e2e-tests.yml

name: E2E Tests

on:
  push:
    branches: [ main, master ]
    paths:
      # Only run on code changes, not documentation
      - 'src/**'
      - 'tests/**'
      - 'pyproject.toml'
      - 'Dockerfile'
      - 'homeassistant-addon/**'
      - '.github/workflows/e2e-tests.yml'
  workflow_dispatch:

# Limit GITHUB_TOKEN permissions to minimum required
permissions:
  contents: read

env:
  PYTHON_VERSION: "3.13"
  UV_CACHE_DIR: /tmp/.uv-cache
  # renovate: datasource=docker depName=ghcr.io/home-assistant/home-assistant
  HA_IMAGE_GHCR: "ghcr.io/home-assistant/home-assistant:2026.8.0"
  # Disable Testcontainers' Ryuk reaper: it has been reported to leave
  # zombie containers on GHA runners (see #366, Ilya0527 2026-05-18).
  # The E2E fixture in tests/src/e2e/conftest.py relies instead on the
  # enclosing ``with container:`` context manager — Python guarantees
  # its ``__exit__`` fires on both normal and exception flows, so the
  # Ryuk safety net is not needed for deterministic cleanup. Refs #366.
  TESTCONTAINERS_RYUK_DISABLED: "true"

jobs:
  # Comprehensive E2E validation for main branch
  e2e-tests:
    name: E2E Tests
    runs-on: ubuntu-latest
    timeout-minutes: 15

    steps:
    - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
      with:
        submodules: true

    - name: Set up Docker Buildx
      uses: docker/setup-buildx-action@bb05f3f5519dd87d3ba754cc423b652a5edd6d2c # v4
      with:
        cache-binary: true

    - name: Install uv
      uses: astral-sh/setup-uv@c771a70e6277c0a99b617c7a806ffedaca235ff9 # v9.0.0
      with:
        version: "latest"

    - name: Set up Python
      run: uv python install ${{ env.PYTHON_VERSION }}

    - name: Install dependencies
      run: uv sync --all-extras --dev

    - name: Cache HA Docker image
      id: cache-ha-image
      uses: actions/cache@55cc8345863c7cc4c66a329aec7e433d2d1c52a9 # v6
      with:
        path: /tmp/ha-image.tar
        key: ha-image-${{ env.HA_IMAGE_GHCR }}-${{ runner.arch }}

    - name: Load cached HA image
      if: steps.cache-ha-image.outputs.cache-hit == 'true'
      run: docker load -i /tmp/ha-image.tar

    - name: Pull HA image (GHCR → Docker Hub fallback)
      if: steps.cache-ha-image.outputs.cache-hit != 'true'
      run: |
        HA_VERSION="${HA_IMAGE_GHCR##*:}"
        HA_IMAGE_DOCKERHUB="homeassistant/home-assistant:${HA_VERSION}"
        for registry in "$HA_IMAGE_GHCR" "$HA_IMAGE_DOCKERHUB"; do
          echo "Trying $registry..."
          if docker pull "$registry"; then
            if [ "$registry" != "$HA_IMAGE_GHCR" ]; then
              docker tag "$registry" "$HA_IMAGE_GHCR"
            fi
            docker save "$HA_IMAGE_GHCR" -o /tmp/ha-image.tar
            echo "Pulled and cached from $registry"
            exit 0
          fi
          echo "Failed to pull from $registry, trying next..."
          sleep 15
        done
        echo "All registries failed" && exit 1

    - name: Run full E2E test suite
      run: |
        echo "🚀 Running full E2E test suite with 3 workers..."
        uv run pytest tests/src/e2e/ \
          -n3 \
          --dist loadscope \
          --tb=short \
          -v
        echo "✅ Full E2E test suite passed"
      env:
        HAMCP_ENV_FILE: "tests/.env.test"
        GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

