From Fedora Project Wiki

Fedora Release Engineering Meeting :: Monday 2008-01-14

Meeting Higlights

DECISION: slip alpha freeze by one week, re-evaluate state next meeting; hold to planned GA date

IRC Transcript

-!- f13 changed the topic of #fedora-meeting to: Fedora Release Engineering Meeting13:00
f13ping jeremy jwb wwoods warren rdieter spot lmacken notting poelcat13:00
* spot is here13:01
rdieterhere (mostly)13:01
-!- f13 changed the topic of #fedora-meeting to: Fedora Release Engineering Meeting - Fedora 9 Alpha13:02
* lmacken 13:02
f13alright that's likely enough to get started.13:02
f13although I'd like wwoods  :/13:02
f13so, alpha, supposed to freeze tomorrow13:03
f13which would normally mean freezing tonight right after rawhide compose.13:03
f13so that rawhide tomorrow == freeze content13:03
f13however, our tree is in an interesting state13:04
f13wwoods: is it installable today?13:04
wwoodsAs far as I can tell.. no13:04
* poelcat here13:05
nottingis it uninstallable in a different way than it has been uninstallable previously?13:05
wwoodstraces back when it tries to start the transaction13:05
wwoodsnotting: yes!13:05
wwoodswe have X and disks now!13:05
wwoodssubject to the normal X caveats, e.g. "no X on sis"13:05
wwoodsI think dhcp is still screwed on x86_6413:06
wwoodsif you click on the "Upgrade existing system" item, the installer crashes13:06
wwoodsthat's what I've got so far13:07
wwoodsit's better than last week but a long way from installable13:07
spoti just pulled out two more fontsets13:08
spotbut that shouldn't affect anything13:08
wwoodsI might try yum upgrading an F8 box just to do *some* functional testing13:09
wwoodsbut it's been weeks since we had an installable rawhide13:09
wwoodsand that's a killer13:09
f13yes.  I'm not comfortable at all snapshotting for alpha at this point.13:09
f13so that would mean talking about a slip13:10
nottingdo we have a quick virtual whiteboard of issues?13:11
f13is in channel OK?13:12
wwoods has what I've got so far13:12
wwoodsI'll be adding the transaction crash and upgrade crash shortly13:12
jwbi just fixed nautilus-open-terminal13:12
jwbbut that shouldn't make any difference either13:13
wwoodsbut yeah, I've done *zero* functional testing13:13
nottingi think 422421 is fixed13:13
wwoods422061 is probably OK too13:13
wwoodsat least it should be non-NEW13:13
f13but more importantly, no new installs are possible13:13
f13so we don't know where there are problems.13:14
f13rather where /else/13:14
wwoodshow long's the freeze?13:14
wwoodsso. we want a week with an installable tree to shape it up13:15
wwoodsdo we push the freeze until we've got an installable tree, or freeze now and extend the freeze until we hit that target?13:15
f13I wanted chris to pop buy and help us make this decision.13:16
jwbhow likely will gdm allow logout in a week?13:16
f13clumens: we're discussing the alpha freeze tomorrow and the current state of things.13:17
* spot would be happy if gdm allowed reliable LOGIN13:17
f13to be perfectly honest, I'd be happy if you can install, and get updates13:17
wwoodsyeah, so long as people can get updates we can push fixes at them later13:18
wwoodsbut if they can't get the bits on their system we're boned13:18
poelcatdo we have an alpha tracking bug so folks can follow along at home?13:18
wwoodspoelcat: not yet. one should be created.13:19
jwbi have some kernel bugs to log myself.  kernel hates my laptop13:19
clumensf13: well we have hard drives today, and X starts on hardware that X runs on13:19
clumensthough apparently there's some rpmdb thing now that i have not gotten a chance to investigate13:19
f13right, and network x86_64 is still boned13:19
clumensstatic networking is probably busted too13:20
wwoodsoh right, forgot about that13:21
wwoodsdcantrell where art thou?13:22
f13wwoods: so what would you like to see before we take the snapshot?13:22
clumenswwoods: 1300-0600 = probably still asleep13:23
* jwb notes that with everything being broken he hasn't done any ppc spins13:23
wwoodsoh he's at UTC-6? good to know13:23
clumenswwoods: EST-613:23
wwoodsf13: installable on i386, at a bare minimum13:24
clumenshawaii and all13:24
wwoodsclumens: oh duh. noted. thanks.13:24
poelcatwwoods: all: tracker created
buggbotBug 428703: low, low, ---, Bill Nottingham, NEW , Fedora 9 Alpha Tracking bug13:24
clumensso if you're asking how i feel about anaconda right now, total mixed bag13:24
wwoodsoh for the love of crap. my kingdom for a working CDRW13:27
wwoodsI haven't seen a crash this colorful since the old DOS days. so that's kinda cool.13:27
wwoodsanyway, yeah, are we going to push the freeze back or extend it?13:27
f13I'd rather push it back13:28
jwbit's a rel-eng only freeze right?13:28
jwbthen why push it back?13:28
