exclude people from giving karma?
Adam Williamson
awilliam at redhat.com
Mon Mar 10 22:55:11 UTC 2014
On Mon, 2014-03-10 at 20:18 +0100, drago01 wrote:
> On Sun, Feb 23, 2014 at 6:12 PM, Reindl Harald <h.reindl at thelounge.net> wrote:
> > https://admin.fedoraproject.org/updates/FEDORA-2014-2922/libreoffice-4.2.1.1-1.fc20?_csrf_token=a6a024f6e2d35ad3f3333b8666c1244e215a6aa2
> >
> > how can people pretend "installation went smoothly, no issue detected during basic
> > document manipulation" for packages which are not installable at all due
> > dependencie problems?
>
> https://admin.fedoraproject.org/updates/mesa-10.0.3-1.20140206.fc20
> ... again broken dep and someone gave it +1 regardless. You should
> know that "someone" very well ;)
>
> Now seriously auto qa detected the broken dep. Maybe it should give
> negative karma even if there are false positives a wrong negative
> karma is not the end of the world ...
FESco just accepted my proposal to disable autokarma if AutoQA checks
fail:
https://fedorahosted.org/fesco/ticket/1242
so that'll get done when Luke and Tim have time
( https://github.com/fedora-infra/bodhi/issues/36 ).
It doesn't make a lot of sense for AutoQA to give a -1, that's just
not...logical. If the test was correct, the update *should not go out*.
A -1 doesn't ensure that, it just adds a rather small amount of weight
on one side of a scale.
The ideal, of course, is to have a test sufficiently reliable that we
can simply bar updates from being pushed stable if it doesn't pass.
We're working on that:
https://bitbucket.org/fedoraqa/depcheck-mk-2/src
until it's done, I think disabling autokarma makes somewhat more sense
than filing a -1, as a compromise for doing some kind of enforcement
with a depcheck that's not sufficiently accurate to be used as a big
gun.
--
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | XMPP: adamw AT happyassassin . net
http://www.happyassassin.net
More information about the devel
mailing list