[08:36] <mlankhorst> feeling a bit sick, but good morning :)
[08:56] <mlankhorst> can someone re-upload libdrm? I fixed the arm build failure
[15:03] <seb128> mlankhorst, there?
[15:25] <mlankhorst> yeah
[15:25] <seb128> mlankhorst, hey
[15:26] <mlankhorst> heya
[15:26] <seb128> mlankhorst, https://bugs.launchpad.net/ubuntu/precise/+source/xorg-server/+bug/1021517 seems a regression in the recent precise SRU
[15:26] <ubottu> Launchpad bug 1021517 in xorg-server (Ubuntu Precise) "Xorg-server crashes reproducible with GIMP usage" [High,Confirmed]
[15:26] <seb128> which got pushed to -security as well
[15:26] <mlankhorst> oh fun
[15:26] <seb128> mlankhorst, can you or somebody else from here have a look soon?
[15:26] <seb128> mlankhorst, comments mention it depends of the trackpad status (like they get it when doing fn-f8)
[15:27] <seb128> see the most recent comments
[15:29] <mlankhorst> yeah seems likely that it was from 10.3 update, bit weird it would only show now though
[15:29] <seb128> mlankhorst, it doesn't only show now
[15:29] <seb128> mlankhorst, we started the bug was filed on 2012-07-06
[15:29] <seb128> ups
[15:29] <seb128> -we started
[15:30] <seb128> mlankhorst, I guess you guys didn't monitor the new bugs after the SRU?
[15:30] <jcristau> no apport gdb trace or anything?
[15:31] <mlankhorst> seb128: I guess those people didn't test -proposed, more likely
[15:32] <seb128> jcristau, only https://launchpadlibrarian.net/109601036/XorgLogOld.txt :-(
[15:32] <jcristau> yeah not quite as useful
[15:32] <mlankhorst> I'll see if I can make a new shiny backtrace
[15:33] <seb128> not sure why we don't get apport report for xorg, I guess the xserver handler block it
[15:33] <jcristau> maybe with those binaries addr2line would work, but i'm not going to download that..
[15:33] <mlankhorst> I'll create a new bt if I can reproduce
[15:46] <mlankhorst> works here :\ bug is useless without backtrace
[15:48] <mlankhorst> oh there we go
[15:53] <seb128> mlankhorst, you got it?
[15:53] <mlankhorst> seb128: hm nope, was a sigpipe which is valid
[15:54] <mlankhorst> might be old xf86-input-synaptics
[15:56] <seb128> mlankhorst, the bug got some pretty responsive users, you can perhaps give them hint to get the stacktrace?
[15:57] <jcristau> the bug already has a pointer to https://wiki.ubuntu.com/X/Backtracing
[15:58] <seb128> ok
[15:58] <seb128> mlankhorst, jcristau: I get the bug here, let me try that
[15:59] <mlankhorst> seb128: oh i just put some qs on the bug
[15:59] <seb128> easy to get for me "gsettings set org.gnome.settings-daemon.peripherals.touchpad touchpad-enabled false"
[15:59] <seb128> well I just gor an xorg segfault this way in a guest session
[16:01] <mlankhorst> ok
[16:01] <mlankhorst> riught that works
[16:05] <seb128_> hum, something the kernel didn't like
[16:05] <mlankhorst> seb128_: seems i can reproduce it, but I'm going to lay down now, not feeling so well
[16:05] <seb128_> mlankhorst, ok
[16:09] <seb128> hum, again
[16:09] <seb128> mlankhorst, sorry I might have missed what you said, but get better!
[16:10] <mlankhorst> seb128: its no problem to get the bt tomorrow, already had one
[16:11] <seb128> mlankhorst, should we hand over that to bryceh or RAOF while you get better?
[16:15] <mlankhorst> it's just something that got my throat quite good, should be over in a day I hope :)
[16:16] <seb128> mlankhorst, jcristau: backtrace on https://bugs.launchpad.net/gimp/+bug/1021517/comments/58
[16:16] <ubottu> Launchpad bug 1021517 in xorg-server (Ubuntu Precise) "Xorg-server crashes reproducible with GIMP usage" [High,Confirmed]
[16:18] <seb128> bryceh, ^ when you are up, it's a segfault and seems a regression from the recent SRU,security update to precise
[16:21] <mlankhorst> it was before the security update so definitely the dix changes in 10.4, maybe cnd would know from the backtrace
[16:21] <mlankhorst> if not with a bit more work anyone could find out what's going wrong :)
[16:23] <seb128> mlankhorst, right, it's likely 10.3
[16:23] <seb128> cnd, ^
[16:26] <cnd> I'll take a quick look
[16:27] <seb128> cnd, thanks
[16:35] <seb128> cnd, you can do that today? it's an xorg segfault regression easy to trigger which made it in a security update of the current lts, i.e not great to have, we should address it
[16:35] <bryceh> heya
[16:35] <seb128> bryceh, hey
[16:35] <seb128> bryceh, how are you? happy monday :p
[16:37] <bryceh> seb128, a bit invalid... threw out my back yesterday while wrestling my son into a pair of pants
[16:37] <seb128> bryceh, :-(
[16:38] <cnd> mlankhorst, seb128: interesting, it's obviously input related, but I have no idea
[16:38] <cnd> mlankhorst, it was in the latest update?
