[00:20] RAOF, on that Xubuntu list there seem to be two chipsets that have bad failures, GeForce 6150SE nForce 430 and RV370/M22. Do we have known bugs for that? [00:30] robert_ancell: I haven't found any yet. [00:56] Off for now! [00:56] ill come back later and iterate on connect-display-request-merge and client-focus-notifications if people have time to review [02:36] Grr. I hate this code. [02:38] ? [02:45] Oh, splitting up the root window. [02:46] I'll just get it to work now, and then refactor out all the ugly. [02:46] Or at least as most ugly as possible. [02:51] Oh, right, multimonitor? === chihchun_afk is now known as chihchun [03:01] Yeah. === chihchun is now known as chihchun_afk === chihchun_afk is now known as chihchun [04:06] duflu: Oh, I seem to remember you having thoughts on a hw cursor api? [04:08] RAOF: Only minor thoughts... [04:08] Also someone started already... [04:09] Oh, yeah. [04:09] That's right. [04:10] RAOF: Bug 1189775, bug 1206780 [04:10] bug 1189775 in Mir "Mir cursor has no hotspot setting, assumes (0, 0)" [Medium,In progress] https://launchpad.net/bugs/1189775 [04:10] bug 1206780 in Mir "Clients cannot change the hardware cursor" [Low,Triaged] https://launchpad.net/bugs/1206780 [04:12] HW cursor is a prerequisite of useful GLX bypass, so I may need to pick that up. [05:15] good morning === forestpiskie is now known as Guest28422 [05:22] tvoss|eod: Good morning === tvoss|eod is now known as tvoss [05:23] RAOF, hey there :) how is it going? [05:23] This code is ugly, and I hate it. [05:23] RAOF, the multi-monitor stuff? [05:23] Yeah. [05:24] But I'll get it to work, and then unuglify it. [05:24] To the extent that's possible. [05:26] RAOF: FYI, a branch adding some of the requested extra display configuration information has landed [05:26] Woot! [05:26] alf__: Thanks. [05:26] RAOF: basically Output::type, Card::max_simultaneous_outputs, Output::preferred_mode [05:27] Nifty. That'll mostly complete the read-end of randr. [05:30] duflu: it seems you were right, since we remove this unity_support_test call from opengl, the ATI hang isn't showing (yet) [05:31] it's still a workaround and not a fix, but good enough for me :) [05:34] didrocks: The guess is well-educated. Mixing multiple GL contexts with fork/exec is apparently a bad idea, sometimes [05:35] ... which means don't run a GL app (unity_support_test) from within a GL app (compiz) [05:35] duflu: I still wonder how/why u-s-c exacerbate this behavior, but at this stage, this is good enough :) [05:35] duflu: but when compiz is starting another app [05:36] from the launcher [05:36] it fork/exec another potential GL app [05:36] Wasn't it hybris that exacerbated this behaviour? [05:36] didrocks: I'm more curious why a few people hit the problem a year ago, but the rest of us did not [05:36] (but I guess you meant during opengl initializion) [05:36] RAOF: no :/ [05:36] RAOF: we got more faulty ran this week-end, after we removed hybris [05:36] duflu: right, that's really weird [05:37] didrocks: One explanation is that very few people actually attempt to use hardware that would fail unity_support_test [05:37] Hardware that's not a VM... [05:39] possibly, but we see that ATI can fail in that case, but it never failed with plan xorg, just with xmir [05:39] so it's not only hardware-related [05:39] there is a chain of event [05:48] duflu: argh, I talked too quick [05:48] RAOF: duflu: ATI machine stuck now :/ [05:48] Mirv: FYI ^ [05:49] * didrocks checks the compiz version [05:50] Grargh! [05:50] 1:0.9.9~daily13.04.18.1~13.04-0ubuntu4 [05:50] ok, we have the one with the fixed version :/ [05:51] RAOF: do you think anything you can do or I should reboot the machine? [05:51] It might be worth attaching gdb again and confirming that it's blocked in the same place? [05:51] RAOF: want access? [05:52] Yeah, why not. [05:52] I've realised that I can't actually get the setting part of multi-monitor to work until I can bind a surface to an output, so let's play SSH! [05:52] ;) === Guest28422 is now known as forestpiskie [07:06] good morning [07:12] racarr: any luck with my nexus4 issue? [07:15] yay [07:15] at least now it's crashing :D [07:19] tsdgeos: Sounds like your expectations have lowered :/ [07:24] duflu: well it was doing nothing, it now crashes [07:24] tbh that is a small improvement :D === chihchun is now known as chihchun_afk [08:02] duflu: hi! It seems the compiz unity-support-test running outside of compiz did not help in the u-s-c crashes on ati, sadly [08:02] sil2100: Then I am out of ideas [08:03] sil2100: stacktrace ? [08:09] smspillaz: not sure if we have any, the machine is usually in really bad condition when that happens === chihchun_afk is now known as chihchun [08:50] tsdgeos: Confirmed, after upgrading to the latest phablet