Skip to content

ci_integration.md

Latest commit

 

History

History
110 lines (86 loc) · 3.77 KB

ci_integration.md

File metadata and controls

110 lines (86 loc) · 3.77 KB
Raw

CI Integration

LabWired CI runs the same labwired test command locally, in GitHub Actions, and in GitLab. Pin the runner release to v0.21.0 so a firmware change is tested against a reproducible simulator version.

GitHub Actions

Use the public LabWired Core action and select the Core CLI release with its version input:

name: Firmware simulation

on: [push, pull_request]

jobs:
  test:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4

      - name: Build firmware
        run: cargo build --release --target thumbv7m-none-eabi -p firmware

      - id: labwired
        name: Run LabWired
        uses: w1ne/labwired-core/.github/actions/labwired-test@bfd879522914b586223081c4c89ba315db4a97ed
        with:
          script: tests/firmware-test.yaml
          version: v0.21.0
          output-dir: out/labwired
          args: --no-uart-stdout

      - name: Link the automatic LabWired artifact
        if: always()
        run: echo "${{ steps.labwired.outputs.artifact-url }}" >> "$GITHUB_STEP_SUMMARY"

The public action reference is an immutable action-source pin to bfd879522914b586223081c4c89ba315db4a97ed. Its only inputs are script (required), version (default v0.21.0), output-dir, and args; it downloads the selected public CLI release archive with curl. The action writes JUnit to output-dir/junit.xml, appends summary.md to the job summary, and always uploads the entire output directory, even when the test fails. Its status, summary-md, report-html, artifact-url, and exit-code outputs are available through the labwired step ID.

Container runner

The release image has labwired as its entrypoint. Pass test directly after the image name; do not repeat labwired in the container command:

docker run --rm \
  --user "$(id -u):$(id -g)" \
  --volume "$PWD:/workspace" \
  --workdir /workspace \
  ghcr.io/w1ne/labwired:v0.21.0 \
  test --script tests/firmware-test.yaml \
       --output-dir out/labwired \
       --no-uart-stdout

When bind-mounting a workspace, pass the caller UID/GID so generated artifacts remain writable on the host. The Docker command and GitHub action accept the same test YAML: it can be a single-machine script or a world script that selects its environment through inputs.env.

GitLab CI

GitLab must clear the image entrypoint so it can start its normal job shell. The active template in integration-templates/gitlab-ci.yml uses the pinned image and then invokes labwired test.

test:firmware:
  image:
    name: ghcr.io/w1ne/labwired:v0.21.0
    entrypoint: [""]
  script:
    - labwired test --script tests/firmware-test.yaml --output-dir out/labwired --no-uart-stdout

Artifacts and reporting

Use --output-dir in every environment. A run writes result.json, snapshot.json, uart.log, and JUnit output under that directory. The GitHub action fixes the JUnit path at output-dir/junit.xml, adds a failure-safe job summary and report, and always uploads the directory; other CI environments should retain the directory so failed assertions keep their diagnostics.

Advanced: build from source

Building labwired-cli from this repository is useful for testing an unreleased commit or a local code change. It is intentionally an advanced alternative to the pinned release archive or runner image:

cargo build --release -p labwired-cli
./target/release/labwired test --script tests/firmware-test.yaml --output-dir out/labwired

Onboarding KPI tracking

For board onboarding competitiveness, core-onboarding-smoke.yml runs a deterministic smoke path and emits onboarding-metrics.json, onboarding-summary.md, onboarding-scoreboard.json, and onboarding-scoreboard.md. Its soft 3600-second threshold tracks time-to-first-smoke without blocking merges.