thomirobert_ancell: I'll email francis and see what he thinks00:11
robert_ancellRAOF, I think I saw you had some bzr magic that tells you if you have merged branches that can now be pruned - is that right?00:34
RAOFrobert_ancell: Correct.00:38
RAOFrobert_ancell: You're looking for the bzr-removable plugin.00:39
RAOFcd ~/.bazaar/plugins; bzr branch lp:bzr-removable removable00:39
robert_ancellRAOF, ta00:46
RAOFfor I in $(bzr removable trunk) ; do rm -r $I ; done00:47
RAOF(Obviously in the repository)00:47
robert_ancellso handy00:49
RAOFYou'll likely find that it misses a number of branches that are actually removable - it's quite conservative. So if you've got any untracked files or uncommitted changes, the branch will not be marked as removable.00:51
RAOF‘bzr unremovable trunk’ will tell you why things aren't removable.00:51
=== owner is now known as Guest76879
thomiduflu: you around?01:49
dufluthomi: Maybe01:49
thomiduflu: any idea how I can narrow down the reason why, as of this morning, compiz is using 80% of my CPU, and Xorg the other 20%?01:49
thomieven when I have no windows open :(01:50
dufluthomi: Wow. What driver/GPU?01:50
thomiwell, other than a terminal01:50
thomiduflu: it's an nvidia GF116M  [GeForce GT 560M]01:50
dufluthomi: Yeah Nvidia's the worst for CPU usage on a good day...01:51
thomiit's never been this bad01:51
thomiI also have a bangin' CPU, so 80% is quite a lot :)01:51
dufluthomi: It's probably some hidden(?) window generating damage01:51
dufluthomi: Try the compiz plugin CCSM>Utility>Show Repaint01:52
thomiduflu: I don't see that plugin in ccsm - is there a separate package I need to install?01:52
dufluthomi: Yes, the compiz-plugins package01:52
dufluthomi: Also, some snapshots of the compiz threads using a script I will email you...01:54
thomiduflu: hmm, interesting. I now have a very colorful desktop01:55
dufluthomi: That's showing the windows generating damage/redraws01:55
thomiduflu: also, it seems like when I enabled that plugin the CPU usage dropped down to 30%01:55
dufluthomi: Perhaps it's the root window; nautilus misbehaving?01:55
dufluOr the unityshell plugin01:56
thomiwow, there we go...01:56
thomiepileptic fit time :-/01:56
thomiback up to 80% :-/01:56
dufluthomi: sudo dstack compiz    # a few times01:57
thomiit seems to be repainting both monitors every frame :-/01:57
dufluAnd then send me the output somehow. In a bug?01:57
dufluI'm not going to *fix* compiz but might be able to tell you what the issue is01:58
dufluthomi: Also try unloading the unityshell plugin. It has a history of doing such things01:58
thomiduflu: http://paste.ubuntu.com/6001417/01:59
dufluthomi: Yep, that confirms something is triggering rendering constantly. And unityshell being slow at rendering makes that 100% CPU02:00
dufluthomi: But unityshell is not always the trigger. Can you try disabling it in CCSM, briefly?02:01
thomiduflu: with unityshell disabled, CPU usage drops to 63%02:02
dufluthomi: Still busy tho?02:02
dufluthomi: OK, try killall nautilus :)02:03
thomihmmm, I don't see the repaint plugin flashing me anymore02:03
* thomi tries re-enabling unity02:04
thomihmm, seems to be better... so far02:05
dufluthomi: Yeah I think I've seen that before. Nautilus randomly goes nuts and since it draws the desktop, that means everything gets repainted02:06
thomiinterestingly, this happened after a reboot as well02:07
thomiit's interesting to see just how often we repaint things, sometimes for no discernable reason02:07
thomifor example, as I type this on the RHS monitor, chromioum is busy repainting all sorts of things on the LHS monitor02:07
dufluthomi: Yes, we browsers will repaint constantly due to animations, somehere, on some tab02:09
duflu-we +web02:09
thomiwell, it seems to be sitting at 60%, which is at least somewhat usable - thanks for your help02:13
dufluthomi: 60% is bad. Is the flashing at least limited to a window now?02:17
dufluI mean, Show Repaint no longer shows the whole screen repainting, right?02:19
duflu... You have to treat the root cause, which sounds like nautilus in this case. Unfortunately using unityshell (< 8.0) and Nvidia drivers will make performance worse02:29
