Jan 23 12:11:07 ok... I guess we've waited long enough... let's get this started... first off, for all of you developers in here that aren't releng... please try to refrain from comments until the end... if you've got a question or feel like you definitely need to say something, do so, but try to hold off, since this meeting will likely be long enough as it is... Jan 23 12:11:33 so... let's get this rolling Jan 23 12:11:51 the first topic is 2008.0's release... Jan 23 12:13:12 I was thinking of doing the initial snapshot on Feb 1... that would give us an approximate release around mid-March... Jan 23 12:13:30 does that work for everyone? Jan 23 12:13:36 your docs liaison sez: works for me. when you need the HB tarballs? :) Jan 23 12:14:21 wolf31o2|work, works for me Jan 23 12:14:27 nightmorph: by Feb 29, I'd say Jan 23 12:14:29 wolf31o2|work: well, I hoped to have gcc-4 finally, but I think I can bury that thought along with some other things. but either way, its fine by me. Jan 23 12:14:52 i'm away on most weekends starting in mid-february until mid-march ... plus several events (fosdem, chemnitz, open-expo bern) are taking place ... not sure if that timeframe will work for me Jan 23 12:15:00 wolf31o2|work: sooo much time. i'm already letting the rest of the GDP do the docs work as it is :) Jan 23 12:15:14 reso worksforme Jan 23 12:15:40 sorry to interrupt, but is anyone else keeping a meeting summary? Jan 23 12:15:51 dberkholz: well, I'm keeping the log :) Jan 23 12:16:02 wolf31o2|work: xfce has only one bug that shouldn't go into release, i'm backporting the fix from upstream _now_ so it can be marked stable now (or in few days) Jan 23 12:16:04 summary can be created from the log post-meeting :P Jan 23 12:16:05 dertobi123: hrrrmn... OK... we're talking about 6 weeks of time... do you think it will be a problem, or... ? Jan 23 12:16:16 drac: thanks Jan 23 12:16:52 drac: xfce is bug free! whatchu talkin' 'bout, willis Jan 23 12:17:02 wolf31o2|work: depends on how many unforseen problems we encounter ... if things go the easy i see no big problem ... if things won't ... Jan 23 12:17:14 the easy way* Jan 23 12:17:53 given the fact that every little stage-rebuild takes around 24h for each hppa1.1 and hppa2.0 Jan 23 12:18:11 dertobi123: understandable... seems to be the status quo for releases, anyway... what machine are you using? Jan 23 12:18:33 wolf31o2|work: The nightmare month will be over for me, so I should have some time to spare to build alpha releases. We still have a nasty showstopper bug with 2.6.23.* and 2.6.24*, but we'll figure it out, I guess. Jan 23 12:18:46 wolf31o2|work: depends ... mostly hake (dev box at osl) plus my c3700 at home which is somewhat faster ... Jan 23 12:19:31 one thing I would like to try to do this release is to have us build on developer machines, rather than our personal machines... this way someone else can help out easily if there's a problem with time/etc... so if we don't have access to sufficient hardware (as a dev box) I'd like to know... Jan 23 12:19:41 dertobi123: there *might* be free hardware availble soon thorugh one of my coworkers, I'll get back to you on that once something materializes. Jan 23 12:19:51 dertobi123: would access to a J5600 help? I have one sitting at home Jan 23 12:20:21 I think hppa SMP support is actually "working" these days Jan 23 12:20:36 wolf31o2|work: not sure ... smp isn't that stable (if stable at all) ... might be worth a try Jan 23 12:20:40 Blackb|rd: there's an alpha machine available as a dev box to the alpha team, right? Jan 23 12:20:54 another option might be to get hosting for my c3700 somewhere nearby Jan 23 12:21:04 dertobi123: OK... I'll pull the machine out and try to get it loaded up for you in the next few days... Jan 23 12:21:16 wolf31o2|work: Yes Jan 23 12:21:18 I can host my box with my company, if you need it... so that's an option Jan 23 12:21:21 Blackb|rd: k... Jan 23 12:21:28 wolf31o2|work: :) Jan 23 12:21:29 wolf31o2|work: the ES40 we spoke of. Jan 23 12:21:34 right Jan 23 12:21:35 wolf31o2|work: something differently. whats gonna be part of 2008.0 ? are we talking about the new profile layout there, or "just" a fresh snapshot w/ security fixes and the usual things ? Jan 23 12:22:02 * nightmorph wasn't aware 2008 profiles even existed yet Jan 23 12:22:10 nightmorph: they don't :) Jan 23 12:22:40 phreak``: was going to get to that in just a second... I am not sure that we'll have time to do the profile