Thanks for the stats and information. How big is the gap in testing. Is there a significant amount of package releases etc walked back because after they passed minimum time in QA and were published it turned out they were broken ? Ie. percent wise or some other metric or in the absence of that a gut assessment. Are certain areas in more need of focus than others due to criticality and lack of testing/testers ? If so what areas are those ? Wanting to get a handle on things around here so I can understand where I can be most effective in helping out. I want to look into the automated qa testing via this link https://fedoraproject.org/wiki/AutoQA Thanks again. On 02/15/2012 10:16 PM, Adam Williamson wrote: On Wed, 2012-02-15 at 21:35 -0500, Vincent L. wrote:On 02/13/2012 09:30 PM, Bruno Wolff III wrote:Note that statistics are still gathered and that future changes might depend on whether or not proventesters do a better job than average of correctly tagging builds as good or bad.Probably stating the obvious, and I am new around here, but the biggest challenge I see is that testing is not well defined. Certainly for the core items standard regressions or checklists of what items should be validated etc do not seem to be present [ or at least i can't find any ]. This naturally leads to inconsistent approaches to testing from tester to tester. There are a lot of packages, and likely a lack of staffing/volunteers to develop and maintain testplans. However as in most commercial release management having these things would help ensure each tester validated things in a similar fashion and ensure better release quality.Yes, this is broadly the problem. We have a system in place that allows you to create a test plan for a package and have it show up in the update request. See it in action at https://admin.fedoraproject.org/updates/FEDORA-2012-1766/dracut-016-1.fc17 - note the links to test cases - and details on how to actually set up the test cases to make this work are at https://fedoraproject.org/wiki/QA:SOP_package_test_plan_creation . We don't have test plans for many packages, really because of the resource issue. Jon Stanley did suggest he might work on this as his 'board advocacy' task.May I ask how many "proventesters" there are ballpark -vs- how many approximate testers of standard status participate at any given time ?We can say with precision how many proven testers there are, because there's an associated FAS group - there are 90 members of the 'proventesters' group in FAS. Active non-proven testers is a bit harder to count, but Luke can generate Bodhi statistics. There's one fairly 'famous' set from 2010 here: https://lists.fedoraproject.org/pipermail/devel/2010-June/137413.html There's a less famous report from March 2011 here: http://lmacken.fedorapeople.org/bodhi-metrics-20110330 you can get some numbers from. At the time of the 2011 report it seems like there was a roughly 1:10 proventester/regular tester ratio for F15 and F14, but it does seem to be slightly unclear. |
-- test mailing list test@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe: https://admin.fedoraproject.org/mailman/listinfo/test