* duflu vanishes to test some ISOs briefly02:54
dufluthomi: I'm curious and concerned... Still 60%? Is showrepaint showing the whole screen repainting or just some windows?03:02
thomiduflu: hey, yeah, still %6403:17
thomibut showrepaint no longer shows whole screens being redrawn03:17
thomiso I think we're one step better than we were03:17
dufluthomi: Without a browser?03:19
thomiduflu: one of my chromium tabs had a notification, which makes chromium fade the tab colour in and out endlessly. if I get rid of the notification, it drops down to 19%03:19
dufluthomi: Also, I think some hacks in unityshell means compiz will get damage events from minimized windows. Which is not ideal.03:20
thomiman, I can't wait until we have mir - I'm sure then we won't see these sorts of problem.... RIGHT GUYS!??03:20
dufluthomi: Yeah Mir has a simpler and more sensible damage model. It also doesn't have Unity7/Nux sitting on top of it to consume yet more CPU/GPU03:22
thomiby then hopefully we'll have better graphics driver support as well03:23
dufluthomi: Tho we can't "fix" browsers or make Nvidia's proprietary drivers any less CPU hungry03:23
dufluBy "fix" I mean stopping them from repainting when they actually need to03:24
dufluYou could "fix" /all/ the web pages to make them more efficient for idle desktops03:24
dufluThere can't be that many web pages03:24
=== chihchun_afk is now known as chihchun
dufluthomi: CPU usage is really an equation with many coefficients. So we don't need to "fix" browsers or drivers if we can lower enough of the coefficients using Mir/Unity803:45
robert_ancellthomi, I'm going to merge in the Ubuntu packaging into LightDM - https://code.launchpad.net/~robert-ancell/lightdm/debian-packaging/+merge/180766. Can you make the appropriate changes to CI / autolanding so it doesn't try and use the obsolete lp:~lightdm-team/lightdm/trunk-packaging?04:59
thomirobert_ancell: yes, but we cannot redeploy those changes until tomorrow05:02
robert_ancellthomi, that's fine05:02
thomiok, cool05:03
thomirobert_ancell: while I'm in here, do you still need/want autolanding & CI for lp:~mir-team/lightdm/unity ?05:05
robert_ancellthomi, nope, that branch is dead05:05
robert_ancellMir support is in trunk now, and in the archive05:06
robert_ancellthomi, oh, did the raring builds get turned off for the staging PPA?05:06
robert_ancellthey seem to still be building05:07
thomirobert_ancell: hmm, I could have sworn I'd landed that change, but it seems they're still configured05:07
thomiremoving them now as well05:07
thomirobert_ancell: FYI: https://code.launchpad.net/~autopilot/cupstream2distro-config/lightdm-config-tweaks/+merge/18076805:11
robert_ancellthomi, does the staging PPA still get lightdm built from trunk? Is that done from somewhere else?05:12
tvoss_good morning05:13
robert_ancelltvoss_, hello05:13
thomirobert_ancell: no, it looks like lightdm trunk gets built into ppa:lightdm-team/daily05:13
tvoss_robert_ancell, hey there :) happy new week05:13
thomihi tvoss_, how's it going?05:13
robert_ancellthomi, ah, ok05:13
tvoss_thomi, good, but my week is roughly 13 minutes old :)05:13
thomirobert_ancell: and in that PPA we build for precise<->saucy and everything inbetyween05:13
tvoss_thomi, how are you?05:13
thomitvoss_: so far so good :)05:14
thomienjoying a quiet monday without all you europeans asking me for things :P05:14
robert_ancellthomi, cool. So I'll delete lightdm from the staging PPA and point people at the lightdm-team/daily PPA if they need to test new changes05:14
robert_ancellthomi, so does the config need updating for the daily PPA to change the packaging?05:14
thomirobert_ancell: yeah. If we want it built into two places we may be able to accomodate that, but TBH that sounds like a bad idea to me05:14
robert_ancellthomi, shouldn't be any need05:15
thomirobert_ancell: that's what I just changed05:15
thomirobert_ancell: diff line 17 from that MP05:15
robert_ancellah, ok05:16
thomialthough, the diff is kind of hard to read, since LP doesn't give us much context around changes :-/05:16