images, N4 hangs instead of rendering. Black screen :/ [08:50] \o/ [08:50] it's not only me being stupid [08:51] Not at all [08:54] tsdgeos: Do you have a bug logged in mir yet? [08:54] i think not [08:54] let me check [08:54] duflu: nope, i was waiting for a second person to confirm [08:55] can do now if you want [08:55] tsdgeos: Yes please. And don't be afraid to log bugs early and confirm later [09:16] * duflu shuts down to swap video cards [09:22] duflu: https://bugs.launchpad.net/mir/+bug/1211694 [09:22] Launchpad bug 1211694 in Mir "Black screen on Nexus4" [Undecided,New] [09:22] tsdgeos: Thanks [09:22] duflu: sorry i did not remember your name [09:22] so i wrote your ircnick [09:23] tsdgeos: /whois duflu [09:23] * tsdgeos is not as good with nicks <-> names <-> faces has he'd like [09:23] duflu: i know but you were not here when i was filling the bug :D [09:23] tsdgeos: Sorry, I am playing with video cards. I won't have much attention on IRC for the rest of the day [09:24] duflu: no worries :-) [09:27] smspillaz: Not (obviously) compiz's fault - X is hung in DRI2Authenticate, waiting on xmir_drm_auth_magic to complete. [09:28] RAOF: Any chance we will/can accept the community-provided patch for https://bugs.launchpad.net/mir/+bug/1195425 ? [09:28] Launchpad bug 1195425 in xserver-xorg-video-ati (Ubuntu) "Corrupted screen using radeon drivers" [Critical,Triaged] [09:28] (comment #12) [09:29] Yes, I can apply that. [09:30] RAOF: For the record, I have nothing to do with it. Just noticed the bug and community activity therein [09:31] Yeah, I was aware of the problem and the patch (which was also a reply to the upstream patch submission bit); I just didn't think we hit it. [09:31] Thanks for pinging me with it. [09:32] RAOF: I'm playing with a Cedar card and have serious corruption in render_surfaces. But haven't tried the patch [09:32] The patch is only for XMir; render_surfaces is going to be a different problem. [09:33] RAOF: You mean https://launchpadlibrarian.net/144789844/mir_fix_tiling.diff ? [09:33] Yes [09:33] Hmm [09:34] Possibly in the mesa patch ;) [09:38] RAOF: Ok, just noticed the render_surfaces bug is a recent regression. Likely caused by the switch branch and the bug is limited only to render_surfaces with its wacky buffer initialization [10:01] * duflu changes graphics cards again [10:04] didrocks: ppa:raof/aubergine contains an annotated mir/xmir stack that'll dump info about the drm auth process. [10:04] Woo, 3 minutes. New record? [10:04] That's nice and fast. [10:05] Can't do that with Windows [10:05] RAOF: excellent! I think next run, when the machine will be freed, you won't be around [10:05] * RAOF uses his bios to switch graphics cards, so it's approximately 1 standard boot time to change. [10:05] didrocks: That is highly likely :) [10:05] RAOF: do you prefer to connect live when we try that? we can do that tomorrow in that case [10:05] didrocks: The interesting stuff will be in /var/log/Xorg.0.log, and /var/log/lightdm/* [10:06] I'm happy for this to just get run and have the logs dumped on me. [10:06] RAOF: ok, let's try that :) [10:06] RAOF: one run passed successfully meanwhile FYI [10:06] so it's not a 100% hit unfortunately :/ [10:06] Yay. Race condition! [10:07] RAOF: oh, I'm afraid we'll get newer Mir though throughout the day [10:07] RAOF: can you try to push mir with a higher version? [10:07] Allow me to upload to my ppa again, this time with an epoch? [10:07] RAOF: sounds good :) [10:07] already? haha [10:07] I would never thought I would say epoch is good :p [10:07] think* [10:30] RAOF: I'm a little bit afraid you uploaded u-s-c to early without an artificial bump on debian/control [10:30] RAOF: so it's getting published before Mir [10:30] and so package dependency won't work === dpm_ is now known as dpm [11:11] didrocks: Oh, arse. Allow me to fix that. [11:12] lol [11:12] one more epoch later! [11:17] Hm. The 3.11 kernel looks like it might have fixed the hsw texture corruption on suspend/resume. === chihchun is now known as chihchun_afk [12:43] sil2100: RAOF: I think I just press the "I'm feeling lucky" button and just get the ati card screwed starting at the first run! [12:43] * didrocks now collect logs [12:55] RAOF: https://bugs.launchpad.net/mir/+bug/1204939/comments/11 [12:55] Launchpad bug 1204939 in Mir "Unity doesn't start on ATI test machine (hang in mir_wait_for())" [Critical,Triaged] [12:57] sil2100: ok, now that those infos are extracted, I'm rebooting the ati machine FYI ^ === pete-woods is now known as pete-woods-lunch [14:17] ;) [14:17] * sil2100 checks the logs [14:23] didrocks: I guess Daniel might take a look at those logs as well, maybe he would have some ideas as well - if he has enough time to look at them [14:24] yeah === alan_g is now known