CI integration

Run the headless lane on every push — it's a fast, hardware-free gate that also produces code coverage (--coverage). Use the device lane occasionally (nightly / pre-release) as a fidelity check via --cross-check; it's the only part that needs hardware.

Exit codes & reports

• brighttest exits non-zero on any test failure — fail the job on it.
• --junit <path> writes a JUnit XML report for test-result UIs.
• --lcov [path] writes an LCOV file for coverage services (Coveralls/Codecov) or genhtml. It works on the headless --coverage lane (no device) and on the device lane.

GitHub Actions — headless gate (with coverage)

Runs on standard GitHub-hosted Linux runners; no Roku required. This is the everyday gate and it uploads coverage.

name: tests
on: [push, pull_request]

jobs:
  headless:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-node@v4
        with:
          node-version: '20'
      - run: npm ci
      - run: npx brighttest --coverage --lcov coverage/lcov.info --junit reports/junit.xml
      - if: always()
        uses: actions/upload-artifact@v4
        with:
          name: junit
          path: reports/junit.xml
      - uses: coverallsapp/github-action@v2
        with:
          path-to-lcov: coverage/lcov.info

For the fastest possible pure-logic gate, use npx brighttest --no-sgnode (skips @SGNode suites); run the --coverage job on a separate trigger if you want to keep the per-push gate sub-second.

Device lane — fidelity check (optional)

The device lane is the only thing that needs hardware, so it runs on a self-hosted runner on the same network as a dev device. You don't need it for coverage — use it to confirm the headless lane still matches real firmware via --cross-check (fails on any divergence). Gate it to a nightly schedule.

  cross-check:
    runs-on: [self-hosted, roku]          # a runner that can reach the device
    needs: headless
    if: github.ref == 'refs/heads/main'   # e.g. only on main / nightly
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-node@v4
        with:
          node-version: '20'
      - run: npm ci
      - name: Confirm headless matches the device
        env:
          ROKU_HOST: ${{ secrets.ROKU_HOST }}
          ROKU_PASSWORD: ${{ secrets.ROKU_PASSWORD }}
        run: npx brighttest --cross-check --host "$ROKU_HOST" --password "$ROKU_PASSWORD"

Never hard-code device credentials

Put the device IP and developer password in CI secrets (ROKU_HOST, ROKU_PASSWORD) and pass them via env:, as above. Don't inline them in the workflow or commit them.

Recommended split

Trigger	Lane	Why
Every push / PR	Headless --coverage	Fast, hardware-free gate; blocks broken logic immediately, and produces coverage.
Nightly / pre-release	Device --cross-check	Needs hardware. Confirms the fast lane is still a faithful proxy for real firmware.
Manual / pre-release	e2e run	UI journeys on a real device. Self-hosted runner near a Roku; can't run on GitHub-hosted runners.

On-device E2E in CI

The e2e lane needs a real Roku on the runner's LAN, so it runs only on a self-hosted runner. The repo ships a ready workflow at .github/workflows/e2e-device.yml — see E2E → Scaling & CI.

Coverage services

coverage/lcov.info is standard LCOV, so it feeds Coveralls, Codecov, or a local HTML report:

genhtml coverage/lcov.info --output-directory coverage/html

brighttest filters framework-internal records out of the LCOV automatically, so reported coverage reflects your code.