Hey! On 31 July 2018 at 06:33, flocculant <[email protected]> wrote: > On 31/07/18 02:19, Lukasz Zemczak wrote: >> >> Hello everyone, >> >> The first set of official working builds for the upcoming xenial >> 16.04.5 point release (due this Thursday, August 2nd) have been added >> to the tracker [1] for all supported flavours. We had a few images for >> this milestone already but had to re-spin due to quickly spotted >> regressions. The ones now seem to be test-worthy at least. > > Would be useful to know what the regressions were ...
The first set of images were completely busted due to an out-of-sync issue on cdimage (cdimage had an older debian-installer than what was in the image itself and archive). The second batch had an issue when using the debian-installer based installer with the default GA 4.4 kernel, causing a kernel panic. This was caused by the last minute SCSI fix we got in (due to 4.4 behaving differently than 4.15 which the fix was targeting). Basically it meant that no server install using regular non-hwe kernels could be done. >> >> >> Which is why please proceed with testing as soon as possible. Even >> though we still have some days until the planned release, don't wait >> until the last minute - the earlier testing starts the more time we >> might have for fixing any blockers found during testing. > > You mean the earlier that testing starts, the sooner we can rebuild, so your > testing gets reset to zero again ;) > > It pretty much makes sense to wait before testing out in the real world > where we test manually. If people wait with the testing and then actually *do* find a release-critical regression last minute, we'll basically won't be able to re-spin and release in time. If we did as you say and did not start performing sanity tests on the two earlier batches as soon as we did, we wouldn't know things are completely busted. Now imagine this happening a day before release. Fast-tracking a fix in through -proposed to -updates and rebuilding all images takes at least a few hours + the re-testing. Yes, it's annoying that testing gets invalidated early and people have to re-test the same thing a few times, but it's a cost worth paying to not have to do an all-nighter before release in case something bad happens. Bugs can pop up everywhere, some can effect only certain flavours and can be potentially caused by even seemingly unrelated changes. This was also the reason why only after the third rebuild batch the call-for-testing was sent out: the first two weren't ready. >> >> As mentioned >> by Adam during the 18.04.1 call-for-testing: images that have no >> testing can't be released! >> >> Good luck! >> >> [1] http://iso.qa.ubuntu.com/qatracker/milestones/393/builds >> >> Cheers, >> > Cheers > > > -- > Ubuntu-release mailing list > [email protected] > Modify settings or unsubscribe at: > https://lists.ubuntu.com/mailman/listinfo/ubuntu-release Cheers, -- Łukasz 'sil2100' Zemczak Foundations Team [email protected] www.canonical.com -- Ubuntu-release mailing list [email protected] Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-release
