test Digest, Vol 77, Issue 32

[Till Maas]

On Thu, Jul 15, 2010 at 02:05:12PM -0700, Adam Williamson wrote:
> On Thu, 2010-07-15 at 22:13 +0200, Till Maas wrote:
> 
> > Yes, but what exactly do you want to have? If you do not want to test
> > updates for a certain package ever, just remove the package from the
> > machine. And if the package is already installed and did not cause any
> > trouble (e.g. because of broken deps or conflicts), then afaik a zero
> > karma comment is ok.
> > 
> > So would it be enough to have a short-cut for this 0-karma comment with
> > a predefined text or should this new command make f-e-k create a local
> > list of updates to ignore?
> 
> No. As I wrote in my initial mail, I've had one maintainer complain
> already about having lots of (to them) useless 0 karma comments on an
> update. Comments saying 'I can't really test this but it didn't break
> boot' are kinda useful for critpath updates, but really useless for
> non-critpath updates which we would never expect to break boot in the
> *first* place. The only kind of feedback that's useful for those is
> feedback from people who've actually tested the package directly.
> Posting useless 0-karma comments just to make the package not show up in
> f-e-k any more is kind of an abuse of process, because f-e-k is just a
> helper widget, the real point of the process is the public feedback it
> generates, and if we generate useless feedback just to make our helper
> widget happy, we're not contributing anything positive.

In general every package might break booting the machine, but this is
not the most likely kind of trouble one can already test with only
installing the package. It's broken deps, file conflicts, broken
scriptlets or broken obsoletes. Even an update not being available in
a repo was noticed by me already three times. And for some updates it
seems that the only test feedback one can get is that it was installed
and did not cause any noticed trouble. Even a 10 days old F12 critpath
update I submitted received only one 0-karma feedback, while the
similiar F13 update got four +1 within three days. And the critpath
build of the update is even core dep. But I agree that it might be
annoying if there are only many 0-comments, but it is imho worse to not
get any feedback for any other update. But the real problem here is that
there is no better way to provide this kind of feedback currently. Or a
way to aggregate this kind of feedback in bodhi without cluttering up
the comments.

> I do agree with 'just remove the package', but there might be cases
> where the tester has it installed for some real reason but isn't exactly
> sure how to manually exercise it, for those cases it'd be nice to have a
> 'please don't show me this one again' button, I think.

So a local list would be enough for this case? How about using a hybrid
approach. If an update is to be ignored, only another 0-comment with a
"installs ok" comment is added if there are not yet say two. And if the
comments exist, the update is added to a local ignore list?

Then there is at least some feedback and people can ignore the update.

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/test/attachments/20100716/6983d9bd/attachment.bin