18:03:50 <jforbes> #startmeeting Kernel Team
18:03:50 <zodbot> Meeting started Fri Jun 28 18:03:50 2013 UTC.  The chair is jforbes. Information about MeetBot at http://wiki.debian.org/MeetBot.
18:03:50 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
18:04:11 <jforbes> #topic who's here?
18:04:19 <davej> woop woop
18:05:04 <jforbes> Okay, we will start with the regular run through...
18:05:13 <jforbes> #topic F17
18:05:32 <jforbes> jwb: want to update us?
18:05:40 <jwb> F17 is at 3.8.13 right now.  As of about 30min ago, 3.9.8 should get pushed out to stable with the next push.  finally
18:05:50 * nirik is lurking around.
18:06:09 <jforbes> Wow, I think 3.9.2 was the first attempt to get 3.9 into F17.  It's been a while
18:06:09 <jwb> i'm sure this will result in a small amount of bug activity, but the release really does some little used at this point
18:06:14 <jwb> yeah
18:06:37 <jwb> with f19 being released next week, that leaves about 5 weeks left of life for F17.  it's effectively dead
18:06:48 <jwb> i don't foresee it being moved to 3.10
18:06:59 <jforbes> make sense.  Anything else on F17?
18:07:07 <jwb> nopes
18:07:09 <jforbes> #Topic F18
18:07:51 <jforbes> F18 has been moving a long. 3.9.8 was  built today and should hit updates testing. Based on other releases this month I expect it will be in stable by Monday
18:08:16 <jforbes> Both F18 and F19 should rebase to 3.10 shortly after release.
18:08:33 <jforbes> Anyone else have anything on F18?
18:08:52 * nirik assumes 3.10 is pretty close now.
18:08:52 <jwb> don't think so
18:08:58 <jforbes> #Topic F19
18:09:01 <jwb> nirik, yeah.  probably this weekend
18:09:12 <jforbes> nirik: either this weekend, or next weekend at the latest I would guess
18:09:20 <jforbes> davej: F19?
18:09:22 <nirik> cool.
18:09:42 * nirik is going to upgrade his main virthost/guest to 19 this weekend, if 3.10 is around will be happy to smoketest it.
18:09:53 <davej> nothing to add on 19 that wasn't already said, rebase to 3.10 once it's out.  bugs already starting to tick upwards for that release
18:10:04 <davej> suspect next week we'll see a lot more
18:10:23 <jforbes> Yeah, with it declared gold, I am sure many are already installing it
18:10:40 <jforbes> #Topic Rawhide
18:10:46 <jforbes> davej: anything on rawhide?
18:10:48 <davej> lots
18:11:09 <davej> the one thing that Linus seems concerned about in rc7 is the soft lockup thing I've been chasing the last two weeks
18:11:40 <davej> Dave Chinner seems to have nailed it down to a lock contention bug, and has a patch that seems to work in my testing so far, but Linus wants something less intrusive given how late we are in -rc
18:12:00 <davej> so he might do an rc8 depending on what that ends up looking like
18:12:01 * nirik is running rc7 on the laptop here, working fine.
18:12:12 <davej> or he might decide "only davej is reporting this so whatever. I'ma release 3.10"
18:12:45 <davej> nirik: if you run something that fills your caches, and then have something that syncs often, you'll hit it
18:12:56 <nirik> huh, ok.
18:13:17 <nirik> happy to help try and duplicate if it helps any...
18:13:19 <davej> it's pretty corner-case stuff but not really infeasible that users will hit it
18:13:37 <davej> nirik: Dave seems to have an easier reproducer than I do, so I think we're good
18:13:47 <nirik> cool.
18:13:51 <davej> my repro case sometimes takes up to 12 hours to hit the bug, which has been painful
18:13:58 <nirik> is there a bug for info?
18:14:02 <jwb> no
18:14:07 <davej> has got me thinking of a bunch of things I need to do to improve trinity though
18:14:30 <davej> yeah, a lot of the stuff I've hit this month and gotten fixed hasn't been in bz because I've hit it before users thankfully.
18:14:46 <nirik> no worries.
18:14:56 <jforbes> That's a good thing
18:15:39 <davej> There's still a bunch of other fun bugs I've walked into too that have no resolution
18:15:58 <davej> btrfs seems pretty fragile for me. Chris can't reproduce the bugs I'm seeing, so that's been fun.