[16:38] <cnd> mlankhorst, when you said changes in 10.4, did you mean 11.4?
[16:38] <cnd> oh wait, nm
[16:38] <cnd> -0ubuntu10.4
[16:38] <cnd> mlankhorst, the attached X log says -0ubuntu10.3
[16:38] <cnd> I may be restating things we already know :0
[16:38] <cnd> mlankhorst, seb128: so my initial thoughts are that I haven't heard of such a bug before based on upstream activity
[16:38] <cnd> and we need to do some deep dive debugging to figure it out, unless we can find a specific commit or patch that affects things
[16:38] <cnd> if you can figure out *what* is going wrong, I can help figure out how to fix it
[16:39] <seb128> cnd, it's trivial to trigger for me, run gimp, create an image, turn touchpad off (that's my latitude e6410 laptop config)
[16:40] <seb128> cnd, I'm happy to test versions but report timing makes it pretty likely the issue is in ubuntu10.3
[16:40] <seb128> cnd, I turn off the pad with gsd by doing the gsettings command I put it comment #58
[16:41] <seb128> cnd, let me know if I can be useful in debugging
[16:41] <bryceh> seb128, have you tested 10.2 to verify that the bug isn't reproducible there?
[16:43] <bryceh> actually, come to think of it I was just yesterday using gimp on a machine that has 10.2 loaded and it did not crash
[16:43] <seb128> bryceh, yes, downgrading to xserver-xorg-core (2:1.11.4-0ubuntu10.2) fixes it
[16:46] <seb128> ok, upgrading to 0ubuntu10.3 and the bug is back
[16:46] <seb128> it segfault on first click in gimp after turning the touchpad off
[16:46] <seb128> so
[16:46] <seb128> bryceh, cnd, mlankhorst: we need an upload to fix that today or tomorrow, do you think debugging the issue and fixing is a reasonable way out or should we look at reverting the ubuntu10.3 changes?
[16:47] <seb128> or do you want to look at fixing it for a bit and after $time go to fallback plan to revert?
[16:51] <cnd> seb128, I'm sort of sprinting with the kernel team
[16:51] <cnd> I didn't bring a second laptop
[16:51] <cnd> so I can't today
[16:51] <cnd> and I'm also not really *supposed* to be helping debug X stuff, since I have other day job stuff
[16:51] <cnd> I know that mlankhorst is quite capable of that :)
[16:51] <cnd> if he can't get to it for some reason, then I can try to step in
[16:51] <cnd> seb128, I could probably reproduce it, but I can't debug it easily without a second machine at least
[16:51] <cnd> that's the blocker for me for today
[16:51] <cnd> but again, I need to focus on other work, and try to be a resource for determining a proper fix after the root cause has been identified
[16:51] <cnd> seb128, it's that critical of an issue?
[16:51] <cnd> seb128, we'd have to figure out what needs reverting first
[16:51] <cnd> the specific patch or commit would help
[16:52] <seb128> cnd, mlankhorst is unwell and went to bed
[16:52] <seb128> cnd, well, it's "xorg segfault when using gimp" which got introduced in a SRU,security update in the LTS, so it's an high importance issue yes
[16:53] <seb128> cnd, xorg closing might mean work and datas lost
[16:53] <seb128> cnd, but if you are busy no worry, I'm trying to estimate our easier way out
[16:53] <bryceh> seb128, the 10.3 change set introduces six patches which fix one bug, which is also a crasher.  Reverting seems feasible and maybe lowest risk solution.
[16:54] <mdeslaur> but wasn't 10.3 to fix a regression introduced in 10.2? :)
[16:54] <seb128> pick your least bad regression :p
[16:56] <bryceh> mdeslaur, hrmm, well the 10.2 fix was to make touchpad work on macbookair
[16:57] <mdeslaur> which was incomplete and causing xephyr to segfault
[16:57] <seb128> well, xephyr to segfault is better than xorg to segfault
[16:57] <mdeslaur> so, it's pick the least evil of regressions :)
[16:59] <bryceh> mdeslaur, right what I mean is that if we rolled that back too, then we would get into a state where nothing is crashing, just that macbooks have broken touchpads
[16:59] <mdeslaur> hrm, I agree, not ideal
[17:00] <cnd> bryceh, seb128: the bug fixed by 10.3 is a much lesser evil than an X server crash
[17:02] <bryceh> seb128, cnd if you'd like, I can take the action to revert the 10.3 and 10.2 changes
[17:02] <seb128> ok, at this point it's easier "we believe the easier way out is to fix the bug and we give it a few hours"
[17:02] <seb128> easier->either
[17:02] <seb128> or "let's revert and give us time then to figure what is wrong"
[17:02] <seb128> bryceh, what do you think?
