Re: KernelCI Workshop in Amsterdam on Aug 28th

Mark Brown <[email protected]> Thu, 21 Aug 2025 20:10:03 +0100

Newsgroups	dev.linux.lists.kernelci
Message-ID	<[email protected]>

On Thu, Aug 21, 2025 at 02:32:23PM -0300, Gustavo Padovan wrote:

[Reflowed into 

> I’ll start by listing a few discussion topics here to warm up:

Some more:

- Lab usage efficiency improvements:
 - Job prioritisation.
 - Job dependency usage (eg, don't try things that need a NFS root
   filesystem if NFS boot didn't work).
 - Deduplciation of tests/builds for the same commit (especially an
   issue at -rc1).

- Should the lab configuration be higher level (more like the
  boards/tests mapping we had with v1)?

- Onboarding documentation/tooling:
 - Programming the API (plus general API docs).
 - Command line tooling (kci-dev is pretty good here, but not well
   signposted from the docs).
 - Configuration?

- Introspection/debuggability of configuration improvements (Denys is
  already doing a bunch of work here).

> * quality of test results - we run many thousands of tests everyday, but the confidence level is not where we want it to be. Some tests and platforms are unstable, but currently we have no tracking of that.

A bunch of these are the underlying thing being tested not being stable
of course and showing that is useful...

> * Defining Hardware/Platform information: Today hw info in pretty much inexistent in KernelCI and a big limitation for showing results filtered by hw capabilities.

We're not quite that badly off - for the class of fairly inflexible SBCs
there's not much information to collect until you start getting into
stuff plugged into USB or whatever.  Not that we don't need to improve
here, but it's not quite so doom and gloom as that.

> * bisection support: how to do it?

I will get to this if nobody else does (we need it for our deployment).

signature.asc (application/pgp-signature, 488 B)