On Fri, 2014-12-12 at 11:00 +0100, Michael Schwendt wrote: > On Fri, 12 Dec 2014 00:12:37 -0800, Adam Williamson wrote: > > > I haven't seen a single person report a dep issue of this nature, > > and I spent most of release day in #fedora, and have been > > following G+, forum, and various news site feedback since. > > Violated upgrade path issues still hit users. One example: > > R in F21 older than in F19 > > https://lists.fedoraproject.org/pipermail/users/2014-December/456108.html > > Other users don't ask the Fedora Project but just give up and hope > it will work some months later (which isn't guaranteed either). That's not a dependency problem. Nothing breaks. The upgrade works, R probably still works, and the docs do suggest running a distro-sync after the upgrade if you want to resolve situations like that. Of course it's not perfect, it's just that I'm not sure anyone's suggested an alternative that's *better* yet. The best combination of ideas I can see is this: Use updates/ as a staging area during Branched freezes, and disallow pushes if the upgradepath would be violated versus stable and the packages currently *pending* for stable. The thing with that is, I'm not sure it's even viable to write a test for that, the 'currently pending' part. Without it, you can't do a simultaneous push of an update to stable for all supported releases, because the upgradepath is violated for the older releases at the time you submit the update. If you want to push the same update to all three releases you have to submit it for the newest, wait a day for it to go stable, submit it for N-1, wait a day, submit it for N-2, wait a day - it's kind of impractical. We struggle enough already with getting updates actually shipped (see the updates-testing status mails with the list of un-pushed security updates). (The other thing we could do is have fedup do distro-sync, but we've already got a whole bug report arguing about that.) -- Adam Williamson Fedora QA Community Monkey IRC: adamw | Twitter: AdamW_Fedora | XMPP: adamw AT happyassassin . net http://www.happyassassin.net -- test mailing list test@xxxxxxxxxxxxxxxxxxxxxxx To unsubscribe: https://admin.fedoraproject.org/mailman/listinfo/test