18:00:02 #startmeeting 18:00:02 Meeting started Fri Jan 6 18:00:02 2012 UTC. The chair is jwb. Information about MeetBot at http://wiki.debian.org/MeetBot. 18:00:02 Useful Commands: #action #agreed #halp #info #idea #link #topic. 18:00:12 #meetingname Fedora Kernel 18:00:12 The meeting name has been set to 'fedora_kernel' 18:00:20 davej, ready? 18:00:43 #meetingtopic Fedora kernel 18:00:53 oops, yeah, gimme a sec. 18:00:57 k 18:01:40 ok 18:01:51 first up? 18:02:06 before we do the usual run-through,.. fudcon 18:02:15 #topic fudcon 18:02:59 we usually do a 'state of the kernel' talk which goes over well, so I suspect we'll do that again 18:03:09 do we have any additional talks we want to give ? 18:03:36 it'd be nice to get something on debugging/triaging kernel bugs, but i'm not sure exactly what we'd pitch 18:03:38 if I can pull my thoughts together, I might do one on some ideas I have for automating kernel qa 18:03:47 or that, yeah 18:04:16 i'm sure there will be general triage and qa meetings, so we might be able to piggy back on those a bit if we don't have enough for our own session 18:04:35 I had thought of a 'how to help us triage' session, but I can't imagine it would be particualrly thrilling to watch us looking through bugzilla. 18:04:46 right, same probably i had 18:04:51 er, s/probably/problem 18:05:21 ok, we'll just roll with it on the day. see how things go 18:05:22 so since this is a community meeting, if anyone in the community has suggestions or questions, chime in 18:05:36 perhaps we'll touch on it in the state-of.. talk 18:05:42 k 18:05:58 f15/f16? 18:06:02 yeah, lets move on 18:06:09 #topic Fedora 15/16 18:06:32 I wouldn't be surprised if we see 3.2.1 around the time of fudcon 18:06:44 so we'll probably have to start thinking about a rebase pretty soon 18:06:52 would be nice. 3.1.x was fairly stable until we hit that resume regression 18:06:53 not expecting it to be too painful 18:07:08 hopefully 3.2.x proves to be similarly stable 18:07:18 yeah 18:07:30 btw... that has the fixes that should help the "hangs on I/O to slow media" problem, right? 18:07:48 I think it's more band-aid type stuff for now, with the real fixes coming in 3.3 18:08:19 ok. i know we had one reporter say they are going to try 3.2. maybe that will be a good guinea pig type report 18:08:31 i don't think they'd get much worse :\ 18:09:08 I've been running self-built ones on 16 for a few weeks, it's held up pretty well 18:09:17 the other thing we've hit lately is the oopses in the f16 compat-wireless stuff 18:09:22 hopefully .1 will fix up the odd bits that were missed 18:10:28 unless 3.2.1 fixes those, we'll still see them in a rebase 18:10:40 I'm hoping that compat-wireless will start shaping up soon, it's been a bit of a horror show so far. 18:11:12 it might be worth keeping track of for eg, just how many wireless / suspend etc bugs we get each month. 18:11:32 in the whiteboard? 18:11:36 yeah 18:11:47 the problem we have with that is remembering to add it 18:12:40 oh, we could just do a bz search for /builddir/compat-wireless ;) 18:12:48 heh, yeah 18:13:09 maybe something could be coded up with python-bugzilla to make search and whiteboard additions easier 18:13:29 feel free to add it to the project list 18:13:38 yep 18:13:47 so something that's been bugging me for a while.. 18:13:51 those sch_generic bugs 18:14:35 I think once we have 3.2 out, if we still keep getting those, I'm going to bug upstream about them again, because they don't seem to be anything we can do anything about 18:14:48 yes. i know the atl1c driver seemed to get some general improvements when one user tested mjg59's driver-specific ASPM patch 18:15:01 but those patches aren't in 3.2. i think more rework is needed 18:15:20 the main ASPM change was backported to f16 though 18:15:32 of course, i'm kinda guessing that is related. 18:15:52 yeah, I think it's a bigger problem tbh 18:16:00 possibly even hardware related in some cases 18:16:38 atl1c seems to be the most frequent by far. the past couple of e1000e that i've seen seem to be in virt guests 18:16:55 beyond that, i haven't noticed any trends 18:17:59 yeah, that's about it really. ~750 f15/f16 bugs right now. Hopefully we knock another bunch out with the rebase. 