switching... but I'm going to try to come up with it... the plan will be to create "normal" 2008.0 profiles, and I'll work on making them multi-parent and such... there's even already a multi-parent SVN repo out there... I just haven't thrown anything in it, yet Jan 23 12:22:54 can we ban that guy (Hz_SNaKE and BillGates)? Jan 23 12:23:14 agaffney: why? he's not doing anything... it's +m Jan 23 12:23:16 wolf31o2|work: sorry for being jumpy then :) Jan 23 12:23:24 phreak``: np... Jan 23 12:23:27 * agaffney shrugs Jan 23 12:23:42 there's always ignore joins parts quits... Jan 23 12:23:43 armin76: you've got dolphin and another box, right? (ia64) Jan 23 12:24:28 just dolphin iirc Jan 23 12:24:40 rangerpb: you have timberdoodle and another box, too, right? plus luna, for G4... Jan 23 12:24:44 phreak``: k... Jan 23 12:24:52 wolf31o2|work: its rooster now ;) Jan 23 12:24:58 wolf31o2|work: timberdoodle is history :S Jan 23 12:25:13 heh... OK Jan 23 12:25:35 wolf31o2|work: mind if i fix a typo on the releng project page? Jan 23 12:25:48 armin76: also, you've got bender for sparc, right? Jan 23 12:25:52 phreak``'s role: "Hardended Liason" Jan 23 12:25:54 * phreak`` kicks over the rocks armin76 and rangerpb are hiding beneath. Jan 23 12:26:03 nightmorph: ehh, I'm going to update it after the meeting, anyway... Jan 23 12:26:15 okay. s/liason/liaison for all instances too :) Jan 23 12:26:18 wolf31o2|work: he has. (armin that is) Jan 23 12:27:03 k Jan 23 12:27:03 nightmorph: fix me instead ... :P Jan 23 12:27:19 woot? Jan 23 12:27:20 yeah Jan 23 12:27:39 k... I'll get w/ the spankster later on his boxes... we can continue Jan 23 12:27:42 wolf31o2|work: yes, dolphin and beluga (ia64) and bender for sparc Jan 23 12:27:56 so... 2008.0... what do you guys want to do with it... is there anything special you'd like to see us do? Jan 23 12:27:57 jaervosz: still here ? Jan 23 12:28:13 wolf31o2|work: let's finally get rid of 2.4 ;) Jan 23 12:28:18 wolf31o2|work: gcc4 for me :P Jan 23 12:28:24 Blackb|rd: that too. Jan 23 12:28:35 Blackb|rd: which means, I do have to restructure profiles :| Jan 23 12:28:36 yay, no 2.4! makes it easier to maintain the docs Jan 23 12:28:38 phreak``: yeah Jan 23 12:28:42 phreak``: gcc4 for hardened? :) Jan 23 12:28:44 also, gentoo-sources on all arches where possible Jan 23 12:28:51 * jaervosz reads backlog Jan 23 12:28:55 I was planning on making a new sub-profile type (if we do multi-parent)... "developer" which would inherit both desktop and profile and set all the pretty "developer" FEATURES by default... Jan 23 12:29:01 nightmorph: which arches aren't on gentoo-sources? Jan 23 12:29:09 Is any arch still depending heavily on 2.4? Jan 23 12:29:18 heavily? probably not Jan 23 12:29:19 wolf31o2|work: developer profile sounds awesome to me Jan 23 12:29:23 but a few machines here and there Jan 23 12:29:26 agaffney: mips Jan 23 12:29:41 Blackb|rd: sparc32 ... which isn't maintained for years :P Jan 23 12:29:43 Blackb|rd: none at all, besides mips and maybe some sparc's Jan 23 12:29:47 wolf31o2|work: devel profile++ Jan 23 12:29:55 wolf31o2|work: sounds like a plan :D Jan 23 12:29:56 Blackb|rd: not anymore... I think alpha/sparc are the few that still have some machines that don't work as well under 2.6 as 2.4, but that number is dwindling Jan 23 12:30:01 agaffney: mips, hppa, and alpha either have a non-gentoo-sources kernel, or else support one equally with gentoo-sources Jan 23 12:30:10 nightmorph: hppa is now on gentoo-sources Jan 23 12:30:12 nightmorph: alpha doesn't Jan 23 12:30:13 sparc at least has said sparc-sources is deprecated, though still available. not recommended though Jan 23 12:30:16 oh? Jan 23 12:30:19 cool! Jan 23 12:30:24 that's new. /me notes that in the handbook changes Jan 23 12:30:25 nightmorph: that's recent Jan 23 12:30:37 wolf31o2|work: We've retired 2.4 completely recently and haven't and any complaints (but von report froma very happy user who's now using 2.6) Jan 23 12:30:49 arm,mips,m68k,s390,sh are the ones that don't have gentoo-sources keyworded Jan 23 12:30:56 agaffney: as long as no pa-patches are releases ... i recall a thread on the parisc list that pa-patches shall be revived (but that was some months ago) Jan 23 12:30:59 nightmorph: yeah, sparc is gentoo-sources, but sparc-sources is still there for people wanting 2.4 (like the crazy people still running sparc32 boxes) Jan 23 12:31:02 released* Jan 23 12:31:10 s390 and arm have 2.6.20 only Jan 23 