18:04:51 #startmeeting Fedora Release Engineering 18:04:51 Meeting started Mon Nov 2 18:04:51 2009 UTC. The chair is Oxf13. Information about MeetBot at http://wiki.debian.org/MeetBot. 18:04:51 Useful Commands: #action #agreed #halp #info #idea #link #topic. 18:04:57 #meetingname fedora-releng 18:04:57 The meeting name has been set to 'fedora-releng' 18:05:02 #topic roll call 18:05:33 ping: notting jwb lmacken rdieter_work wwoods warren dgilmore spot poelcat 18:05:43 plong 18:05:51 here 18:05:54 flap flap flap flap bonk 18:06:00 * poelcat here 18:06:06 * dgilmore is present 18:07:59 * notting is here 18:09:15 #topic Fedora 12 18:09:19 It looms. 18:09:29 Supposed to be in RC state on Wed 18:09:34 lots of blockers still in the way 18:10:11 I made a promise earlier to create /something/ on Wed/Thur for testing, be it an RC, or be it a "test compose" 18:10:51 and I still need to setup all the tickets for the f12 final milestone 18:11:10 to be fair the blocker list is *mostly* MODIFIED (13/25 at last check), which indicates we probably have builds fixing them 18:11:12 so that's all I have to say about that. 18:11:24 and a lot (all but ~5?) of the things in ASSIGNED actually need moved to MODIFIED 18:11:30 so don't let the blocker list scare you too much. 18:11:48 scared-- 18:11:48 just, you know, some. 18:15:18 what's the bodhi plan? 18:16:03 We previously talked about opening bodhi for submission when we enter RC phase 18:16:21 as that's the point where we'd stop taking random crap and only take things that would block the release of F12 18:17:29 so basically it's just head down, keep going as we get closer to release 18:17:31 so, wednesday? i mean, i'd also assume we wouldn't be taking too much random stuff if we're just waiting on one or two blockers, too 18:17:46 notting: yeah, I suppose we can just pick a date now 18:18:29 no lmacken though 18:18:44 #action Oxf13 will ask lmacken to enable bodhi for F12 submissions as of Wed 18:18:58 * poelcat thinks we have that as #76 18:18:59 http://poelstra.fedorapeople.org/schedules/f-12/f-12-releng-tasks.html 18:19:49 yep 18:22:35 * warren is here, just nothing useful to add. 18:22:46 #topic open floor 18:22:57 I got nothing else for this meeting. 18:23:10 well, I don't see a point to requiring non-spin packages to be day-0 updates past RC. 18:23:18 new packages and stuff that totally doesn't effect any spin 18:23:41 warren: except that we could wind up with tagging things that never make it into the Everything tree 18:23:59 warren: it's just easier to redirect that action over to bodhi, so that there is no confusion or lost builds 18:24:24 might be good to use Bodhi for tag requests in the next cycle 18:24:38 warren: wow, that's such a great idea. I wonder why I didn't think of that. 18:24:46 Oxf13: i posted to rel-eng list about the Fedora 13 schedule 18:24:50 i know you were planning to review 18:24:57 poelcat: I saw that, it's flagged for me to read/review 18:25:16 warren: at some point, you really should read the no frozen rawhide proposal that went through releng and fesco 18:27:40 poelcat: my concern is that we'd end up with an ever-rotating release date 18:28:05 notting: that's one question I'm asking 18:28:27 in the past we said it was important to have predicatable release dates twice each year 18:28:50 but I'm wondering how important it is and to who? 18:29:13 as a result we constantly short change our release cycle (at least the initial schedule) 18:29:16 until we slip 18:29:17 poelcat: i think it was important to help users schedule there upgrades 18:31:34 It's important to ensure *something* gets released 18:31:38 and on a regular schedule 18:31:59 it was also theorized that it would help upstreams plan their major releases, and publications plan their issues 18:32:04 the length and endpoints of the cycle were basically arbitrary 18:32:16 in reality I'm not sure if that's happening, mostly because we haven't been accurate to our release dates 18:32:40 one could argue we might be better off with (e.g.) a 9-month cycle 18:34:34 but it'd take a *lot* to convince me that we'd be better off without a regular release cycle 18:34:48 a gestation period of 9 months - how very homosapien! 18:34:57 combined with no frozen rawhide, a 9 month schedule that leaves more time for bugfixing before the release (not necessarily more time for feature development) might be more acceptable 18:35:39 but this is a Big Idea and it might be better suited to a different time or place 18:35:52 I think what we want is more time to stabilize, bugfix, and polish, and the only way to do that with our 6 month schedule is to eat into the time to develop features, which is already very short 18:36:28 indeed, and I've definitely heard complaints from developers about that problem 18:36:42 that means 19 month life span for Fedora releases? 18:36:43 * poelcat doesn't think "more time" is the answer if there is no discipline around when change does or doesn't happen 18:37:09 poelcat: I'm not sure we're suggesting that there wouldn't be discipline 18:37:21 i am :) 18:37:30 IOW things would need to be different from today 18:37:56 poelcat: I'm all ears on how we can enforce our policies in any scalable way across the entire package set. 18:38:14 I think we've done a pretty darn good job with the crit path packages this freeze cycle. 18:39:04 and we're only going to get better with AutoQA coming online, which can be a fence between "package got built" and "package ends up in a published tree" 18:39:36 there is not likely to be any magic bullet here 18:40:00 it's going to have to be multiple solutions touching on different parts of the problem 18:40:31 and one part of the problem, getting feature like changes late in the cycle, is very much a product of not having enough time to do the feature development within our short window 18:40:55 not to mention the features being untestable because rawhide is very hosed in the early cycle 18:41:17 late cycle this time, I didn't need to do emergency untags late at night to prevent rawhide from being hosed the next day 18:41:21 like I had in the past 3 cycles 18:42:16 early cycle rawhide being hosed for days/weeks straight seems to be a remaining problem 18:42:33 anyway I agree with wwoods that this is a Big Picture item, one that's going to take more than just us to fix it, and at a more appropriate time than right now where we're all head down trying to get F12 out the door 18:42:54 yup. This is a FAD-worthy discussion. 18:43:00 warren: I'm attempting to fix that by removing the ebb/flow of rawhide, and just turning it into a constant flow 18:43:08 not really something to wedge into the middle of a rel-eng meeting. 18:43:14 warren: removing the build up of changes and wreakage 18:43:55 #info lots of talk about F13 schedule and release schedules in general, to be talked about more later 18:43:57 Oxf13: if everyone is testing late distro before release and a ton of untested packages build up in rawhide with few testers, is that really better? 18:44:15 warren: you assume "everyone" is testing the late distro 18:44:44 when in reality I suspect that there will be a fair amount of people who want to move on to testing things for the next release, as they're "done" with the current one 18:45:39 and a fair amount of people who will never look at anything /but/ rawhide 18:46:46 But anyway I think we said this should be talked about more Later 18:46:53 yep. 18:47:04 if there are any other topics, you got 4 minutes to bring them up 18:53:28 other than we have a few tickets waiting for more testing/tagging, i don't have any new business 18:54:33 I need to restart to test a few of the tag requests. 18:54:38 waiting for this meeting to be over 18:55:50 #endmeeting