robert_ancellyeah, that's what go me05:16
robert_ancellRAOF, can haz review plz? https://code.launchpad.net/~robert-ancell/unity-system-compositor/apport-pci-info/+merge/18077105:31
robert_ancellstolen shamelessly from xorg packaging05:31
* robert_ancell -> EOD05:32
RAOFdidrocks: Are we able to reproduce the ATI bug on the jenkins machine with the most recent mir?05:42
didrocksRAOF: not sure, as we stopped building it05:42
didrocksRAOF: do you think there is hope?05:42
RAOFdidrocks: Well, https://bugs.launchpad.net/mir/+bug/1212516 is maybe related and might be fixed...05:43
ubot5Launchpad bug 1212516 in Mir "valgrind integration-tests: [ FAILED ] BespokeDisplayServerTestFixture.*" [Undecided,In progress]05:43
didrocksRAOF: ok, I'll ask to have a try later today05:43
dholbachgood morning07:01
tvoss_dholbach, salut07:16
dholbachhi tvoss_07:18
Saviqmorning all07:24
RAOFAh, yes. When changing XMir ABI it's important to actually update the drivers…07:26
* duflu is still impressed every time he sees Mir smoothly and reliably use multiple monitors07:28
RAOFThen prepare to be amazed by XMir smoothly and reliably using multiple monitors!07:28
RAOF*: FSVO ‘smoothly’07:28
RAOF*: FSVO ‘reliably’ ☺07:29
RAOFFor Some Values Of.07:29
dufluAh, my math terminology is rusty07:30
dholbachsudo apt-get install bsdgames; wtf fsvo      :)07:31
RAOFGone from no output to infinite loop after presenting the first frame!07:32
dufluIt's /a step/07:32
dufluOne small step for correctness, one giant leap for fan noise07:33
RAOFGah. Stupid closures in C and their stupid type-erasing void *ctx.07:41
tvoss_RAOF, https://code.launchpad.net/~thomas-voss/mir/fix-fd-sharing/+merge/18080107:49
tvoss_mzanetti, good morning :)07:50
mzanettigood morning tvoss_07:50
tvoss_mzanetti, hey :) so can you check if our friend the upstream merger is still alive?07:50
mzanettihehe, sure07:50
mzanettitvoss_: what makes you think it isn't?07:52
tvoss_mzanetti, I pushed to https://code.launchpad.net/~thomas-voss/platform-api/location-service/+merge/18069207:53
tvoss_mzanetti, an hour ago, and still no vote07:53
mzanettitvoss_: right. I'll check07:53
mzanettitvoss_: otherwise it seems to do stuff. not sure why this one wasn't picked up07:53
tvoss_mzanetti, thx :)07:53
didrockstvoss_: "no vote" -> maybe he's undecided and wants to vote blank ;)07:54
tvoss_didrocks, well, I always figured that our jenkins setup might evolve to become skynet, but I haven't expected it that early07:54
mzanettitvoss_: oh... the build queue holds like 50 jobs :/07:55
tvoss_mzanetti, oops07:55
mzanettithis is not looking good.07:56
didrockstvoss_: ahah ;)07:56
alf__duflu: Hi! Bypass question: why do we care about transformations of non-fullscreen windows in the bypass filter?07:58
duflualf__: Because no one uses the fullscreen state yet. It's not "wired up"07:58
dufluWhen it is, the certainly we'll use it07:59
alf__duflu: no, I mean we are checking for orthogonality of all windows, why is that?07:59
duflualf__: Because a rotated window should not get bypassed. It would lose transformation08:00
dufluOh, I see08:00
dufluHang on08:00
dufluThat was to fix a bug, let me find it08:01
duflualf__: It's to reliably detect overlaps. If a window overlaps a fullscreen window then the fullscreen window can't be bypassed. And I do calculate the overlap for "orthogonal" windows, but for rotated windows it's too hard/unreliable. Safer to assume if anything is rotated (not orthogonal) then some fullscreen window "might" be overlapped.08:02
dufluAnd there are test cases which will fail otherwise...08:03
alf__duflu: ok, thanks08:04
duflualf__: Also, I suspect in future that rotation will not be an attribute of a surface, but rather a function of an effect or animation that's in progress. If that changes then we will test for "any effects in progress" instead08:05
duflualf__: I now see however that I shouldn't be caring about rotated windows that are underneath08:09
dufluUnless we start using Z-buffers08:10
alf__duflu: I think that's ok for now (for the XMir use case)08:10