12:31:11 agaffney: fortunately, sparc32 is not supported ;) Jan 23 12:31:17 nightmorph: indeed :P Jan 23 12:32:21 OK... so I'd also like to change a couple ways we do things... for one, as soon as we finish up a release, I'd like to go ahead and create the new release stuff in the repos... so the releng repo, we'd create the specs... in the tree, we create a new dev profile... I'd *like* to do the same w/ the Handbook, but that's up to the docs guys... the idea is that by doing this, we can update the profiles and such immediately, and it's Jan 23 12:32:21 also useful for the next topic (automated builds), so I'll hold off until then... Jan 23 12:32:53 hmm, a profile for the handbook? whatcha mean? Jan 23 12:33:27 no no... sorry... I mean as soon as we finished up 2008.0, we'd break off the 2008.1 draft... so we can make changes as we go so you don't end up with all the changes at the end Jan 23 12:33:52 of course, in your case, it likely means having to update multiple handbooks, so if you'd prefer we just bug everything along the way, that's fine, too Jan 23 12:33:57 well...many of the changes afterward get merged into the current handbook Jan 23 12:33:59 wolf31o2|work: that'd also facilitate early tweaking which makes it easier to try out new stuff. Jan 23 12:34:07 we also keep a section on errata and changes since release Jan 23 12:34:20 but i see what you're saying Jan 23 12:34:38 nightmorph: OK... so you already have a decent process in place... if we don't need to change anything there, we won't... Jan 23 12:34:52 Blackb|rd: exactly... we can try out really stupid stuff along the way... Jan 23 12:34:54 wolf31o2|work: *shrug* our process is open to change Jan 23 12:34:55 hehe Jan 23 12:35:11 we like trying out really stupid stuff Jan 23 12:35:20 actually, if that rapid rolling release thing we talked about ever gets going, we won't have to worry about creating separate 200*.* dirs. which would be nice. Jan 23 12:35:25 wolf31o2|work: +1 from me on that matter :P Jan 23 12:35:33 nightmorph: honestly, you tell me if anything needs to change... the only other docs-related change I see is trying to force the handbook updates prior to the last minute (as has been the case for a long time) Jan 23 12:35:36 agaffney: yeah, we do. thats why opensource is sooo great :P Jan 23 12:35:55 nightmorph: sure, but that would be a ways off, anyway... Jan 23 12:36:04 agaffney, hey, we are good at that Jan 23 12:36:11 wolf31o2|work: what do you think about a deadline for stuff that'd change the handbook? Jan 23 12:36:25 wolf31o2|work: well, assuming we don't have to worry about releases being changed (2007.1 to 2008.0), actually, i like your idea of just creating the new dir immediately Jan 23 12:36:49 nightmorph: well, its a svn repo :P Jan 23 12:36:54 nightmorph: svn mv ftw! Jan 23 12:36:55 the dirs can always be renamed/moved Jan 23 12:36:56 Blackb|rd: well, there's been one in the past and nobody met it... hehe... this time, we'll set one... and if you don't get your stuff in, it'd be bad Jan 23 12:37:13 yes, you'll get a stern talking-to Jan 23 12:37:14 wolf31o2|work: i.e. "one week before release, nothing may be changed (except security yada yada) that'd affect docs" Jan 23 12:37:27 renaming dirs sucks. i still haven't come up with a migration plan. potential problem is that even with a separate next release handbook, double-commits would have to be made Jan 23 12:37:30 wolf31o2|work: splendid, I like that :) Jan 23 12:37:32 in some case Jan 23 12:37:39 phreak``: hah, the handbooks are in CVS, foo Jan 23 12:37:48 nightmorph: well, ..... :P Jan 23 12:37:59 nightmorph: though luck then :) Jan 23 12:38:07 nightmorph: you can do the handbooks however is easiest for you Jan 23 12:38:10 it's just a suggestion Jan 23 12:38:17 don't let us tell you how to do your job :) Jan 23 12:38:25 agaffney: easiest for me is tracking what YOU guys are doing! Jan 23 12:38:28 RCS all the way. Jan 23 12:38:29 and then cursing CVS Jan 23 12:39:02 we just need to move anything that's not gentoo-x86 over to SVN Jan 23 12:39:11 all that reminds me yet again, I *really* need to kick vapier's rock Jan 23 12:39:24 phreak``: careful...he might enjoy that Jan 23 12:39:24 phreak``: stomp his foot, too. Jan 23 12:39:29 also, I'd like to try to release a 2008.0 beta... which is why we want the handbooks as soon as possible, so we can put them on the beta... the beta would essentially be an entire release... handbook and all... and I want to *force* a QA period after the beta... say, a week Jan 23 12:39:35 agaffney: he actually does. ha. Jan 23 12:39:55 wolf31o2|work: how official would a beta be? how much in advance of the actual march-ish release? Jan 23 12:39:57 wolf31o2|work: ain't a week a bit short ? Jan 23 12:40:04 wolf31o2|work: experience tells me to make that two weeks. or at least expect them. Jan 23 12:40:23 I tend to agree with phreak`` and Blackb|rd Jan 23 12:40:24 I think the beta should probably be half the time of the release cycle frankly Jan 23 12:40:32 if i can just roll up the existing draft handbooks with the draft disclaimer on them (i.e. not make them replace the "live" toplevel dir), then i'm fine with that Jan 23 12:40:33 that gives them time to test and us time to fix whatever Jan 23 12:40:42 nightmorph: ehh... it would be after the Feb 29 docs due date... Jan 23 12:40:51 k Jan 23 12:41:04 rangerpb: that might be too long: things start to linger once you don't have a tight schedule. Jan 23 12:41:26 that's rather my fear... getting swamped by security bugs at the last minute again Jan 23 12:41:28 rangerpb: uh, thats a bit much frankly :) sure, having the beta 2-4 weeks before release, ain't giving us much time, but doing the beta 3 months before release, puts stuff kinda outa the frame. Jan 23 12:41:30 two weeks beta time plus another two weeks for another (if needed) beta would make sense imho Jan 23 12:41:37 uh, it's a six week cycle, half would be 3, which is only 1 more than the suggested two Jan 23 12:41:43 if things go well with the second beta we could s/beta/release Jan 23 12:41:56 me and my handbooks wouldn't mind Jan 23 12:42:09 well, my plan for betas is we release them as needed... so if beta1 is really good, we tweak the few things and release... if we need another beta, we do it and push the release date out Jan 23 12:42:31 * nightmorph is all in favor of the idea Jan 23 12:42:31 sound reasonable? Jan 23 12:42:40 +1 Jan 23 12:42:46 wolf31o2|work: then I think a two week test period would be nice. You can always shorten the span with the consecutive beta(s) and rc(s) Jan 23 12:43:01 wolf31o2|work: so, ++ from me. Jan 23 12:43:17 one thing I'll want all of you guys to come up with (on gentoo-releng list, please) is a list of things that you would like for us to have checked on the media... I want to come up with a comprehensive checklist... I'll start the threads after the meeting... Jan 23 12:43:50 wolf31o2|work: so when should the first beta be released? given the initial snapshot is done on february 1st - mid-february gives enough time for building and initial testing, plus enough time for bug-fixing and polishing the media? Jan 23 12:43:56 wolf31o2|work: arch specific stuff too? Jan 23 12:44:03 Blackb|rd: yes Jan 23 12:44:17 wolf31o2|work: mmmmmhmmm. Jan 23 12:44:19 i thought sometime after the 29th? Jan 23 12:44:25 wolf31o2|work: then the arch times will need to do some QA themselves, of course. Which is a-ok Jan 23 12:44:26 dertobi123: one second... I've got a tentative schedule running as we discuss, and I'll propose the whole schedule once we've got our ideas down Jan 23 12:44:37 nightmorph: ^^^ Jan 23 12:44:41 k Jan 23 12:45:17 Blackb|rd: they're supposed to be doing that already... hehe... I've mailed the QA team, and (at least Halcy0n) they're willing to help out a bit... so we'll coordinate with them once we have these lists Jan 23 12:45:33 wolf31o2|work: splendid Jan 23 12:45:37 * phreak`` giggles about QA & Halcy0n Jan 23 12:45:52 also, we'll want a list of "stuff you want to know" from users about their machines/environments for the beta... things like chipsets/cpu/ram/etc... Jan 23 12:46:20 basically, a text doc that people can grab along w/ the beta so they'll know what info we definitely want from them when they file bugs Jan 23 12:46:26 * phreak`` goes subscribing to the gentoo-releng ml Jan 23 12:46:32 didn't we use to have beta testing to some extent in the past? Jan 23 12:46:34 wolf31o2|work: yeah, getting a list of hardware that's *Worthwile* to care about would be nice. Jan 23 12:46:49 welp: we had something like it for 2006.1 (or was it 2006.0?) Jan 23 12:46:51 welp: private, yeah. but no public beta testing. Jan 23 12:46:55 wolf31o2|work: especieally for niche archs. Jan 23 12:46:56 i seem to remember emailing someone, and then getting a location where i can download stages from Jan 23 12:46:57 welp: not really... there were 1.4 release candidates, but that was because 1.4 