18:00:13 #startmeeting 18:00:13 Meeting started Fri Mar 30 18:00:13 2012 UTC. The chair is jwb. Information about MeetBot at http://wiki.debian.org/MeetBot. 18:00:13 Useful Commands: #action #agreed #halp #info #idea #link #topic. 18:00:28 #meetingname Fedora Kernel meeting 18:00:28 The meeting name has been set to 'fedora_kernel_meeting' 18:00:43 #meetingtopic Fedora Kernel meeting 18:00:54 #addchair jforbes davej 18:01:08 #topic init 18:01:16 everyone ready? 18:01:20 yup 18:01:22 yep 18:01:36 who's the token community person today? nirik ? :) 18:01:47 * nirik waves 18:02:05 ok, so should probably start with the release recaps 18:02:14 #topic F15 18:02:37 davej, you're on f15 right? 18:02:38 15 is kinda stagnating until 3.3.1 lands. 18:02:49 I've got the rebase ready, just waiting for Greg to release it 18:03:06 had hoped that would have been this week, but it's looking more like next week 18:03:44 * brunowolff I mostly watching the blocker meeting, but will try to follow here as well. 18:03:56 last kernel pushed was 2.6.42.13 (3.2.13), right? 18:04:07 yeah 18:04:21 if something else comes out between now and 3.3.1, I'll push it if it's important 18:04:27 it's the only reason I've not committed the rebase yet 18:04:44 but it would have to be something pretty critical 18:04:49 #info F15 currently at 2.6.42.13 (3.2.13). Waiting for 3.3.1 18:05:06 ok. seems pretty quiet otherwise 18:05:16 considering how long it takes to push an F15 update to stable 18:05:36 #topic F16/F17 18:05:38 F15 users could really help that out by testing and giving karma as appropriate 18:05:46 jforbes, true 18:06:03 #info testers/karma needed for F15 as it stagnates in updates-testing 18:06:22 ok, so F16/F17 are both on 3.3 at the moment 18:06:27 and should be almost identical 18:06:54 we rebased F16 early this time, and it closed out quite a few bugs 18:07:08 of course, more got opened, but it seems to have been a net benefit for us 18:07:46 we'll pick up 3.3.1 as soon as it comes out 18:08:15 since last week, 159 new f16 bugs opened. 76 closed. 18:08:25 (including dupes) 18:08:40 looks like still quite a few dupes open there too. 18:08:55 yeah, probably. i know some of the new ones aren't on 3.3 either. still from 3.2 18:09:08 so two things we fixed this week of note 18:09:13 1) ASPM regression 18:09:48 3.3 has a commit that causes some machines to either not boot at all due to ASPM changes, or it will do weird things like cause jmicro sata controllers to basically break 18:10:02 I just ran this weeks bug stats for those curious.. http://codemonkey.org.uk/2012/03/30/weekly-fedora-kernel-bug-statistics-march-30-2012/ 18:10:04 mjg59 figured that one out and we included the patch. should be in updates-testing right now 18:10:21 #info Weekly bug stats: http://codemonkey.org.uk/2012/03/30/weekly-fedora-kernel-bug-statistics-march-30-2012/ 18:10:38 bug 2) i915 hibernate fix 18:10:50 i really hesitate to call it THE fix, but it's certainly at least a fix 18:10:56 I'm pretty confident in that fix 18:11:02 * jsmith cheers 18:11:04 I don't think airlied needs to worry about buying beer for a while 18:11:13 davej, yeah, definitely 18:11:24 it just isn't going to fix everyones issues, that's all 18:11:36 yeah, still a lot of unrelated hibernate problems to fix 18:11:49 but that should close so many 'weird shit happened' bugs over the next few weeks 18:11:53 somewhat disappointingly, it seems the unlock_new_inode issue still hits for a couple of people after including that 18:12:04 i need to go look and see if they were using i915 though 18:12:44 anyway, that patch went into f16/f17 today. should get pushed to an update really soon now 18:12:59 anything else on f16/f17 18:13:36 #info i915 and ASPM fixes queued up in f16/f17 18:13:49 is there anything we need to discuss for F17 beta before we move on to rawhide? 