18:16:10 <davej> I've got a patch from josef that I need to try for one of the issues
18:16:26 <davej> xfs has a recovery bug that still isnt' fixed
18:16:42 <davej> I think ext is ok right now.
18:17:08 <davej> aisdes from all that, some small driver bugs etc, little things like always.
18:17:33 * nirik s sound is still broken. Really need to try and find time to gather more debugging. ;(
18:17:53 <jforbes> Anything else on rawhide?
18:18:05 <davej> nirik: with those sorts of bugs you're better off going straight to the maintainer than through us
18:18:29 <nirik> yeah, I have/did... I'm not sure I understand what they want me to gather, but also haven't sat down to look hard at it either.
18:18:59 <jforbes> #Topic Test infrastructure
18:19:44 <jforbes> The test infrastructure is moving along now. The harness identifies builds, starts testing. If a test is already in progress, it will wait until it finishes before starting the next one
18:20:48 <jforbes> We do have an issue with fedmsg where it can drop off the bus completely and it doesn't know it. I have filed an issue requesting some sort of callback so we can monitor the connection, and reconnect as necessary
18:21:32 <jforbes> In the meantime, we are restarting the harness regularly in cron in hopes that it will stay on, and we run a catchup to test everything in the middle of the night
18:22:15 <nirik> cool. :)
18:22:17 <jforbes> I don't know how frequently it actually falls off the bus.  I have left the harness running for days and never had it fall off.  The one time it did fall off, I think the harness had been running for 24 hours
18:23:55 <jforbes> jwb dropped the libhugetlbfs test from the default because it was always failing on the guests, probably due to memory contention. I will look for a better solution soon, but for now that is enough that I can start turning on failure notification
18:25:24 <jforbes> The other biggy is we have no way to tell if a guest hangs booting a new kernel. Ping monitoring doesn't entirely work because guests are only supposed to be up for a little bit while testing, so I am writing something that will add  a guest to the monitor list when the harness kicks it off
18:26:11 <jforbes> Once the guest is not responsive, it will check the guests status in libvirt to see if it finished testing, or hung. If it finished testing, it gets removed from the monitor list. If it isn't finished, we get notified
18:27:00 <jforbes> It's not such a theoretical thing, rc6 actually hung consistently on 32bit guests, but the fix was found quickly
18:27:34 <jforbes> I think that's about it.  Any questions on the test infra?
18:28:13 <jforbes> #topic open floor
18:28:22 <jforbes> Anyone have anything else?
18:28:23 <davej> I think people might have more questions when we start blogging aobut how it works etc
18:28:36 <davej> and maybe after your talk at flock
18:28:48 <nirik> I have one item
18:29:12 <jforbes> Yeah, we are getting to the point that I can lay out the actual architecture pretty well, maybe before the next meeting.  It changed a lot as we got actual systems and our requirements became more defined
18:29:27 <jforbes> what's that nirik
18:29:28 <nirik> I added someone to the koji kernel perm... if you don't want me to add people or want some kind of formal process, please let me know. ;)
18:29:45 <nirik> otherwise I'll try and get an ack from you folks before making any changes there.
18:30:14 <jwb> yeah, that's fine
18:30:17 <davej> we should probably have some process there, at least to ACK/NACK additions.
18:30:33 <davej> who got added btw ?
18:30:35 <jforbes> that would probably make sense
18:30:37 <jwb> kyle
18:30:38 <nirik> kylem
18:30:42 <davej> oh, that guy
18:30:44 <davej> NACK
18:30:46 <nirik> :)
18:30:54 <jwb> nirik, just send an email to kernel-team@fedoraproject.org
18:31:14 <davej> I suspect it wouldn't be a regular occurance
18:31:14 <nirik> sure, can do that... note also that this perm is shim/grub2/etc too tho
18:31:26 <nirik> yeah, possibly not common enough to even care. ;)
18:31:28 <dgilmore> nirik: who did you add?
18:31:38 <nirik> see above ^ kylem...
18:32:02 <jforbes> okay, anything else?
18:32:07 <jforbes> Closing in 2 mins if not
18:32:08 <dgilmore> okay, we should have some propper proces to follow for adding folks
18:32:34 <jforbes> dgilmore: I think we just defined one.  email to kernel-team@fedoraproject.org
18:32:37 <nirik> yeah, how about a mail to all people in there now and discussion/time before adding?
18:34:55 <jforbes> That works
18:35:07 <jforbes> Okay, closing out, Thanks for coming
18:35:11 <jforbes> #endmeeting