=== yofel_ is now known as yofel === ara_ is now known as ara === doko_ is now known as doko [09:27] skaet, cjwatson: just reading allison's email; didn't we move the DebianImportFreeze? [11:38] * ScottK thought we did. [11:39] I thought so too [15:27] Laney: could you change the ocaml-[a-z-] bits in the ocaml transition regexes to (ocaml-[a-z-]|camlp4-) please? [15:29] Laney: I think the current version actually does catch everything, but better safe than sorry [15:29] cjwatson: sure. If you want to commit it to the branch then I can just pull that. [15:33] Laney: which branch, the ~ubuntu-transition-trackers one? [15:33] or are you using git? [15:34] cjwatson: yeah, that's the one I'm using now [15:37] Laney: ok, cool - done [15:38] pulled, thanks [16:23] skaet: ping [16:24] hi micahg [16:24] hi skaet [16:25] I was wondering would it make sense to make alpha 2 after the platform rally? [16:25] or was it made the week of on purpose? [16:25] micahg, I was talking to jibel about it as well. [16:25] was planning on starting a discussion off on the subject after get off vacation next week. [16:26] skaet: ok, also, I was wondering why DIF was so early [16:26] key is going to make sure we have image build and testing over july holiday weekend. [16:28] micahg, it was brought up at UDS, and after looking at the back scroll, I see cjwatson, ScottK, and doko commenting as well. [16:28] ah, I must have missed that discussion [16:28] I'll take another pass at the schedule on Monday, and see if these two issues can get sorted. [16:28] skaet: I'm just curious why, if there's a good reason for DIF being so early, it's obviously fine [16:32] micahg, in maverick (https://wiki.ubuntu.com/MaverickReleaseSchedule ) it was between alpha 1 and alpha 2. Same story with Natty (its just that Natty had Christmas breaks, etc. in it so it seemed longer). Oneiric is following the same pattern as both of those (between Alpha1 and Alpha 2). I'll go back through the notes from UDS, and see what options might work. [16:33] I think raw elapsed time is more important to DIF than alignment with milestones [16:33] +1 [16:33] that's what I was noticing, natty was week 11, oneiric week 7 [16:33] since our milestones have very little effect on Debian development [16:34] on the other hand, maverick was week 8 [16:34] we stop autosyncing for a few days around milestones anyway [16:37] ok, this cycle is closer to maverick in terms of timing. If we push Alpha 2 out to July 7th, we can move DIF to week 8. We'd then have the rally week, to settle things down, before pushing out an image. [16:37] skaet: I don't see why it needs to be dependent on Alpha 2 at al [16:37] *all [16:38] cjwatson, desire to minimize the surprises the week before, and not have to ask folk to work the weekends to get things sorted. [16:38] it doesn't affect that [16:38] people just switch to filing big slews of manual sync requests instead [16:39] it really isn't a work-weekends kind of thing IMO [16:39] most of the worst surprises are internally generated :-) [16:40] * skaet bows to your experience on that ;) [16:40] we quite happily do autosyncs around Alpha 1, for instance [16:41] and I know A1 is a bit more free-and-easy but it still isn't generally a problem - we just stop autosyncing for a few days to let things settle down [16:41] yeah, but expectations are a bit lower. [16:41] mostly just to free up build time [16:41] * skaet nods [16:41] right, but I'm having a hard time thinking of problems caused by autosyncing for an A1 [16:41] related: could DIF for universe be later than for main? [16:41] Laney: I wouldn't recommend it [16:42] Laney: you'll get all sorts of awkward skew [16:42] I can see how it could cause problems [16:42] right [16:42] it's technically possible (if fiddly) - but I think I'd rather have manual sync requests for that [16:42] I just tend to think that on balance Universe benefits from having autosync rather than DIF [16:43] because fewer people are looking specifically at most of the packages [16:43] maybe that's true for main too.. [16:43] cjwatson, do you remember what the options/preference for DIF was from UDS? [16:43] Laney: it does, but there are enough subsystems spread across main+universe that it gets painful [16:44] skaet: I remember my opinion, but I can't remember whether it carried ... :-) [16:44] skaet: there are hopefully recordings? [16:44] * skaet looking for them now... [16:44] cjwatson: when did you want to see it? [16:45] the other problem with having DIF too early is that nominally, before DIF is when most merges are supposed to happen (and it's good for the bulk of merges to happen in a period when we're autosyncing, for similar reasons of avoiding skew) [16:45] and I think seven weeks is too short a time to get most merges done in, especially if the argument is trying to reduce the amount people have to work weekends [16:46] of course, it's not possible to determine statically which syncs are going to cause problems [16:46] maybe I'm in favour of DIF=FF :-) [16:46] * Laney ponders some more [16:47] Laney: sure, anything with lots of rdepends :) [16:47] micahg: heh, there are definitely some heuristics, but you're never going to be able to predict accurately [16:47] not that blanket autosyncing is without its problems [16:48] skaet: we went between weeks 8 and 9 for quite a few releases, and I think that generally worked out OK from the point of view of having just enough time to get merges done in [16:48] How about Alpha 2 July 7, DIF July 14, 10.04.3 July 21 ? Alpha 3 is on Aug 4, and FF is on Aug 11th. Any concerns? [16:49] alternatively, Alpha 2 June 30, DIF July 7, and leave the rest where it is [16:49] or DIF June 30, Alpha 2 July 7 [16:50] I'm not looking for *that* much extra time on DIF, just a couple of weeks - week 11 when we've done that was unusually late, I think [16:50] ok, [16:50] (and December is weird) [16:50] * skaet nods [16:50] DIF June 30 sounds better [16:50] * micahg thought week 11 for natty was early, but it was a long cycle :) [16:51] doko, are we getting any toolchain changes around then? [16:51] Lucid was week 15, but that was because of a combination of it being an LTS and Squeeze being frozen, so we wanted as much time as possible to sync fixes in from Debian testing [16:51] aside from that week 11 was definitely later than normal [16:52] how does DIF interact with the Rally? [16:52] the other thing I often say about DIF is that it's much more when you *stop* doing something than many of the other dates on the calendar [16:52] it doesn't [16:52] do we want it before or at the end? [16:52] Linaro has releases planned for Jun 16 and Jul 21 [16:52] cjwatson: the cycle was 2 weeks longer than normal though [16:52] it really doesn't matter IMO [16:52] micahg: even comparing with other 28-week cycles, it was late [16:53] I think, anyway [16:53] hm, maybe we haven't been doing 28-week cycles for long enough that I can say that [16:53] * micahg thought it was the first [16:53] LTSes do tend to throw things off [16:53] you may be right [16:54] cjwatson, June 30 DIF, July 7 Alpha 2, July 21 10.04.3 , then rest the same. [16:54] based on this discussion, june 30 seems just right :) [16:54] (also, I expect the dynamics around DIF to change in the future, since people will be able to do syncs themselves in the near future rather than having to go through archive admins) [16:55] that's OK with me - that puts DIF at the end of the rally, right? [16:55] if anything that's probably better, it means we don't need to spend quite as much time processing manual sync requests through the rally [17:02] ok, lets see if anyone can spot flaws over next couple of days, and if not will send out email early next week, and adjust the schedule appropriately.