2.7 KiB

Raw Blame History

Testing

Regression tests live under tests/, one topic per file, grouped by domain (routing/, hbp/, obp/, …). Full file index: tests/README.md (repo root, maintainer checkout).

They use an in-process deterministic harness (no Twisted reactor, no UDP sockets).

Install dev dependencies

python3 -m pip install -e ".[dev]"

Use the project interpreter, e.g. /opt/.pyenv/versions/3.11.8/bin/python3.

Run tests

Full suite:

python3 -m pytest tests/ -q

By domain (while working on one area):

python3 -m pytest tests/routing/ -q
python3 -m pytest tests/hbp/ -q
python3 -m pytest tests/echo/ -q

By file (recommended for point checks):

python3 -m pytest tests/routing/test_unit_data_routing.py -q
python3 -m pytest tests/echo/test_rekey_playback.py -q

Single test:

python3 -m pytest tests/routing/test_startup_subscriptions.py::test_startup_bridge_routes_voice_after_apply -q

Collect only:

python3 -m pytest tests/ --collect-only -q

Markers

Registered in pyproject.toml:

Marker	Use
`@pytest.mark.behavior`	Integration-style regression (P0/P1)
`@pytest.mark.smoke`	Quick routing smoke checks

Harness overview

Component	Role
`DeterministicScenario`	Wires `RoutingUseCases` with fakes and packet capture
`scenario.routing`	Routing use-case facade (`dmrd_received`, timers)
`inject_hbp` / `inject_unit` / `inject_obp`	Public ingress paths into `dmrd_received`
`PacketCapture`	Records outbound `send_to_system` + parsed DMR fields
`FakeReportSender` + `ReportingUseCases`	Reporting events (via `scenario.report_factory.events`)
`tests/harness/assertions.py`	Reusable asserts: `assert_forwarded`, `assert_report_event`, …

Full audit (scores, red-test table, inventory): internal docs-priv/en/test-audit.md in maintainer checkouts.

Writing a regression test

Docstring: Regression: if X breaks, this test fails because Y.
Prefer a dedicated file (or the smallest existing file for the same topic).
Enter through a public path (inject_*, use-case API, or @pytest.mark.unit for pure domain).
Assert an observable outcome (capture, master.sent, report strings), not only private flags.
Add a negative path where behaviour differs.
Red-test: break the production condition → test must fail → revert.
Mark @pytest.mark.behavior or @pytest.mark.unit.

Do not chase coverage percentage; chase regressions operators would notice on RF.

v2 policy

Each v2 feature branch should include or extend regression tests in the same PR. Baseline task V2-TST-001 tracks bringing tests/ into the repo on develop.

2.7 KiB Raw Blame History