18:14:09 seems the Beta kernel is going to be 3.3.0-1.fc17 18:14:21 Dan Walsh mentioned something yesterday about needing something in for beta, but didn't give me any details. 18:14:26 which is kinda old, but adamw might kill me if i ask for a newer one 18:14:32 jwb: I told him to bug you 18:14:36 fair 18:14:52 but yeah, unless it's something critical, sounds like post-beta 18:14:56 going to drive adamw crazy (or me) if it's really critical 18:14:59 jwb: yes. yes, you would be dead. 18:15:17 #info F17 Beta kernel is 3.3.0-1.fc17 18:15:32 at this point the only things we care about for beta kernel is "anaconda doesn't work unless we have this patch" 18:15:39 yeah 18:15:44 and even then, it would have to affect a significant number of people 18:15:56 i wonder if 3.3.0-1 has the broken irqpoll patch 18:16:12 I think it did 18:16:49 17 also still has a bunch of annoying things, like that timekeeping printk/spew that we should silence for final 18:16:55 it does 18:17:01 getting tired of seeing that bug reported 18:17:27 davej, did you write up a patch for that? 18:17:46 not yet. I can do it though. 18:18:03 #action davej/jwb to write patch to shut up timekeeping 11% skew message 18:18:35 at least the irqpoll thing is just a "update your kernel" kind of issue 18:18:39 ok, move on? 18:18:51 #topic Rawhide 18:19:14 jforbes, you wanna cover this one? 18:19:31 So rawhide is rebasing daily through the merge window, we are building now, and yesterdays rebase is in the repo 18:19:59 There are a couple of interesting pieces, PTP support being one 18:20:30 #info rawhide currently at 3.4-rc0 (post 3.3 git snapshots) 18:20:46 I expect Sunday to be the end of the merge window, so we will know exactly what we should end up with featurewise in 3.4 18:20:47 jforbes, there's a legal block on the ptpd-phc userspace package now 18:21:05 jwb: ugh, license problem ? 18:21:23 jwb: fair enough, but at least now people wishing to build the userspace themselves won't have to also build a new kernel 18:21:31 davej, someone pointed out a couple of previous reviews of different forks and something about IETF 18:21:35 spot is looking at it 18:21:43 I blame spot. 18:21:44 jforbes, true 18:22:08 In other good news, our patch count is going down quite a bit, bug fixes are making the merge window 18:22:17 nice 18:22:52 anything else on rawhide? 18:22:54 our NFS patches, a few random bug fixes and virtio-scsi have all been dropped as patches because they are upstream 18:23:20 nothing else yet. I will likely build today, and hopefully monday or tuesday we will have the rc1 build 18:23:51 oh, I have something 18:24:02 thoughts on increasing the size of the dmesg buffer again ? 18:24:14 with systemd spewing stuff to it, the early boot stuff is always cut off. 18:24:18 for rawhide only? 18:24:31 at the least, we should do it for debug/rawhide builds 18:24:46 yeah, i was going to say it would probably be good in debug builds too 18:24:59 I don't see a reason not too, especially for debug builds 18:25:16 #agreed increase the size of the dmesg buffer for rawhide/debug builds 18:25:18 Even for release builds, it's not a bad idea 18:25:34 I note we're currently at 128K, this moves it to 256K. 18:25:41 it's not like it's megabytes of ram here 18:26:17 so move LOG_BUF_SHIFT=18 then 18:26:29 I should bug util-linux to change the default -s for dmesg too. (It's something silly like 16K) 18:26:31 Right, and it is a very worthwhile use of another 128k, though I expect some of the arm folks might prefer something smaller 18:26:53 they can override it in config-arm-generic 18:26:55 yeah, arch overrides would be understandable for lowmem platforms 18:27:06 and given arm doesn't even use config-generic right now.. 18:27:16 it sort of does for some platforms 18:27:21 it's just there are too many of them 18:27:23 Hmph 18:27:25 oh, I missed that 18:27:38 they have config-arm-generic, config-arm-omap-generic 18:27:47 i think those actually are smaller overrides 18:27:59 there are a few that are just full configs though 18:28:02 right, but those don't inherit from config-generic do they ? 