duflualf__: I think it may even become preferable to be "less correct" in future to stop more edge cases from accidentally un-bypassing a fullscreen window08:11
dufluAlthough that's less important for Mir where bypass can be switched on/off between frames. It was important in Compiz where there's a 2-frame lag08:12
duflualf__: Actually that's a fun multimonitor test for bypass: Run egl{plasma,triagnle} -n -f, and then a second smaller windowed client. See how the first fullscreen client speeds up when there's nothing overlapping it08:15
alf__duflu: ack, will try it out08:15
Saviqtsdgeos, so it's not looking like racarr got anywhere on Friday? :/08:17
tsdgeosi think not :_/08:18
tvoss_Saviq, is ogra_ back?08:18
tvoss_ogra_, ping08:18
Saviqtvoss_, was he away?08:18
tvoss_Saviq, ah no, that was rsalveti08:18
Saviqtvoss_, canonicaladmin.com says he should be back today08:19
tvoss_Saviq, ah08:19
Saviqgreyback, tsdgeos let me know what I can halp with, will try the app auth out now08:19
ogra_tvoss_, yup08:20
greybackSaviq: thanks08:20
tvoss_ogra_, hey, did something change in the lxc-android-container configuration in terms of memory limits?08:20
ogra_tvoss_, nope, not that i'm aware of08:21
ogra_hmm ... unless ...08:21
ogra_ask the security team, they might have tightened apparmor rules08:21
ogra_thats the only thing that comes to mind08:22
tvoss_ogra_, for some weird reason, when Mir tries to access the framebuffer, we see ENOMEM in ptrace08:22
tvoss_ogra_, thanks for the hint08:22
ogra_tvoss_, oh, there were also apparmor kernel changes now that i think of it, but i dont think xnox has rebuilt the android package for them yet08:24
ogra_might be that something is out of sync there08:24
xnoxogra_: ack.08:25
tvoss_ogra_, thanks08:25
mzanettitvoss_: jenkins doesn't come up again after a restart. I fear we need to live without it for the next couple of hours until some of our jenkins experts are online08:39
tvoss_mzanetti, thx08:40
didrockstvoss_: maybe ask gema? her team should have some people for this kind of production issue?08:41
alan_ghikiko: welcome back. Good break?08:49
hikikohi alan_g :) yes it was nice!08:50
alan_ghikiko: Are you caught up yet? We landed some nested mir stuff while you were away and more is waiting for review: https://code.launchpad.net/~alan-griffiths/mir/native-platform/+merge/18061808:52
hikikolet me get a look at it :)08:52
alan_ghikiko: Once you've done that we should plan the next steps together.08:55
hikikoalan_g, you merged the nested-display branch and you added the NestedGbmPlatform isn't it?08:59
Saviqtvoss_, oh, interesting find (re ENOMEM)09:00
alan_ghikiko: *Native*GbmPlatform (and NativeAndroidPlatform)09:01
tvoss_Saviq, yup, that was the only thing that was hinting to an issue09:02
hikikosorry that's what I meant I am pulling :)09:04
tvoss_didrocks, can you give https://code.launchpad.net/~thomas-voss/mir/fix-fd-sharing a try on the ati machine09:19
didrockssil2100: ^09:20
didrockstvoss_: weird that RAOF told me that latest trunk may fix it09:21
didrocksas it's not merged yet09:21
didrocksso we will need a ppa + rebuild09:21
RAOFdidrocks: That's another possible fix.09:21
didrockssil2100: handling that?09:21
didrockssil2100: maybe take latest mir trunk + merge that09:22
didrocksand push to a ppa09:22
didrocksonce built, push u-s-c there09:22
didrocksand we can then just test installing from the ppa?09:22
dufluHmm, I could return to trying to figure out test case side-effects and whether they're important...09:49
duflu... or I could finish early and work on preparing a healthy meal09:50
dufluRight then09:50
hikikoalan_g, when I build trunk09:51
hikikoI get this:09:51
hikikomake[2]: *** No rule to make target `/usr/lib/libboost_chrono.so', needed by `lib/libmirclient.so.1'.  Stop.09:51
hikikoalthough I have libboost-chrono-dev installed09:51
alan_ghikiko: your cmake cache is out of date with saucy09:52
alan_gdelete it and run cmake again09:52
hikikobut I just did bzr branch lp:mir :s09:52
hikikoI will try09:52
alan_ghikiko: that's odd then09:53
alan_ghikiko: are you using any ppas?09:54