[17:04] <sbeattie> bryceh: well, if we revert one or both, we should probably revert from the -security pocket as well.
[17:07] <bryceh> seb128, it looks to me like reverting the patches from 10.2 and 10.3 would be the way to go, until this is better understood.
[17:08] <seb128> bryceh, great, can you get that done today and coordinate with sbeattie,mdeslaur to get the rollback in -security as well?
[17:08] <bryceh> seb128, certainly
[17:08] <seb128> bryceh, then mlankhorst and cnd can probably investigate the segfault tomorrow
[17:09] <seb128> bryceh, thanks
[17:11] <mlankhorst> cnd: ^
[17:11] <mlankhorst> but yeah it's fine if it gets reverted for now
[17:11] <bryceh> sbeattie, I'm thinking just comment out the patches in question rather than doing a literal rollback to 10.1... sound right?
[17:12] <mlankhorst> sbeattie: the bug was filed before the security sru got released..
[17:12] <sbeattie> bryceh: yes, that's fine.
[17:12] <mlankhorst> so I don't thin kit had anything to do with it
[17:13] <mlankhorst> I'm not going to object to a rollback for now though :)
[17:14] <cnd> the bug would result in the touchpad doing wonky things if you have disable trackpad while typing and you use it a bit, and then it starts to act like extra touches are on the touchpad when more than one touch is used
[17:14] <cnd> IIRC
[17:14] <cnd> the bug can also occur just by doing suspend/resume cycles
[17:14] <cnd> but it's stil not a crasher
[17:14] <cnd> bryceh, I thought the 10.2 was ok?
[17:14] <cnd> so just reverting the 10.3 changes would be good enough
[17:14] <cnd> seb128, correct?
[17:14] <cnd> btw, I have really terribly high latency here
[17:14] <cnd> wifi is terrible
[17:15] <seb128> cnd, right, 10.2 is correct (but it had a xephyr segfault issue)
[17:15] <mlankhorst> seb128: not just xephyr
[17:15] <mlankhorst> xorg-server under some conditions as well
[17:16] <mlankhorst> it's mentioned in the changelog, but xephyr was the easiest way to trigger it (sorry really slow today)
[17:17] <bryceh> cnd, right; 10.3 attempted to fix the regression caused by 10.2.  10.2 itself was an attempt to fix a touchpad activation issue.
[17:18] <cnd> bryceh, I don't think I'll be much use today between the sprint, wifi instability, and lack of machines, so if you can take care of the revert please do
[17:18] <mlankhorst> so yeah rollback to 10.1
[17:18] <mlankhorst> + security fix
[17:18] <bryceh> cnd, yep I'll take care of the revert
[17:18] <mlankhorst> thanks bryceh :)
[17:19] <cnd> ok
[17:19] <bryceh> mlankhorst, 10.4 added 516-dix-dont-emulate-scroll-events-for-non-existing-axes.patch which I'm guessing we'd like to keep?
[17:20] <mlankhorst> bryceh: I don't know, it should be in -proposed atm, not sure though if it is
[17:20] <cnd> +1 :)
[17:21] <mlankhorst> if not just comment it out again
[17:21] <bryceh> ok
[17:22] <mlankhorst> bryceh: I updated some drivers in xorg today btw to x1.13rc1
[17:22] <mlankhorst> likely going to need to touch all of them
[17:22] <mlankhorst> since xaa was removed
[17:27] <bryceh> sbeattie, btw the security patch 509 probably should have been numbered 228.  The comments in debian/patches/series are confusing but generally 5xx we were using just for the input frankenserver bits
[17:28] <bryceh> I'll fix the comments so this is clearer
[17:28] <seb128> bryceh, btw do you know why apport doesn't trigger on xorg segfaults?
[17:28] <seb128> bryceh, is that a known issue,being worked?
[17:30] <bryceh> seb128, yes it's a known issue discussed at UDS.  I believe RAOF has the action item to work on it.
[17:30] <seb128> bryceh, ok, thanks, I will check with him, if we had that we would probably have noticed that one earlier through errors.ubuntu.com
[17:30] <bryceh> seb128, it's not that they're not being triggered, but that they trigger only in some cases
[17:31] <seb128> bryceh, well, that gimp segfault doesn't trigger it
[17:31] <seb128> it might turn out to be a good testcase ;-)
[17:31] <bryceh> the xserver has it's own crash handling code, which we had to bypass to make it work with apport, and it's always been a bit flaky
[17:31] <sbeattie> bryceh: ah, okay. kees was the one who handed it to me, so I'd assumed he'd gotten the numbering correct.
[17:32] <sbeattie> bryceh: anyway, let me know and I can shepherd things through the -security pocket.
[17:33] <bryceh> sbeattie, yeah no prob, it's our fault that the sections weren't labeled clearly
[18:08] <bryceh> sbeattie, ok, 1.11.4-0ubuntu10.6 uploaded
[18:22] <sbeattie> bryceh: thanks.
[19:33] <kees> bryceh, sbeattie: oh, whoops. sorry about that patch numbering. I missed that note :(
=== yofel_ is now known as yofel