ci.yml

# GitHub Actions CI for Poseidon (ROS Noetic / Catkin)
# - Builds and tests the Catkin workspace on the self-hosted ROS Noetic runner (Ubuntu 20.04/Focal)
# - Publishes JUnit test results and optional coverage output
# - Keeps configuration minimal and aligned with repository conventions

name: CI (ROS Noetic)

on:
  push:
    branches: ["**"]
  pull_request:
    branches: ["**"]
  workflow_dispatch:
    inputs:
      run_hardware_e2e:
        description: "Run hardware E2E (backend websocket + headless UI)"
        required: false
        type: boolean
        default: false

defaults:
  run:
    shell: bash

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

jobs:
  build-and-test:
    name: Build and Test
    runs-on: [self-hosted, Linux, ARM64]

    env:
      CATKIN_WS: src/workspace
      # Adjust blacklist to skip heavy/non-critical packages on CI if needed
      CATKIN_BLACKLIST_PACKAGES: "mavros;inertial_sense;velodyne;ins_piksi;velodyne_pointcloud;velodyne_driver;velodyne_laserscan;gnss_mosaic_x5;tf"

    steps:
      - name: Ensure base packages
        run: |
          set -eo pipefail
          sudo apt-get update
          DEBIAN_FRONTEND=noninteractive sudo apt-get install -y --no-install-recommends \
            git ca-certificates \
            build-essential cmake ccache \
            python3-pip python3-rosdep \
            gcovr \
            ros-noetic-catkin ros-noetic-rostest ros-noetic-rosbash \
            ros-noetic-diagnostic-updater ros-noetic-angles \
            ros-noetic-tf2-geometry-msgs \
            ros-noetic-pcl-ros ros-noetic-pcl-conversions libpcl-dev \
            ros-noetic-sbg-driver \
            ros-noetic-roslint ros-noetic-nodelet ros-noetic-dynamic-reconfigure ros-noetic-sensor-msgs \
            libopencv-dev libexiv2-dev libwebsocketpp-dev \
            rapidjson-dev libssl-dev libgps-dev libboost-system-dev \
            python3-lxml libxml2-dev libxslt1-dev \
            lcov \
            python3-websockets
          sudo -H python3 -m pip install --no-cache-dir catkin-lint flake8 coverage pytest
          if [ ! -f /etc/ros/rosdep/sources.list.d/20-default.list ]; then
            sudo rosdep init
          fi
          rosdep update

      - name: Pre-clean workspace (sudo)
        run: |
          set -eo pipefail
          sudo rm -rf "$GITHUB_WORKSPACE"/*
          sudo mkdir -p "$GITHUB_WORKSPACE"
          sudo chown -R "$USER:$USER" "$GITHUB_WORKSPACE"

      - name: Checkout repository
        uses: actions/checkout@v4
        with:
          submodules: recursive

      - name: Clean Catkin workspace (self-hosted)
        run: |
          set -eo pipefail
          if [ -d "$CATKIN_WS" ]; then
            sudo rm -rf "$CATKIN_WS/build" "$CATKIN_WS/devel" "$CATKIN_WS/logs"
          fi

      - name: Resolve ROS package dependencies
        run: |
          set -eo pipefail
          source /opt/ros/noetic/setup.bash
          rosdep install --rosdistro noetic --from-paths "$CATKIN_WS/src" --ignore-src -r -y

      - name: Lint (Python and Catkin)
        run: |
          set -eo pipefail
          # Keep lint non-blocking initially; tighten later as needed
          flake8 georef || true
          catkin_lint "$CATKIN_WS/src" || true

      - name: Setup Node.js
        uses: actions/setup-node@v4
        with:
          node-version: 18

      - name: JavaScript tests (www/webroot/js)
        run: |
          set -eo pipefail
          JS_DIR="www/webroot/js"
          if [ ! -f "$JS_DIR/package.json" ]; then
            echo "No package.json in $JS_DIR; skipping JS tests."
            exit 0
          fi
          cd "$JS_DIR"
          npm ci || npm install
          npm test --if-present

      - name: Configure ccache
        run: |
          set -eo pipefail
          export CCACHE_DIR="$HOME/.ccache"
          ccache -M 1G

      - name: Build (Debug + sanitizers)
        env:
          CC: "ccache gcc"
          CXX: "ccache g++"
        run: |
          set -eo pipefail
          source /opt/ros/noetic/setup.bash
          cd "$CATKIN_WS"
          catkin_make \
            -DCMAKE_BUILD_TYPE=Debug \
            -DCATKIN_BLACKLIST_PACKAGES="$CATKIN_BLACKLIST_PACKAGES" \
            -DCMAKE_CXX_FLAGS="-O0 -g --coverage" \
            -DCMAKE_EXE_LINKER_FLAGS="--coverage"

      - name: Run tests
        run: |
          set -eo pipefail
          source /opt/ros/noetic/setup.bash
          cd "$CATKIN_WS"
          catkin_make run_tests -DCATKIN_BLACKLIST_PACKAGES="$CATKIN_BLACKLIST_PACKAGES"
          # Report results but do not fail the whole job on first adoption
          catkin_test_results build/test_results || true