had no schedule and dragged on and on forever Jan 23 12:46:57 welp: very limited and nothing "official" Jan 23 12:47:00 phreak``: i don't think it was private Jan 23 12:47:40 welp: we had a release tester program... and I want to reinstate that, actually... I'm going to try to solicit help from other teams in doing so, so I'm not trying to coordinate it all Jan 23 12:47:52 wolf31o2|work: that must be what i'm thinking of.. Jan 23 12:48:33 one thing I am definitely doing is trying to make it so that we don't rely so much on me for things to get done... people will be able to update things on their own more... like, we're converting the releng repo to SVN and I'm going to keep the release checklist in there, so everyone can update themselves as they get things done Jan 23 12:48:39 roger55 was involved in it iirc Jan 23 12:48:45 welp: yes, he was coordinating it Jan 23 12:48:56 and it wasn't private Jan 23 12:49:02 yeah Jan 23 12:49:04 it just wasn't publicized well Jan 23 12:49:12 wolf31o2|work: same for the snapshot, right (at least you mentioned it in the past) ? Jan 23 12:49:22 phreak``: I'm going to try to, yes... Jan 23 12:49:24 that was fun, would be cool if it was better publicized and possibly more arch-oriented? Jan 23 12:49:40 wolf31o2|work: just yell if you need any help, please. Jan 23 12:49:41 it didn't work out all that well. we fixed a few problems, but many people just wanted the media early and didn't report problems, or gave us worthless bug reports Jan 23 12:49:46 welp: yeah, depends on what we can accomplish in a short time... but we'll have the beta, no matter what Jan 23 12:49:52 right Jan 23 12:49:53 wolf31o2|work: cool Jan 23 12:50:03 agaffney: that'll be the main problem: getting good reports. Jan 23 12:50:07 welp will do it Jan 23 12:50:12 not having some kind of form or anything to let people know what we wanted to know was one of the major reasons it didn't work out so well Jan 23 12:50:17 Blackb|rd: which is why wolf31o2|work asked for a list of things that people will want with a bug report Jan 23 12:50:18 we never got good reports Jan 23 12:50:24 right Jan 23 12:50:32 agaffney: half the rent, yes. Jan 23 12:50:39 ok... so I've got a tentative schedule Jan 23 12:50:54 What about a bug wrangler for release bugs? Jan 23 12:51:09 So some are filtered beforehand Jan 23 12:51:13 Opfer: have one? Jan 23 12:51:14 *G* Jan 23 12:51:22 we could use jakub! Jan 23 12:51:23 * agaffney runs Jan 23 12:51:24 Opfer: there really aren't that many of them to warrant it, I don't think... we'll have to see how much our bug reports explode when it comes time for the beta Jan 23 12:51:32 agaffney: i was just thinking that. but at least i didn't say it ;) Jan 23 12:51:35 agaffney: somehoe I expected that line :)) Jan 23 12:51:49 agaffney: oh, save it. Jan 23 12:51:50 buhwuh? Jan 23 12:52:06 Some good bug reports are better than none...me out. go on. :) Jan 23 12:52:11 anyway, show us this proposed timeline. i've got a notepad all ready Jan 23 12:52:23 nightmorph: windows notepad? :> Jan 23 12:52:47 welp: yeah, exactly that! Jan 23 12:52:59 * Blackb|rd uses EDLIN Jan 23 12:53:07 * agaffney cringes Jan 23 12:53:33 ok... schedule... Snapshot on Feb 1, Docs Due (Releng)/Snapshot frozen (pre-beta) on Feb 25, Docs Finalized/Due on Feb 29, Beta1 released on Mar 3, Final upload on Mar 14, Release on Mar 17... Jan 23 12:54:02 that's doable Jan 23 12:54:04 basically, everything after Beta1 is *very* dependent on the quality of the beta... Jan 23 12:54:14 so this assumes no major problems Jan 23 12:54:22 i'd like a beta *before* a snapshot freeze Jan 23 12:54:32 umm... that's the beta snapshot freeze Jan 23 12:54:40 wolf31o2|work: how do you plan to avoid getting rear-ended by sec bugs? Jan 23 12:54:51 wolf31o2|work: (schedule is fine by me) Jan 23 12:55:10 Blackb|rd: still working on that one, but coming up with a way for people to update the snapshot themselves will help out a lot Jan 23 12:55:22 hah, beta1 release one day before my birthday, good timing Jan 23 12:55:23 Blackb|rd: part of the problem last time was that wolf31o2|work was trying to update the snapshot for *every* sec bug, isntead of just the ones that affected the media Jan 23 12:55:24 wolf31o2|work: *nod* Jan 23 12:55:31 wolf31o2|work: was about to say that :) since it keeps you from doing all the work ;) Jan 23 12:55:35 angelos will be too drunk to do anything! Jan 23 12:55:36 agaffney: I kinda noticed. Jan 23 12:55:41 bah, even andew