On Fri, 2010-06-25 at 16:27 -0700, Jesse Keating wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > Can anybody tell me what went wrong with this update? It was submitted > at 15:09 on 06-23, then made it into testing at 16:19 on 06-24 and was > submitted for stable two hours later. Between that submission and the > push to stable (push to stable happened at 18:09 on 06-25) numerous > negative karma came in due to broken deps. The update went out anyway, > and now we have a mess of broken updates on a critical path package, and > have to scramble to fix it with more updates, on a Friday. > > Can anybody help me understand the scenario here? Should we start > filtering out push requests that have more than -2 karma? I talked to notting &c about this earlier, and we've hit this situation before. The 'scenario' is simply that there's really no screening between 'submit' and 'push' for stable updates, and this one was submitted to stable before any negative karma came in. There's no reliable process whereby whoever's doing the push to stable actually looks at feedback before doing it; unless they happen to have been made aware that a particular package shouldn't be pushed, they just push everything that's been submitted. Clearly the maintainer did not allow sufficient time for testing here; there's a grand 4 hour window between the update being 'pushed to testing' and 'submitted to stable'. That probably wasn't long enough for it even to hit any public mirrors. AutoQA will also help a lot here, obviously, when it's fully implemented. If implemented as planned, the depcheck tests would have prevented the update from being pushed. The requirement for proventester feedback for critpath updates, when we turn it on, should also catch problems like this in the critpath. Evo isn't critpath, though, I believe. Until AutoQA is in place to tackle this, the obvious option is for there to be a process improvement whereby whoever's doing stable update pushes at least gets notified if a package has received negative karma since the submission. I think we discussed something similar on -devel last time this happened. Not sure if there's a ticket, or how hard that would be to implement. -- Adam Williamson Fedora QA Community Monkey IRC: adamw | Fedora Talk: adamwill AT fedoraproject DOT org http://www.happyassassin.net -- devel mailing list devel@xxxxxxxxxxxxxxxxxxxxxxx https://admin.fedoraproject.org/mailman/listinfo/devel