      - name: Setup Playwright (hardware E2E)
        if: ${{ (github.event_name == 'workflow_dispatch' && github.event.inputs.run_hardware_e2e == 'true') || (github.event_name == 'push' && (github.ref == 'refs/heads/main' || github.ref == 'refs/heads/master')) }}
        run: |
          set -eo pipefail
          cd test/e2e
          npm install --no-package-lock
          npx playwright install --with-deps chromium

      - name: Run hardware E2E (backend + headless UI)
        if: ${{ (github.event_name == 'workflow_dispatch' && github.event.inputs.run_hardware_e2e == 'true') || (github.event_name == 'push' && (github.ref == 'refs/heads/main' || github.ref == 'refs/heads/master')) }}
        env:
          # Hardware-specific: set this to the USB serial adapter used for injecting fake NMEA DBT strings.
          # Must NOT point at /dev/sonar.
          DIAGNOSTICS_FAKE_SERIAL_PORT: "/dev/ttyUSB1"
          # Relaxed defaults; adjust if the bench needs longer to come up.
          POSEIDON_E2E_WAIT_SECONDS: "120"
          DIAGNOSTICS_WS_LAUNCH_TIMEOUT: "120"
          POSEIDON_E2E_TIMEOUT_MS: "120000"
          POSEIDON_E2E_LAUNCH_AS_ROOT: "1"
          # Fail early if these expected rows disappear.
          POSEIDON_E2E_REQUIRED_DIAGNOSTICS: "GNSS Fix,IMU Communication,Sonar Communication"
          # Avoid port/node conflicts with a system-installed Poseidon service on the bench.
          POSEIDON_E2E_EXCLUSIVE: "1"
          # Make comm diagnostics more tolerant on benches where rates can be lower than nominal.
          POSEIDON_DIAG_IMU_HZ: "5"
          POSEIDON_DIAG_IMU_WINDOW: "2.0"
          POSEIDON_DIAG_IMU_RATIO: "0.5"
          POSEIDON_DIAG_SONAR_HZ: "1"
          POSEIDON_DIAG_SONAR_WINDOW: "5.0"
          POSEIDON_DIAG_SONAR_RATIO: "0.2"
          POSEIDON_E2E_DIAGNOSTICS_REFRESH_DELAY_MS: "7000"
        run: |
          set -eo pipefail
          bash test/e2e/run_poseidon_e2e.sh

      - name: Upload hardware E2E artifacts
        if: ${{ always() && ((github.event_name == 'workflow_dispatch' && github.event.inputs.run_hardware_e2e == 'true') || (github.event_name == 'push' && (github.ref == 'refs/heads/main' || github.ref == 'refs/heads/master'))) }}
        uses: actions/upload-artifact@v4
        with:
          name: e2e-artifacts
          path: test/e2e/artifacts
          if-no-files-found: ignore

      - name: C++ coverage report (gcovr)
        run: |
          set -eo pipefail
          # gcovr 4.x cannot emit XML and HTML in one invocation; split to avoid skipping HTML output.
          gcovr -r "$CATKIN_WS/src" "$CATKIN_WS" --xml -o coverage-cpp.xml || true
          gcovr -r "$CATKIN_WS/src" "$CATKIN_WS" --html --html-details -o coverage-cpp.html || true

      - name: Python coverage
        run: |
          set -eo pipefail
          cd "$GITHUB_WORKSPACE"
          coverage run -m pytest src/workspace/src || true
          coverage xml -o coverage-python.xml || true
          coverage html -d coverage-python || true

      - name: Upload JS coverage
        if: always()
        uses: actions/upload-artifact@v4
        with:
          name: coverage-js
          path: www/webroot/js/coverage
          if-no-files-found: ignore

      - name: Upload JUnit test results
        if: always()
        uses: actions/upload-artifact@v4
        with:
          name: junit-test-results
          path: |
            src/workspace/build/test_results
          if-no-files-found: ignore

      - name: Upload C++ coverage artifacts
        if: always()
        uses: actions/upload-artifact@v4
        with:
          name: coverage-cpp
          path: |
            coverage-cpp.xml
            coverage-cpp.html
          if-no-files-found: ignore

      - name: Upload Python coverage artifacts
        if: always()
        uses: actions/upload-artifact@v4
        with:
          name: coverage-python
          path: |
            coverage-python.xml
            coverage-python
          if-no-files-found: ignore

      - name: Collect C++ coverage (optional)
        run: |
          set -eo pipefail
          # Try to collect coverage; ignore failures if toolchain is not configured for it
          lcov --directory . --capture --output-file coverage.info || true
          lcov --remove coverage.info '/usr/*' '*/devel/*' '*/opt/ros/*' --output-file coverage.info || true
        continue-on-error: true

      - name: Upload coverage artifact
        if: always()
        uses: actions/upload-artifact@v4
        with:
          name: coverage-lcov
          path: coverage.info
          if-no-files-found: ignore

      - name: Coverage summary (CPP/Python/JS)
        if: always()
        run: |
          set -eo pipefail
          python3 .github/workflows/scripts/coverage_summary.py

    # Self-hosted runner notes:
    # - Runner must be registered on the repository/org with labels: self-hosted, Linux, ARM64.
    # - Machine should run Ubuntu 20.04 with ROS Noetic installed in /opt/ros/noetic.
    # - Ensure runner user has passwordless sudo for apt/rosdep during the workflow.