Testing

Testing

The test suites, what they prove, and two traps that produce confidently wrong answers. Written against 178a0c2, the commit NebulaOS 1.0.2 ships.

Per module, the way the repository's own CI does it:

cd <repository root>
python3 -m unittest extras.test_nebulaos_beeper

(From the repository root, not from extras/ - the module path is extras.<name>, which is also how scripts/ci-tests.sh invokes it.)

At 178a0c2: 29 of 31 suites, 912 tests, all passing, re-run 2026-11-10, plus test_nebulaos_compat.ManagedModulePresenceTest (5 tests), the part of that suite which is a property of this tree rather than of Klipper. The two not in that run (test_nebulaos_compat, test_nebulaos_temperature_mcu) import real Klipper modules and need the pinned Klipper composed in, exactly as scripts/ci-tests.sh documents. They are a different run, not failures.

Two traps that produce confidently wrong answers:

• Strip ANSI before matching OK. unittest colours that line when it has a TTY, so grep '^OK' never matches and every suite looks like it failed. CI is unaffected, having no TTY. A first run here reported 29 of 29 failing, entirely because of this.
• Do not use unittest discover. It reports dozens of errors on this suite that are rootdir and import artifacts. test_tmcstatus fails under discovery and passes perfectly as a module.