is faster :P Jan 23 12:55:49 right... I was definitely the single point of failure on that one... heh Jan 23 12:56:11 ok... so schedule looks alright (it can be adjusted as we go, of course...) Jan 23 12:56:19 tinderbox and make binpkgs available for the security updates? Jan 23 12:56:28 i think the arch releng coordinator definitely needs to keep in touch with the arch security liaison here Jan 23 12:56:36 nixnut: won't help us all that much Jan 23 12:56:39 the 29/2 mark for b1 should be the point where stuff hinges on. Jan 23 12:56:52 agaffney: beats rebuilding all the stages I reckon Jan 23 12:56:58 Blackb|rd: pretty much, yes Jan 23 12:57:05 nixnut: you'd still have to rebuild the stages Jan 23 12:57:08 welp: they have to, still amd64 keeps falling behind ;) Jan 23 12:57:14 after updating the snapshot and pushing a new one out Jan 23 12:57:14 agaffney: why? Jan 23 12:57:16 nixnut: we have lots of caching in catalyst... Jan 23 12:57:34 phreak``: well, i'm the new security liaison, so hopefully things will improve :P Jan 23 12:57:40 wolf31o2|work: ah, ok. Not that much work as I thought then Jan 23 12:57:40 nixnut: subsequent runs use binpkgs from previous runs Jan 23 12:57:45 nod Jan 23 12:57:50 agaffney: that ain't much of a problem (rebuilding the stage) when you have the binpkg handy ;) Jan 23 12:57:53 nixnut: only the sec bumped package would need to be rebuilt Jan 23 12:57:59 k... so let's move on (this is taking longer than I hoped) Jan 23 12:58:09 ha. then move on :) Jan 23 12:59:17 automated regular internal release builds... this is basically extending andrew's stage-building script to include *everything* for a particular arch (though only necessary sub-arches for functionality)... Jan 23 12:59:55 the reason for us splitting off the new release as soon as the old is done is for these auto-builds... they'll check out the SVN for the new release and use the new dev profile... so we find problems and fix them, year-round... instead of just during the release cycle... Jan 23 13:00:15 wolf31o2|work: the problem with doing anything more than stages is that you need a per-arch template for livecd-stage[12].spec Jan 23 13:00:19 wolf31o2|work: we had thought about internally doing continous building for alpha (since we got shiny new disk space). I think it's a good idea. Jan 23 13:00:23 how "internal" do you want this builds to be? Jan 23 13:00:48 agaffney: that's fine... we're going to be pulling the specs from SVN and using them directly... so it really be the same as what we do for releases Jan 23 13:00:49 dertobi123: run on a dev box and not distributed to mirrors Jan 23 13:01:15 wolf31o2|work: Of course, this will help with 2008.1 earliest, not 2008.0. Jan 23 13:01:27 dertobi123: what andrew said... you could distribute them to people you wanted to test them and such... but they won't hit the main mirrors Jan 23 13:01:42 Blackb|rd: right... since 2008.0 starts $soon Jan 23 13:01:53 wolf31o2|work: ok, I think I'd like some clarification on the auto-build. Like do you want it refreshing every time someone updates the specs, once a day, once a week ... ? Jan 23 13:01:56 wolf31o2|work: and we'd need to get the process to run smoothly, too. Jan 23 13:01:56 :) Jan 23 13:02:01 ok, let me propose to manually distribute a selected subset of automated builds to /experimental "if necessary". i.e. new kernel with extended hardware support has been released and the latest "release" install media doesn't work for a number of users Jan 23 13:02:18 phreak``: we were running weekly automated builds from the live portage tree on poseidon for x86/amd64 Jan 23 13:02:26 agaffney: I know :) Jan 23 13:02:37 phreak``: well, we've been doing once a week... I think that works... we can adjust the schedule if we find that we need to do so... Jan 23 13:02:44 agaffney: keep in mind that a too tight schedule might pose a problem for the less-endowed archs Jan 23 13:02:46 wolf31o2|work: aight. Jan 23 13:02:57 the "best" way would be to re-pull the specs from CVS/SVN once a week and create a new snapshot from the live tree at that time Jan 23 13:03:00 dertobi123: that's totally up to the individual arch team... releng doesn't control /experimental Jan 23 13:03:05 similar to what we were already doing Jan 23 13:03:15 that was the plan Jan 23 13:03:35 I guess it gets a +1 from me, then :P Jan 23 13:03:38 wolf31o2|work: k ... Jan 23 13:03:43 same 'ere Jan 23 13:03:44 dertobi123: I mainly meant that they're not going to