[OpenAFS-devel] hudson success (i.e. automatically building)

Simon Wilkinson sxw@inf.ed.ac.uk
Tue, 13 Jul 2010 14:40:37 +0100


On 13 Jul 2010, at 14:28, Jeffrey Altman wrote:
>=20
> There are several things that I believe should be done:
>=20
> (1) setup an account that is not your personal account for the =
execution
> of your Hudson instance.  This would be for two reasons:
>=20
>  (a) distinguish what you have personally done from the automated
>      server.
>=20
>  (b) permit all of the e-mail messages generated as a result of
>      testing to be filtered.

This should be done through gerrit's support for role accounts. I can =
either create them manually, or this could wait until we upgrade to the =
latest gerrit version when they can be created automatically (this would =
be preferred)

> (2) figure out someway of logging the configuration that is being =
tested
>    (configure script output?) and submitting it with the verified
>    action.   This is critical for determining what platforms and
>    configurations have been tested.  It will be even more important
>    when reported failures to Gerrit or openafs-bugs.

We should also upload a known broken change to gerrit and check that =
Hudson correctly identifies it. At the moment, it's sprinkling =
'verified' over every commit it sees, and we don't actually know whether =
it's telling the truth or not.

S.