PR Tests workflow (NVIDIA/tilus)

name: PR Tests
 
on:
  push:
    branches:
      - "pull-request/[0-9]+"
 
concurrency:
  group: ${{ github.workflow }}-${{ github.ref }}
  cancel-in-progress: true
 
jobs:
  check-changes:
    timeout-minutes: 30
    if: github.repository == 'NVIDIA/tilus'
    runs-on: latchkey-small
    outputs:
      should_run_tests: ${{ steps.changed-tests.outputs.any_changed }}
      should_run_examples: ${{ steps.changed-examples.outputs.any_changed }}
    steps:
      - name: Checkout Repository
        uses: actions/checkout@v4
        with:
          fetch-depth: 0
 
      - name: Check for changes in test-relevant directories
        id: changed-tests
        uses: step-security/changed-files@v46
        with:
          files: |
            tests/**
            python/**
            docs/source/_static/layout-demo/**
            pyproject.toml
          base_sha: 'origin/main'
 
      - name: Check for changes in examples
        id: changed-examples
        uses: step-security/changed-files@v46
        with:
          files: |
            python/**
            examples/**
            pyproject.toml
          base_sha: 'origin/main'
 
  test-python-compat:
    timeout-minutes: 30
    needs: check-changes
    if: github.repository == 'NVIDIA/tilus' && needs.check-changes.outputs.should_run_tests == 'true'
    runs-on: linux-amd64-gpu-l4-latest-1
    container:
      image: nvidia/cuda:13.0.0-devel-ubuntu22.04
      options: --gpus all
    steps:
      - name: Checkout Repository
        uses: actions/checkout@v4
        with:
          fetch-depth: 0
 
      - name: Install dependencies
        env:
          DEBIAN_FRONTEND: noninteractive
        run: |
          apt-get update && apt-get install -y git cmake nodejs software-properties-common
          add-apt-repository -y ppa:deadsnakes/ppa
          apt-get update && apt-get install -y python3.10 python3.10-venv python3.10-dev \
            python3.11 python3.11-venv python3.11-dev \
            python3.12 python3.12-venv python3.12-dev \
            python3.13 python3.13-venv python3.13-dev
 
      - name: Mark repo as safe for git
        run: |
          git config --global --add safe.directory "$GITHUB_WORKSPACE"
          git config --system --add safe.directory "$GITHUB_WORKSPACE"
 
      - name: Run smoke tests across Python versions
        shell: bash
        run: |
          # Derive version from git so setuptools-scm doesn't need git access during build
          export SETUPTOOLS_SCM_PRETEND_VERSION=$(python3.10 -c "
          from setuptools_scm import get_version
          print(get_version())
          " 2>/dev/null || git describe --tags --always 2>/dev/null || echo "0.0.0")
 
          echo "Using version: $SETUPTOOLS_SCM_PRETEND_VERSION"
 
          for pyver in 3.10 3.11 3.12 3.13; do
            echo "========================================="
            echo "Testing Python $pyver"
            echo "========================================="
            python${pyver} -m venv /tmp/venv-${pyver}
            source /tmp/venv-${pyver}/bin/activate
            pip install --upgrade pip
            pip install ".[dev]"
            pytest tests/kernels/matmul/test_matmul_v2.py -v
            deactivate
            rm -rf /tmp/venv-${pyver}
          done
 
  test-core:
    timeout-minutes: 30
    needs: check-changes
    if: github.repository == 'NVIDIA/tilus' && needs.check-changes.outputs.should_run_tests == 'true'
    continue-on-error: true
    strategy:
      matrix:
        runner:
          - linux-amd64-gpu-l4-latest-1
    runs-on: ${{ matrix.runner }}
    container:
      image: nvidia/cuda:13.0.0-devel-ubuntu22.04
      options: --gpus all
    steps:
      - name: Checkout Repository
        uses: actions/checkout@v4
        with:
          fetch-depth: 0
 
      - name: Setup and Install Tilus
        id: setup-and-install
        uses: ./.github/actions/setup-environment
        with:
          python-version: '3.10'
 
      - name: Run main tests
        run: pytest ./tests --ignore=tests/examples/
 
  test-examples:
    timeout-minutes: 30
    needs: check-changes
    if: github.repository == 'NVIDIA/tilus' && needs.check-changes.outputs.should_run_examples == 'true'
    runs-on: linux-amd64-gpu-l4-latest-1
    container:
      image: nvidia/cuda:13.0.0-devel-ubuntu22.04
      options: --gpus all
    steps:
      - name: Checkout Repository
        uses: actions/checkout@v4
        with:
          fetch-depth: 0
 
      - name: Setup and Install Tilus
        id: setup-and-install
        uses: ./.github/actions/setup-environment
        with:
          python-version: '3.10'
 
 
      - name: Run unit tests
        run: pytest tests/examples/ -v
 
  all-functional-tests:
    timeout-minutes: 30
    name: All Functional Tests
    needs: [test-python-compat, test-core, test-examples]
    if: always() && github.repository == 'NVIDIA/tilus'
    runs-on: latchkey-small
    steps:
      - name: Check job results
        run: |
          echo "Python compat result: ${{ needs.test-python-compat.result }}"
          echo "Core tests result: ${{ needs.test-core.result }}"
          echo "Example tests result: ${{ needs.test-examples.result }}"
 
          # Check if any required job failed
          if [[ "${{ needs.test-python-compat.result }}" == "failure" ]] || [[ "${{ needs.test-core.result }}" == "failure" ]] || [[ "${{ needs.test-examples.result }}" == "failure" ]]; then
            echo "One or more functional tests failed"
            exit 1
          elif [[ "${{ needs.test-python-compat.result }}" == "cancelled" ]] || [[ "${{ needs.test-core.result }}" == "cancelled" ]] || [[ "${{ needs.test-examples.result }}" == "cancelled" ]]; then
            echo "One or more functional tests were cancelled"
            exit 1
          else
            echo "All functional tests passed or were skipped"
          fi