Please don't let updates sit in testing forever

[David Timms]

Matthias Clasen wrote:
> On Fri, 2007-06-15 at 23:06 +0200, Hans de Goede wrote:
>> Matthias Clasen wrote:
>>> On Fri, 2007-06-15 at 20:33 +0000, Kevin Kofler wrote:
>>>> Maintainers, please don't let your F7 updates linger in updates-testing 
>>>> forever. I've posted nag comments to the updates 2 weeks or more old, but I 
>>>> guess a lot of the 1-2 weeks old updates could also use getting pushed to 
>>>> stable.
>>>>
>>>> Either the update is broken or it's not. If it's broken, it should be withdrawn 
>>>> (unpushed) from updates-testing. If it works, why not mark it as stable? And I 
>>>> don't think waiting for feedback for more than 2 weeks makes sense, 1 week was 
>>>> pretty much the consensus in the discussions here and on fedora-maintainers. If 
>>>> there are no complaints in 2 weeks, the package can't be that broken, or if it 
>>>> is it most likely won't affect many people or someone would have noticed. :-)
>>>>
>>>> And no, pushing to stable after a timeout is currently not automatic.
>>> The old fedora updates tool sent nag mail...I assume that updates will
>>> be pushed to stable more regularly once that is implemented in bodhi.
>>>
>> Nag mail is not the answer, an option saying push this automatically to stable 
>> after X days is the answer. And yes I know not everyone wants this to happen 
>> for their packages, thus it should be an option.
> 
> Pushing automatically is inferior to nag mail since nag mail makes me go
> back to the update, look at possible feedback in fedora-test-list and
> bugzilla, and then make a conscious decision to push it. Automating this
> just further devalues updates-testing. IMO
Would it be useful to get some sort of estimate of download stats for 
each file in -testing ? I don't expect that it would be possible to 
count downloads from the mirrors, however, even the download stats for 
the repodata and the rpms {complete, not just the header {byte range}} 
from the sites within fedora control would give some {poor} impression 
of whether _anyone_ has actually at least downloaded it yet. A download 
doesn't mean installed nor tested.

The same availability of stats for the new build system in relation to 
counting rpm's that get downloaded directly would possibly be useful as 
well.

The actual testing side could be improved by requiring that _at least 
one_ independent {ie not the code developer nor the package maintainer} 
person reports success in bugzilla for the issue that the update is 
supposed to fix. Then it would be up 2 the package maintainer to solicit 
feedback if none has been forthcoming.

Perhaps there could be a way for non-packager users of packages to add 
themselves to a hidden list of people interested in the progress of 
certain packages, and who may be willing to run the updates-testing 
packages. If no feedback had been forthcoming the maintainer could click 
a link that sends email to the "interested" users soliciting feedback. 
This can sort of be done at the moment in bugzilla, but only if the user 
had previously found the particular bug and had added themselves to the 
cc list.

DaveT.