f13jwb: because it's not as if we're comfortable of have any idea whats going on outside of installer, so there isn't really a point to freeze that right now anyway13:28
poelcatand the whole point of the alpha that it installs13:29
jwbf13, i wouldn't say we have no idea13:29
jwbbut ok13:29
f13usb fob13:30
* wwoods confused by this directory13:30
wwoodshm. I should use mine.13:30
clumensso, you're saying i should get back to work?13:31
wwoodsyeah. slacker.13:31
wwoodsI'll get you some info on that transaction crash RSN13:31
wwoodsahh, I love the new URL entry screen (with proxy support and everything!)13:32
f13so how do we feel about a week slip, and re-evaluate next monday?13:32
wwoodsI feel OK about it13:32
jwbi feel it would be good13:32
nottingwe have nowhere to go but up?13:33
clumenswe came a long way this weekend13:33
jwbwell... we could go sideways... get anaconda working, and then everything after crashes?13:33
wwoodserror: no dbpath has been set (?!)13:35
wwoodserror: cannot open Packages database in /mnt/sysimage/%{_dbpath}13:35
wwoodssounds like that might be an easy thing to fix, though13:37
f13ok, so I'll announce the slip and we'll have a work to get things better.13:38
wwoodsa week to work?13:38
-!- f13 changed the topic of #fedora-meeting to: Fedora Release Engineering - Open Discussion13:39
jwbF9 names?13:40
clumensi wonder if rpm has changed13:40
jwblegal got back to us13:40
f13jwb: Yeah, can you get a vote setup with toshio?13:40
jwbyep, was just going to see if that's who i had to bug13:40
jwbbtw, most of the good names were yanked :(13:40
clumensso it's down to tasticles and popplers?13:40
jwbclumens, there's a much larger list than last time.  but most of the ones people said "yeah!" at were cut13:41
f13I liked bathysphere13:41
jwbi think that one's still there13:42
f13notting and I made some progress at making the rawhide creation process less tearful13:42
f13we'll have to enact it soon, which includes an upgrade of releng1 to rhel513:43
nottingalthough it's in a holding pattern for a bit13:43
nottingwell, skvidal will fix up createrepo. but more fixage requires func-ification13:43
f13notting: I'm not convinced that func is the answer13:43
f13notting: more realisticaly we need koji integration13:43
f13or else it's just a slightly better method of what we hav enow.13:44
nottingwell, there are two routes13:44
notting- a func module that controlls mock (you can set it up to create buildroots, etc.)13:44
notting- grand unified koji bits, so koji has a 'make me a rawhide' button13:44
nottingthe second seems overkill13:44
f13that's not what I'm talking about at all13:44
nottingthe first is nice in that you don't actually need to worry about what the koji repos are13:44
f13we just need a koji runroot function13:45
f13run "command" in chroot of <tag>13:45
f13upload results to /path/13:45
nottingright, but rawhide is already running on its own repos13:45
nottingnot the koji ones13:45
f13which is fixable.13:45
nottinghow is running on a potentially stale (vis-a-vis static-repo creation) repo better?13:45
f13because then we allow koji to better pick a builder to use, registers the weight of the run on the builder itself, and handles passing the bits around for us.13:46
notting.... passing the bits?13:47
f13notting: if you're that worried, you can trigger a newRepo task right before starting mash, and taskwait said newRepo for the buildinstall phase.13:47
f13notting: kojibuilder can upload the output via xmlrpc to the hub, which writes it out in /mnt/koji/ somewhere.13:47
jwbpoelcat, where is the rel-eng schedule at?13:47
nottingyeah, but 'output' is a pretty amorphous concept for pungify13:48
f13either way we're going to have to modify koji13:48
f13or we'll just keep ignoring the fact that rawhide tasks will abuse builders and not register as weight and we'll continue having problems of delayed rawhide due to builds.13:48
f13and builds getting delayed due to rawhide.13:49
nottingyes. the other thing is that a mock controller in func might be picked up quicker in general by others, but that shouldn't be a deciding factor13:49
nottingwhich led me wondering if koji would have been written differently if func predated it. which is neither here nor there.13:49
jwbpoelcat, ah thanks13:50
nottingf13: short-term, we can do 'check load ; koji disable-builder', etc. which would require some auth tricks (certificate or otherwise)13:50
f13notting: yeah, we could do that.13:51
nottingother note: as mentioned on fedora-admin, the current schedule for reversing the sync streams is february 4th13:53
nottingwhich will require some changes around rawhide & bodhi, although not necessarily *in* bodhi13:55
f13ok, we're off in the weeds.13:57
f13poelcat: decision: slip alpha freeze a week, re-evaluate state next meeting.13:57

Generated by 2.3 by Marius Gedminas - find it at!