18:28:08 i think they do 18:28:11 oh, great 18:29:32 #action set CONFIG_LOG_BUF_SHIFT=18 in config-generic. other arches can override in arch specific configs 18:29:59 anything else on rawhide? 18:30:10 nope, that should make today's buidl 18:30:24 #topic kernel autotest 18:30:56 we had a discussion brainstorming what we want to do for performance and regression testing of the kernel 18:31:09 this is one of the goals the kernel team is trying to focus on this year 18:31:50 the summary is that we're going to use the autotest framework and start with small 'does it boot' tests 18:31:56 jforbes, you have that pretty much done, right? 18:32:02 or it's trivial to do anyway 18:32:21 Yes, that part is done from the local side, I suppose I should post the control script somewhere 18:32:38 #info kernel performance/regression testing one of the goals the Fedora kernel team is working on this year 18:33:03 yeah, would be good to get out there 18:33:13 We actually assume in the control script that you already have a VM ready for testing, though we can post the instructions for it to do an auto install too 18:33:38 might be good to do a small write up to the mailing list? 18:34:28 Yup, actually probably best to do a wiki page, and follow it up with a write up to the mailing list that points to it 18:34:36 yeah, perfect 18:34:41 hang it off the main Kernel page 18:35:13 so we could also ask the community if they know of small, self-contained tests we could run as well, right? 18:35:26 So the next thing we want to look at is writing more comprehensive regression tests, or getting some contributed from others 18:35:31 heh 18:36:17 ideally I think we have a 'make test' target, or a test launching script in a kernel-tests package. Autotest can call it, and people who don't want to mess with autotest can run it locally on their machines 18:36:22 #help If community members want to contribute regression tests, please email the kernel list with suggestions/tests. 18:36:46 not that single machine autotest is very difficult 18:37:21 right, but it lowers the barrier if we don't require it 18:37:35 at least for local runs 18:37:50 sure, and fairly unnecessary if people are booting into the kernel anyway 18:39:03 use vms? 18:39:21 vms == Virtual Machines ? 18:39:52 autotest does with the current control script 18:39:56 jwb: yes 18:40:03 yeah, what jforbes said :) 18:40:19 ok ;) 18:40:23 though it can be made to control a test host that isn't a vm it is just a more elaborate setup 18:41:47 anything else on this? 18:42:12 No, though I would love to hear some community input 18:42:26 I will specifically ask for it in the mailing list post 18:42:30 cool 18:42:54 oh, it's worth pointing out we're going to start with regression tests. performance testing will come later, and probably on a specific set of machines 18:44:22 ok, that pretty much covers the (one line) agenda we had 18:44:32 #topic Open Floor 18:44:40 questions, other topics? 18:45:08 jwb: I should probably file a bug instead of complaining in every meeting / on irc but ... also didn't shut up yet ;) 18:45:13 *alsa 18:45:24 yeah, i emailed about that last time. i got no reply 18:45:33 i can poke them again 18:45:37 I did the same in 2008 ;) 18:45:46 drago01, or file a bug in the alsa project might be best to be honest 18:45:59 jwb: ok 18:45:59 they seem somewhat responsive on their bug tracker thing 18:47:34 anything else? 18:48:08 quiet week. 18:48:25 got a number of bugs fixed though, so not a bad week 18:48:29 even if bugzilla says otherwise 18:48:56 ok, going to close out the meeting in 30 sec 18:49:36 #endmeeting