go to the mirrors automatically or anything Jan 23 13:03:59 agaffney: only if you add hardened foo to your buildlist ;) Jan 23 13:04:08 phreak``: use miranda :P Jan 23 13:04:11 wolf31o2|work: yea, I'll help you come up with some checklists and crap to give testers. Currently I'm the only active QA members really (maybe one or two others, so expect some slight latency in replies :) ) Jan 23 13:04:21 phreak``: I'll get with you on that... but yes, I was planning on including hardened, of course Jan 23 13:04:38 wolf31o2|work: no need to push them out automatically, but having a x86 install-media which works on "current" hardware somewhere would be nice ;) Jan 23 13:04:45 agaffney: *grmpf* fix my issues building x86 stages on a amd64 host :P Jan 23 13:04:56 Halcy0n: no problem... just join the gentoo-releng list and participate in the discussion (once we start it) Jan 23 13:04:59 phreak``: WFM Jan 23 13:05:10 agaffney: you ain't using gcc-3.4.6 :P Jan 23 13:05:21 phreak``: I'm not insane :P Jan 23 13:05:24 x86 and amd64 are moving very wuickly currently, so I think it's a good idea to have "quickfix" images in /exp Jan 23 13:05:31 agaffney: damn, just shutup :P Jan 23 13:05:48 Blackb|rd: yeah, that's what i meant to say :P Jan 23 13:06:10 dertobi123: I got bitten by that with our amd64 machines at work :/ Jan 23 13:06:19 Blackb|rd: dertobi123: again, that's up to the individual arch team... anybody can do an interim release now (in /experimental)... that won't change... Jan 23 13:06:25 Blackb|rd: use ting :P *cough* Jan 23 13:06:32 Blackb|rd: I'd certainly have no problem doing a quick minimal CD rebuild with a new kernel Jan 23 13:06:35 for x86 Jan 23 13:07:09 agaffney: So do I. Many user don't have the expertise. Jan 23 13:07:14 do we will have access to internal stages on dev boxes or is it closed to arch teams and releng? Jan 23 13:07:30 ColdWind: I'm sure something can be arranged Jan 23 13:07:33 ok... so my long-term goal for the automatic builds is for us to be able to use them as our initial beta... we pick one that's as close to release quality as possible, release it as beta, *then* start the release cycle (snapshot/qa/another beta/possible rc/release) Jan 23 13:07:49 there's no reason for us to "hide" them, especially if someone needs something Jan 23 13:07:51 ColdWind: it'll all be open to devs... it is now Jan 23 13:07:56 wolf31o2|work: Yep. Jan 23 13:08:05 wolf31o2|work: ok Jan 23 13:08:10 wolf31o2|work: sure, but this problem affects releng (and imho this problem is somewhat essential for us) Jan 23 13:08:23 dertobi123: huh? Jan 23 13:08:33 * agaffney senses a delayed reply Jan 23 13:08:46 A disturbance in the laf. Jan 23 13:08:48 lag* Jan 23 13:08:58 fail Jan 23 13:08:59 wolf31o2|work, although I would suggest most of our problems with releases are not so much the stages as media right? Jan 23 13:09:06 wolf31o2|work: wrt experimental stuff ... whatever Jan 23 13:09:08 least it is for my archs Jan 23 13:09:09 agaffney: I always ruin the good lines :( Jan 23 13:09:20 keeping it up the arch teams only didn't work in the past Jan 23 13:09:25 dertobi123: if you want to build a new CD for your arch right now... you can... the *only* things I ask is you don't name it a "release" name, like 2008.0... and it goes to /experimental... Jan 23 13:09:48 rangerpb: I'd agree. the most common problem is outdated kernel for new hardware Jan 23 13:09:54 wolf31o2|work: maybe that needs to be stated more explicitly Jan 23 13:10:00 that's of course usually only a problem on "main" arches like x86/amd64 Jan 23 13:10:17 at least, more of a problem on those arches Jan 23 13:10:19 dertobi123: I'm not sure I'm following... both the arch teams and releng can so interim stuff now... we just generally don't... Jan 23 13:10:25 wolf31o2|work: one thing wrt the openness of releng and the build process is that people know they can burn wtf they want Jan 23 13:10:46 wolf31o2|work: and that "we don't" is exactly the problem Jan 23 13:11:08 agaffney, I have some ideas around that if we get to a "feature" part of this meeting Jan 23 13:11:10 dertobi123: releng doesn't because most of us don't want to look at another stage tarball for a few months after a release cycle Jan 23 13:11:10 dertobi123: you're thinking of an explicit interim release? Jan 23 13:11:12 dertobi123: ok... this has nothing to do with the automated builds... can we hold it off until later Jan 23 13:11:30 dunno how some interim is stuff is called, but