCH_FINAL_PACKAGE_DIR}" \
            -w / \
            "${DOCKER_IMAGE}"
          )
          docker exec -t -w "${PYTORCH_ROOT}" "${container_name}" bash -c "bash .ci/pytorch/binary_populate_env.sh"
          # binary_linux_test.sh writes its test commands to OUTPUT_SCRIPT (defaults to a
          # legacy CircleCI path); /pytorch and /final_pkgs are hardcoded inside the script.
          docker exec -t -w "${PYTORCH_ROOT}" -e OUTPUT_SCRIPT=/run.sh "${container_name}" bash -c "
            ln -sf '${GITHUB_WORKSPACE}' /pytorch
            ln -sf '${PYTORCH_FINAL_PACKAGE_DIR}' /final_pkgs
            bash .ci/pytorch/binary_linux_test.sh
          "
          docker exec -t "${container_name}" bash -c "source ${BINARY_ENV_FILE} && bash -x /run.sh"

      - name: Cleanup docker
        if: ${{ always() && !inputs.use_container_directive }}
        shell: bash
        run: |
          # On bare-runner mode the container is left running by docker run --detach;
          # stop it for clean worker shutdown.
          # shellcheck disable=SC2046
          docker stop $(docker ps -q) || true