  # In-process MCP server backend (#1527): the SAME full E2E suite, but the
  # server-under-test is the in-process MCP server (ha_mcp_tools entry) running inside
  # each worker's testcontainer (E2E_BACKEND=embedded), driven over its ingress
  # webhook. Runs in parallel with the container lane above (a separate job — it
  # does not extend that job's runtime). Matrixed x64 + arm64 to mirror the PR
  # lane (pr.yml's e2e-validation-embedded). -n2/-n3 (vs the container lane's -n3):
  # each worker additionally builds a ha-mcp wheel and its container preinstalls
  # the fastmcp dependency tree before HA boots, so a lower worker count keeps peak
  # memory + concurrent-pip pressure in check (arm gets +1). The larger timeout
  # covers that per-worker preinstall window.
  e2e-tests-embedded:
    name: E2E Tests (Embedded Server, ${{ matrix.os }})
    runs-on: ${{ matrix.os }}
    timeout-minutes: 25
    strategy:
      fail-fast: false
      matrix:
        os:
          - ubuntu-latest           # Linux x64
          - ubuntu-24.04-arm        # Linux ARM64
        include:
          - os: ubuntu-latest
            pytest_workers: 2
          - os: ubuntu-24.04-arm
            pytest_workers: 3

    steps:
    - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
      with:
        submodules: true

    - name: Set up Docker Buildx
      uses: docker/setup-buildx-action@bb05f3f5519dd87d3ba754cc423b652a5edd6d2c # v4
      with:
        cache-binary: true

    - name: Install uv
      uses: astral-sh/setup-uv@c771a70e6277c0a99b617c7a806ffedaca235ff9 # v9.0.0
      with:
        version: "latest"

    - name: Set up Python
      run: uv python install ${{ env.PYTHON_VERSION }}

    - name: Install dependencies
      run: uv sync --all-extras --dev

    - name: Cache HA Docker image
      id: cache-ha-image
      uses: actions/cache@55cc8345863c7cc4c66a329aec7e433d2d1c52a9 # v6
      with:
        path: /tmp/ha-image.tar
        key: ha-image-${{ env.HA_IMAGE_GHCR }}-${{ runner.arch }}

    - name: Load cached HA image
      if: steps.cache-ha-image.outputs.cache-hit == 'true'
      run: docker load -i /tmp/ha-image.tar

    - name: Pull HA image (GHCR → Docker Hub fallback)
      if: steps.cache-ha-image.outputs.cache-hit != 'true'
      run: |
        HA_VERSION="${HA_IMAGE_GHCR##*:}"
        HA_IMAGE_DOCKERHUB="homeassistant/home-assistant:${HA_VERSION}"
        for registry in "$HA_IMAGE_GHCR" "$HA_IMAGE_DOCKERHUB"; do
          echo "Trying $registry..."
          if docker pull "$registry"; then
            if [ "$registry" != "$HA_IMAGE_GHCR" ]; then
              docker tag "$registry" "$HA_IMAGE_GHCR"
            fi
            docker save "$HA_IMAGE_GHCR" -o /tmp/ha-image.tar
            echo "Pulled and cached from $registry"
            exit 0
          fi
          echo "Failed to pull from $registry, trying next..."
          sleep 15
        done
        echo "All registries failed" && exit 1

    - name: Run full E2E test suite (embedded backend)
      run: |
        echo "🚀 Running full E2E suite through the in-process MCP server (embedded backend) with ${{ matrix.pytest_workers }} workers..."
        uv run pytest tests/src/e2e/ \
          -n${{ matrix.pytest_workers }} \
          --dist loadscope \
          --tb=short \
          -v
        echo "✅ Full E2E test suite (embedded backend) passed"
      env:
        E2E_BACKEND: "embedded"
        HAMCP_ENV_FILE: "tests/.env.test"
        GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