18:18:17 #info Revisit sched_generic bugs with upstream if 3.2 still shows frequent occurances 18:19:03 #info f15/f16 rebase to 3.2.x around FUDCon timeframe 18:19:47 so one thing i don't think the rebase will help is the hibernate things... 18:20:02 where stuff seems randomly broken on resume. like unlock_new_inode 18:20:07 that bug is a nightmare 18:21:00 we're still no closer to really understanding that one. until one of us can reproduce it so we can bisect it, I'm not optimistic. 18:21:24 i'll try harder. i can't remember the last time i even wanted to hibernate 18:21:29 * jwb is non-mobile 18:22:33 move on? 18:22:35 yep 18:22:44 you had one more thing for f15/f16 didn't you? 18:22:51 oh yeah 18:22:57 I think it's a no-brainer though 18:23:19 things like UAS which aren't ready for mass consumption, but we want to give them a try in rawhide.. I thought about adding them to 'make release/debug' 18:23:45 yeah, nice idea 18:24:02 some kind of comment or separate section for them so they're noticable? 18:24:04 that's the only one that came to mind initially, but there might be more 18:24:12 yeah, needs a comment too 18:24:30 a lot of the decisions we've made in the config options really could use commenting tbh 18:24:42 I forget myself sometimes why we set something a certain way. 18:24:57 yeah. i'm afraid to go turn stuff off or change it because i have no idea why any of them are set the way they are 18:25:15 i cleaned up powerpc recently, but stopped there 18:26:06 #agreed Put various config options we want to try in rawhide but aren't ready for a stable release in 'make release/debug' 18:26:31 so, rawhide ? 18:26:37 sure 18:26:43 #topic f17 (rawhide) 18:26:57 3.2 seemed to drag out a little 18:27:08 but I feel optimistic we can still squeeze 3.3 into 17 18:27:32 hopefully. should we hold off on a rebase until we see how -rc1/rc2 look? 18:27:48 we'll be in -rc by alpha 18:28:05 assuming the schedule on the wiki is current 18:29:04 right 18:29:52 too bad it won't be released before we branch. having to do bodhi updates for 3.3 -rcs will be a bit slower 18:30:17 ugh, yeah, I dislike that process 18:31:36 so I had an item for modules-extra discussion, but I think we really need to see how that is accepted post alpha to really know whether it's a good idea or not 18:31:52 i'm still very undecided about it myself 18:32:09 * nirik has a quick question. 18:32:18 shoot 18:32:37 whats the status/plans for debug options? always on in rawhide, on in branched until X? or was it going to get split into a kernel-debug ? 18:32:55 ah, right. 18:33:04 I think the only thing we really decided so far was to do one non-debug build per -rc 18:33:26 might be good to document somewhere... 18:33:31 i think we should stick with that for a kernel cycle and see how much benefit it has 18:33:44 before we go implementing kernel-debug/kernel-nodebug 18:34:03 so you can find the non debug one via changelog? 18:34:07 yep 18:34:15 nirik, yep. i'll look at writing it up on the kernel wiki page too 18:34:22 cool. 18:34:31 as of a few days ago, it's going to be the same commit msg every time too 18:34:35 so that same plan will hold for f17 once it's branched? 18:34:44 or undecided yet? 18:34:48 #info rawhide will have kernels built with debug options disabled once per -rc 18:34:59 should be 18:36:32 this sort of ties into some of the ideas I have for automated benchmarking of rawhide kernels too. (which is one of the things I might talk about at fudcon) 18:37:13 * nirik updates his rawhide vm to the non debug build. ;) 18:37:59 any other questions? 18:38:48 davej, so i saw josh stone will be at fudcon. maybe we can corner him on utrace/uprobes... 18:38:52 ;) 18:38:59 maybe he plans on cornering us about same :) 18:39:13 a double trap... 18:39:18 we'll just have to decide who plays good/bad cop beforehand. 18:39:54 from the fedora side, that has gone awfully quiet. 18:40:30 yeah. the patch for 3.2 was identical to 3.1 except some line number changes in the patch itself 18:40:41 minor enough that the 3.1 patch still applied 18:41:58 anything else? 18:42:01 I'm all out of things I had. 18:42:18 * nirik has nothing more 18:42:40 i'm sure knurd_afk will have some questions for us on the list ;) 18:43:40 ok, if there is nothing else, let's wrap up 18:43:43 word 18:43:51 same bat-channel 2 weeks time. 18:44:15 #info Remember to post questions and suggestions to the fedora kernel list 18:44:19 #endmeeting