we need such interim stuff on a somewhat "official" base for x86/amd64 Jan 23 13:11:40 Blackb|rd: somewhat, yeah Jan 23 13:11:43 wolf31o2|work: k Jan 23 13:12:56 ok.. so on the subject of the regular builds... everybody seems in agreement? Jan 23 13:13:02 seems so Jan 23 13:13:06 * Blackb|rd is all for it. Jan 23 13:13:13 yup Jan 23 13:13:13 I'd like to run these on dev boxes, of course... Jan 23 13:13:21 cool... then to the next point... Jan 23 13:13:30 how do we involve the community more? Jan 23 13:13:37 betas will *definitely* help a lot Jan 23 13:13:43 wolf31o2|work: QA testing is one area Jan 23 13:13:53 wolf31o2|work: well, for starters, the beta cycle. (you already mentioned it) Jan 23 13:13:57 wolf31o2|work: I.e. public betas and explicitly asking for feedback Jan 23 13:14:14 I think the latter is more important than the former. Jan 23 13:14:31 Usually, you have to pull quality feedback out of people's noses. Jan 23 13:14:41 Blackb|rd: right... we're definitely going to spend more time on QA throughout the entire process Jan 23 13:14:46 wolf31o2|work: not having to go through an intermediary (like we did with roger55) will probably help Jan 23 13:15:05 sadly you almost want a beta that phones home as to what it was successfully run on Jan 23 13:15:13 agaffney: for release testers, agreed... and for the betas, we'll ask people to file bugs, like normal... Jan 23 13:15:18 users testing betas will want a list of thing they're expected to test Jan 23 13:15:24 Also, it might be nice to get feature requests/wishes earlier, for example directly after a release. Jan 23 13:15:25 ColdWind: we already discussed that Jan 23 13:15:39 Blackb|rd: indeed... that would help, for sure... Jan 23 13:16:03 When people think 2008.0 sucked for some reason, they will have forgotten until 2008.1 beta time. Jan 23 13:16:09 yep, I mean it should be clear what they should *test* not just the info they should report with bugs Jan 23 13:16:32 did i miss the meeting? Jan 23 13:16:38 ColdWind: right... it'll have both Jan 23 13:16:41 dsd_: still ongoing Jan 23 13:16:42 dsd_: nope Jan 23 13:16:43 wolf31o2|work: also, the hardware survey is a good idea. Especially for the more obscure arches. Jan 23 13:16:46 Blackb|rd: there was a thread on -releng after 2007.0 asking for features people wanted...hardly any response Jan 23 13:16:57 good ;) Jan 23 13:17:16 May I throw in app-admin/hwreport for hardware survey? Jan 23 13:17:17 agaffney: yeah. Maybe involving the forum peole might be a good idea. Have a poll citing ten things we might include. Jan 23 13:17:18 Blackb|rd: I find it more important for the least obscure arches... since there's only so many alpha boxes out there... but literally thousands of combinations of x86/amd64 stuff... Jan 23 13:17:22 dsd_: my very rough, still in-progress summary: http://dev.gentoo.org/~dberkholz/20080123_releng_summary.txt Jan 23 13:17:26 agaffney: most likely because people don't watch releng Jan 23 13:17:26 thanks Jan 23 13:17:34 (as in the mailinglist) Jan 23 13:17:40 Opfer: interesting... let's discuss that after the meeting... Jan 23 13:17:41 agaffney: if people just ahve to click to vote, they do so more freely than when they have to type up a mail Jan 23 13:17:58 wolf31o2|work: good point. Jan 23 13:18:17 drop a news item to the front page that you need feedback on what users need for next release Jan 23 13:18:41 rane: good idea Jan 23 13:18:50 dberkholz can help you with that Jan 23 13:18:55 :-) Jan 23 13:19:01 robbat2's been talking about getting his survey framework up for the user survey. use the same setup. Jan 23 13:19:08 rane: providing feedback should be as easy as possible without encouraging spamming Jan 23 13:19:19 dberkholz: I was about to suggest the same. Jan 23 13:19:21 sure... until that's up and running, we'll use the list Jan 23 13:19:25 Yea, I think the more we try to actually involve our users with the entire releng and development process will go a long way in improving developer<->user relationships. Jan 23 13:19:37 I'll just announce to -user and -dev-announce when I start the threads Jan 23 13:19:52 Halcy0n: indeed... *and* it'll improve our quality as a side-effect Jan 23 13:20:28 wolf31o2|work: Also, it might be a nice occasion to point out resources releng needs direly. Like arch teams without build machines. Jan 23 13:20:34 cool... I think we've got a few good ideas on this... let's move on Jan 23 13:20:41 Blackb|rd: that's on the agenda Jan 23 13:20:42