Saviqgreyback, oh, tricky... webbrowser creates a separate surface it seems09:54
Saviqgreyback, and it gets rejected09:55
greybackSaviq: quite likely, as the webprocess separate. Good catch...wonder how to work-around09:55
Saviqgreyback, indeed09:55
hikikoalan_g, mir-team/staging/ubuntu and http://ppa.launchpad.net/phablet-team/ppa/ubuntu09:56
Saviqgreyback, it feels like unity8 starts faster with Mir, are you seeing the same?09:57
greybackSaviq: hud is disabled :)09:57
greybackmust re-enable it09:57
hikikoalan_g, I am dist-upgrading maybe that will solve the problem, I didn't upgrade for 10 days or more09:57
Saviqgreyback, desktop hint → happroved09:57
greybackSaviq: thanks.09:58
alan_ghikiko: you shouldn't need the ppas as everything ought to be in the distro09:58
hikikook, I ll remove them and try again, thanks!09:59
hikikoalan_g, the branch looks good but shouldn't we call the NativeGbmPlatform NativeGBMPlatform to match the GBMPlatform? could I rename it?10:21
alan_ghikiko: we should, but as alf__ top-approved without that tweak we can do it in a future MP (it only affects a couple of lines in a .cpp we need to change anyway)10:22
alan_ghikiko: you building OK now?10:26
hikikoyes alan_g10:27
hikikoafter upgrading everything is fine10:27
=== hikiko is now known as hikiko|lunch
=== tvoss_ is now known as tvoss|lunch
Saviqtsdgeos, hey, I might take over QDesktopServices::openUrl from you, interested?10:49
tsdgeosSaviq: all yours, i'm doing some unit/auto/something-tests for greyback10:49
Saviqtsdgeos, yup, know that10:50
Saviqtsdgeos, any quick pointers where I should put that in (e.g. the DBus service backing it)?10:50
tsdgeosSaviq: tbh not sure what's needed, so we want QDesktopServices::openUrl to do what? open the appropiate app for that url? do we have such system in place?10:51
Saviqtsdgeos, a preliminary one, yes10:52
tsdgeosand that's going to call ApplicationManager::startProcess or something unrelated to start the app?10:52
Saviqtsdgeos, https://launchpad.net/url-dispatcher10:52
Saviqtsdgeos, upstart10:53
tsdgeosSaviq: ok, why not ApplicationManager::startProcess? it needs to have upstart support afaics from the comments10:53
tsdgeosmay as well have the upstart code only once?10:53
Saviqtsdgeos, yeah, it seems url-dispatcher is a self-contained thing now10:53
Saviqtsdgeos, will talk to tedg on that later10:54
tsdgeosSaviq: i would understand that it should go in unity-mir, probably10:56
Saviqtsdgeos, yeah, that's for sure10:56
tsdgeosnot sure which class exactly10:56
tsdgeosi don't think ApplicationManager is a bad place10:57
tsdgeosbut gerry may disagree10:57
Saviqtsdgeos, we agreed on that before already10:57
greyback_fine with me10:57
tsdgeosah, did we?10:57
* tsdgeos can't read lately10:57
* greyback_ rebooting as Mir seems to freeze his machine10:58
greyback_alan_g: ping10:59
alan_ggreyback_: hello10:59
greyback_alan_g: hey there, I need some advice from you10:59
greyback_alan_g: I am trying a simple unity-mir server on my desktop. Something is not happy, as when running it, my machine's graphics completely hangs. I can ssh in, but could you give me tips on how to debug?11:00
mlankhorstdoes ssh work?11:01
greyback_mlankhorst: yes11:01
greyback_I suppose catching output into a file is necessary for a start11:02
greyback_alan_g: also, if using XMir, would my Mir instance with XMir collide? (I've disabled xmir)11:02
alan_ggreyback_: I usually try running from an ssh session so that I can see errors. (and maybe turning on some logging)11:02
alan_ggreyback_: yes, we don't have nested mir sessions working yet11:03
greyback_alan_g: ack11:03
greyback_alan_g: any way to confirm xmir not running?11:04
alan_ggreyback_: I'm the wrong one to ask (I have never enabled it)11:05
greyback_alan_g: ok, thanks11:05
=== greyback_ is now known as greyback|unstabl
alan_ggreyback_: I'd guess there's be a u-s-c process running (but I don't actually know)11:06
Saviqgreyback|unstabl,  lp:platform-api/mir is merged into lp:platform-api already?11:06
greyback|unstablSaviq: yes11:08
Saviqgreyback|unstabl, good, thanks11:08