  # Update-path e2e lane (#1783/#1785): mirrors pr.yml's e2e-validation-update-path
  # on master push. Each run parametrizes over the component source — the released
  # ``stable`` git tag (proves a new server release won't break existing installs)
  # AND this checkout's working tree (proves the PR's own update machinery still
  # works) — with the ha-mcp server installed from PyPI stable in both. The test
  # then drives the in-process update that installs THIS commit's freshly built
  # wheel over the running server (the real reload/purge/restart path) and asserts
  # the server survives. A single ubuntu-latest lane runs both scenarios serially
  # (no -n), not the parallel suite. fetch-depth 0 is REQUIRED — the default
  # shallow checkout omits the ``stable`` tag the stable scenario checks out from.
  e2e-tests-update-path:
    name: E2E Tests (Update Path)
    runs-on: ubuntu-latest
    # 30m: each of the two scenarios (component@stable and component@working-tree)
    # does a first bring-up that installs ha-mcp plus the full fastmcp dependency
    # tree from PyPI stable before the in-process update swaps in the PR wheel. The
    # two containers run sequentially in this one job (observed ~50s per scenario),
    # so 30m still fits comfortably.
    timeout-minutes: 30

    steps:
    - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7
      with:
        submodules: true
        # fetch-depth 0: the update-path scenario checks the custom component
        # out at the released ``stable`` git tag, which the default shallow
        # checkout does not fetch.
        fetch-depth: 0

    - name: Verify stable tag is present
      run: |
        if ! git rev-parse --verify refs/tags/stable >/dev/null 2>&1; then
          echo "::error::The 'stable' git tag is missing — the update-path lane checks the released component out from it. Ensure checkout uses fetch-depth: 0."
          exit 1
        fi

    - name: Set up Docker Buildx
      uses: docker/setup-buildx-action@bb05f3f5519dd87d3ba754cc423b652a5edd6d2c # v4
      with:
        cache-binary: true

    - name: Install uv
      uses: astral-sh/setup-uv@c771a70e6277c0a99b617c7a806ffedaca235ff9 # v9.0.0
      with:
        version: "latest"

    - name: Set up Python
      run: uv python install ${{ env.PYTHON_VERSION }}

    - name: Install dependencies
      run: uv sync --all-extras --dev

    - name: Cache HA Docker image
      id: cache-ha-image
      uses: actions/cache@55cc8345863c7cc4c66a329aec7e433d2d1c52a9 # v6
      with:
        path: /tmp/ha-image.tar
        key: ha-image-${{ env.HA_IMAGE_GHCR }}-${{ runner.arch }}

    - name: Load cached HA image
      if: steps.cache-ha-image.outputs.cache-hit == 'true'
      run: docker load -i /tmp/ha-image.tar

    - name: Pull HA image (GHCR → Docker Hub fallback)
      if: steps.cache-ha-image.outputs.cache-hit != 'true'
      run: |
        HA_VERSION="${HA_IMAGE_GHCR##*:}"
        HA_IMAGE_DOCKERHUB="homeassistant/home-assistant:${HA_VERSION}"
        for registry in "$HA_IMAGE_GHCR" "$HA_IMAGE_DOCKERHUB"; do
          echo "Trying $registry..."
          if docker pull "$registry"; then
            if [ "$registry" != "$HA_IMAGE_GHCR" ]; then
              docker tag "$registry" "$HA_IMAGE_GHCR"
            fi
            docker save "$HA_IMAGE_GHCR" -o /tmp/ha-image.tar
            echo "Pulled and cached from $registry"
            exit 0
          fi
          echo "Failed to pull from $registry, trying next..."
          sleep 15
        done
        echo "All registries failed" && exit 1

    - name: Run update-path e2e test
      # E2E_BACKEND deliberately UNSET (container backend): the test drives its
      # own dedicated container end to end and never touches the session
      # backend, but conftest's autouse session fixture boots one regardless —
      # the default container backend boots it cheaply, while
      # E2E_BACKEND=embedded would add a wheel build + in-container preinstall
      # of the whole fastmcp tree that nothing in this lane uses.
      run: |
        echo "🚀 Running the update-path e2e lane (released component + PyPI server → PR wheel)..."
        uv run pytest tests/src/e2e/workflows/embedded/test_embedded_update_path.py \
          -m update_path \
          --tb=short \
          -v
        echo "✅ Update-path e2e lane passed"
      env:
        E2E_UPDATE_PATH: "1"
        HAMCP_ENV_FILE: "tests/.env.test"
        GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}