depcheck test (was Re: measuring success)

[Till Maas]

On Tue, Jul 06, 2010 at 10:09:34PM +0200, Till Maas wrote:
> On Tue, Jul 06, 2010 at 03:06:37PM -0400, Will Woods wrote:
> > On Tue, 2010-07-06 at 19:21 +0200, Till Maas wrote:
> > > On Tue, Jul 06, 2010 at 09:40:01AM -0700, Jesse Keating wrote:
> > > > On 7/6/10 8:52 AM, Till Maas wrote:
> > > > > IMHO it should not be a +1 karma but some different flag that is set for
> > > > > updates that passed the tests.
> > > > 
> > > > Using karma is viewed as the path of least resistance to getting support
> > > > in current bodhi for this.  For future bodhi yes, it makes some sense to
> > > > use some different flagging mechanism.
> > > 
> > > Essentially using a different flag is just re-using the code used to
> > > flag a package as critpath-approved only with a different name.
> > > Therefore it should not need that much more effort.
> > 
> > Feel free to help write the code to prove this point!
> > 
> > > Btw. using the "path of least resistance" to implement policy
> > > changes seems to be what makes the new workflows suck for package
> > > maintainers, e.g. with the change in place using a auto-karma value of 1
> > > will become 0.
> > 
> > Well that's only one *proposed* idea. We could just as easily have
> > autoqa give a comment with neutral (0) karma on updates which pass, and
> > -1 on failed updates, which would serve all the same purposes. That
> > might be a better idea, actually.
> 
> Using karma 0 the patch could be this one:
> http://till.fedorapeople.org/tmp/0001-support-passed_autoqa.patch
> 
> Tested with:
> http://0.0.0.0:8084/updates/sos-2.2-0.fc13
> 
> To make it pass autoqa run in sqlite3 /var/tmp/bodhi.sqlite
> update comment set author = "autoqa" where update_id = 1435;
> 
> Instead of making it a bool, it might be also a good idea to use three
> values: untested, passed, failed and in case if failed a pointer to the
> test results.

Also the patch is not quite correct depending on how autoqa is supposed
to provide comments. In case it really does provide a -1 comment in case
of a broken dep, it also needs to provide a +1 comment afterwards once
the dep is fixed. This is currently not implemented. But in case autoqa
only ever adds a comment in case the update is ok, which is unlikely,
because a later update might break the deps again, then it would work.

A better documentation about what autoqa actually does would help to
write a proper patch.

Regards
Till
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: not available
Url : http://lists.fedoraproject.org/pipermail/devel/attachments/20100706/96448500/attachment.bin