as alan_g|tea [14:31] didrocks: I think we need to make sure everyone knows what happened [14:31] didrocks: did you poke Pat and Kevin? [14:31] sil2100: oliver is doing the communication work :) [14:32] Awesome === alan_g|tea is now known as alan_g [15:01] didrocks: Apps finished, all green :O [15:02] sil2100: wooow \o/ green again [15:02] :O <- shocked face [15:02] sil2100: running more make it less a burden to fix issues [15:02] didrocks: so it seems, as less changes are in during releases, so even less possible stacks to fail at once ;p [15:03] right :) [16:30] is ickle around? [16:30] Chris Wilson === bodya_leone is now known as bodyaFCSD [20:55] RAOF, ping [21:04] thomi, ping [21:04] olli_: Hi [21:04] thomi, hey [21:05] thomi, mind having a look at http://bazaar.launchpad.net/~ubuntu-testcase/ubuntu-manual-tests/trunk/view/head:/testcases/packages/1572_xMir to see if this is in good shape for a round of call for testing as part of balloons cadence testing [21:05] sure [21:05] thomi, focus for this should be to test mir out of archive [21:05] and to get some wider HW coverage from the community [21:05] balloons mentioned he is accepting changes via branches ;) [21:06] thomi, thx [21:06] yeah OK, there are a few tweaks I'd make [21:06] olli_: what's the timeline for this? [21:06] the CFT will happen on Sat 8/17 [21:07] so before EOW, preferable earlier [21:07] olli_: OK. [21:07] I'll finish writing this looooong bug report, then get to it :) [21:07] thomi, awesome [21:07] thomi, we will provide them with additional information on [21:07] how to install etc [21:08] as well as release notes type of information (i.e. don't file MM not working...) [21:08] so you don't have to worry about it [21:08] but you can if you want ;) [21:08] olli_: if MM doesn't work, we should remove it from that list [21:08] I'll start putting something together and copy you on it for comments [21:09] cool [21:09] thanks [21:09] coolio [21:09] thx [22:19] Client-focus-notifications v5! [22:28] Is anyone besides tsedgos able to reproduce this mir not working on nexus 4 thing I have been hearing about [22:28] I just tested fresh everything and it seems fine === bschaefer_ is now known as bschaefer [22:42] racarr, ev was having issues [23:08] Ok. What the hell is going on with the jenkins ATI machine? [23:25] RAOF, hey, so I've been seeing these memory leaks, but they look like libEGL and not the clients fault: https://bugs.launchpad.net/mir/+bug/1211982 [23:25] Launchpad bug 1211982 in Mir "Memory leaks in libEGL while running mir" [Undecided,New] [23:25] unless i've missed cleaning something up, or the demos missed cleaning something up [23:26] That does look a lot like a leak in Mir's EGL, doesn't it. [23:27] RAOF, yeah, i've seen the dri2_create_screen leak for sometime now...but figure to at lease poke you about it [23:27] Pity about those ??? frames :) [23:27] yeaah [23:27] Got the -dbgsym packages installed? [23:27] * bschaefer checks [23:27] well im assuming I don't :) [23:29] RAOF, http://paste.ubuntu.com/5983019/ [23:29] for the first one, still some ?? for the second, let me dig for those packages [23:29] Yup, there's the leak. [23:31] RAOF, interesting, well let me collect better info and update the bug [23:32] dang... just empty ??? that are left: http://paste.ubuntu.com/5983021/ [23:33] * bschaefer wondres which package that would be [23:33] libgl1-mesa-dri-dbg, I wager. [23:34] * bschaefer installing [23:35] hey all [23:35] btw, I am collating Mir testing of gpus at https://wiki.ubuntu.com/Mir/GPUTesting [23:35] Hey jono! [23:35] still question marks for that one...but it filled in a bunch of ??? for an uninted var: [23:35] hey RAOF :-) [23:35] http://paste.ubuntu.com/5983027/ [23:35] this should provide a better idea of overall gpu coverage [23:36] RAOF, ill look for some dbg packges, and take a look at the source code to figure out which lib is being called [23:36] thanks for the info! [23:37] the full log atm: just 3 more ??? to go... http://paste.ubuntu.com/5983033/ [23:38] Ah. That ioctl is going to be noise. [23:39] cool, then just 2 leaks that I see in egl apps, and with sdl on the mir side of things :) which is not a lot [23:39] How annoying are these leaks for you? I(should)'ve plugged the create_window_surface one, and I can check out the other one. [23:40] RAOF, not at all annoying, just saw them when I was checking my own memory leakage [23:40] Ok. [23:40] I'll fold them in next time I need to update mesa then. [23:40] RAOF, everything is still running smooth, and its only during creation, and termination of the program so its not a huge deal [23:40] Yeah. You'll also leak once per surface you create, but you're unlikely to do that in a loop :) [23:41] one would hope :), depending on the application...but yes