Testing¶

How work gets checked before it is trusted, and how results on real hardware are recorded.

What we test¶

To write: the layers of testing that exist — what can be checked on a development host, and what can only be checked by booting a machine. Set the expectation early that most meaningful testing here needs hardware.

Running the tests¶

To write: the test invocations per tree, what they cover, and what a pass actually proves.

Testing on hardware¶

To write: the procedure for testing a build on a real machine, including how to tell the difference between a bad build and a bad flash, and how to get the machine back afterwards.

Recording a result¶

To write: how a result becomes a recorded support state on feature support. Cover the rule that matters most — a result is recorded against a device ID, never a chip or a marketing name, and never generalised to a sibling machine.

Evidence¶

To write: what has to accompany a claim for it to be believed — logs, the exact build, the firmware revision, and the machine it ran on. This is the section that keeps the support tables honest, so it is worth being strict in.

Reporting a failure¶

To write: where failures go and what to include. Distinguish a failure worth filing from a machine that simply has not been brought up yet.