On Fri, 20 Mar 2015 04:55:34 -0400 (EDT), Kamil Paral wrote: > Just one note - A great one, IMO! > the number of people giving karma is not the same as the > number of people actually using that proposed package build. For example, > some of system-critical libraries don't usually receive too much karma, > because people are not sure how to test them, and whether running app X or > Y is sufficient to test them. So they don't report any feedback for > them. But if they have updates-testing enabled, they still test them at > least unknowingly. If those libraries got broken, they would know it very > fast, and bugs would get discovered. So, in this case, we have almost no > positive feedback "things work", but the important thing here is the > absence of the negative feedback. And that's the purpose of the timeout. Exactly. Just because a package is installed does not mean it is "in use". That the very latest kernel works for me does not imply it works for everyone else with different hardware, different filesystems, and so on. I consider it dangerous to file +1 in such a case. I find it more important to actually run with updates-testing enabled and watch out for breakage to stop from entering the stable repo. And for lots of core components, for each package, a tiny guide on how to test basic funtionality would be good. From time to time, a few packagers add such testing instructions to their updates. For example, they request a specific test to be performed which they consider essential/fundamental. Still, with the steady flood of updates (and possibly a lot of flawless ones in there), it would be a tedious task to perform individual tests manually every other day and submit positive karma points manually, too. What's much more important is _the chance_ to give test-updates a try before it will be declared stable officially. That may involve - waiting for the stuff to appear on your nearby mirror, - installing the updates (possibly on another different machine, too), - rebooting (perhaps more than once to be certain everything still works), - daily usage specially of the changed software/packages. Meanwhile, if expert testers are much faster than the timeout and vote on the update to make it reach its karma threshold, fine. They are responsible for doing so. In some cases it may be more clever, more helpful, to vote +0 and point out that the test-update may affect other users differently than the tester. That is also covered by http://fedoraproject.org/wiki/QA:Update_feedback_guidelines#Neutral_feedback_or_no_feedback.3F as the case where the update seems to work for you but you don't have the insight to claim that it will be troubleless also for other users. > This also applies to many leaf packages. We have many more people running > with updates-testing than people regularly giving feedback to everything > they have installed (my personal guess would be by several orders of > magnitude). Even if the leaf package doesn't receive any karma, that > timeout interval is very useful, because it gives people time to report > issues, if they spot any. From my QA point of view, that timeout might be > even more important than the feedback provided. +1 ;-) Except for negative feedback, which is most important when it is not a false positive. > And I have been very angry about some critical path packages pushing to > stable with just 10 karma in _one or two days_. There are millions of > Fedora users, with various hardware and software combinations - we can't > afford to push something like kernel, mesa or X to stable if only 10 users > give it a thumbs up. We need to let those hundreds or thousands of people > running with updates-testing to install it as well, and give it the > invisible thumbs up, which is not present in bodhi, but which is expressed > by the lack of critical issues reported in bodhi, bugzilla or on mailing > lists. And the best way to achieve that at the moment is to make the > updates sit it updates-testing for at least a certain time. I sit here stunned, not believing what I just read. Thank you very much for that post.Great! Not everything is lost it seems. -- test mailing list test@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe: https://admin.fedoraproject.org/mailman/listinfo/test