greyback|unstablSaviq: same with qtubuntu11:08
greyback|unstablalan_g: ever got this exception: "Failed to mute keyboard" when launching over ssh? Would it be that I'm sshing into my primary machine (that I'm typing this on)11:10
alan_ggreyback|unstabl: yeah, sounds like X owns the keyboard. You need to be in a VT (and tell your server which one with --vt)11:12
greyback|unstablalan_g: aha, --vt! Thanks11:12
alan_ggreyback|unstabl: ./mir_demo_server --help is wonderful11:13
tsdgeosgreyback|unstabl: how do i "stop" a QMirServer?11:18
tsdgeosit seems that returning from the function i pass to runWithClient is not enough11:18
=== hikiko|lunch is now known as hikiko
=== alan_g is now known as alan_g|lunch
kgunnalf__: hey...nice job on mm...12:49
alf__kgunn: thanks, did you try it out?12:53
kgunnalf__: not yet....was just looking at what landed while i was out....12:54
kgunnalf__: and what new problems we have :)12:54
=== kode54_ is now known as kode54
=== alan_g|lunch is now known as alan_g
kgunntvoss|lunch: so are we able to pilot https://code.launchpad.net/~thomas-voss/mir/fix-fd-sharing/+merge/180801 on didrocks machine? (sorry if this is old news)13:52
tvoss|lunchkgunn, I think sil2100 did13:54
=== tvoss|lunch is now known as tvoss_
kgunnsil2100: how many successful runs? i only ask because this is a heisenbug....i know didrocks got 10 in a row once13:55
kgunnwith the "current failing config"13:55
tvoss_didrocks, got an update on https://code.launchpad.net/~thomas-voss/mir/fix-fd-sharing/+merge/18080114:02
didrockstvoss_: as you told, sil2100 is on it14:03
didrocksI didn't got any update since, he had some issue with his ppa, then I think he copied to another one but didn't get feedback14:03
didrockslet's wait for him :)14:03
sil2100didrocks: running the tests, but so far no failures - but no many tests have passed yet14:09
didrocksif 2 tests pass, we're fine TBH14:10
didrockssil2100: you will rerun it then?14:10
didrocks(hum, aren't we at next tick though?)14:10
didrocksstill 30 minutes until the first stack starts I guess14:10
sil2100I know ;p14:10
sil2100didrocks: hmmm14:12
sil2100didrocks: looks to me like 'it' ;/14:12
sil2100didrocks: since it seems that it stopped on the opengl14:12
sil2100kgunn: ^14:12
didrockstvoss_: ^14:13
didrockssil2100: we'll need to free it for production in some minutes (reboot electrically directly the machines to avoid having next test failing)14:13
didrockssil2100: let's wait a little bit to see if kgunn/tvoss have a chance to get to it14:14
sil2100tvoss_: ?14:24
sil2100didrocks: I guess we might have to reset the machine?14:24
kgunnsil2100: didrocks ...can we get another opportunity to build in the patch14:24
didrockssil2100: agreed14:24
kgunnfrom RAOF http://paste.ubuntu.com/5983202/14:24
kgunnunderstanding this might be a while14:25
didrockskgunn: IIRC, this one was already tried14:25
didrockskgunn: without any success14:25
kgunndidrocks: its not a solution...14:25
kgunnits debug14:25
didrocksyeah, we tried it14:25
didrocksand couldn't get any blocking situation14:25
didrocks(after 20 runs)14:25
didrocksdamn heisenberg… :p14:26
kgunndidrocks: oh...the debug patch changes timing...14:26
didrocksright :/14:26
kgunndidrocks: sil2100 ...can you at least grab the logs off it before rebooting14:27
didrockssil2100: mind doing that? see the logs I attached to the bug and paste the same14:28
didrockssil2100: just log into the container and installs pastebinit14:28
didrocksit's the easiest14:28
didrockssil2100: can help you if needed (but in meetings)14:28
sil2100Doing ;)14:29
sil2100Looking at the bug first14:29
sil2100kgunn, tvoss_: attached the logs to the bug as a comment14:35
kgunnsil2100: thank alot14:35
sil2100didrocks: would it be enough if I just restart the machine through ssh?14:36
tvoss_sil2100, thx14:36
didrockssil2100: yeah14:37
sil2100didrocks: do I have to detach the lxc containers first or nothing needed ;p ?14:37
didrockssil2100: no no, it's fine14:37
didrockssil2100: just reboot outside the container though :p14:38
didrocksor you just shut down the container then :p14:38
sil2100Rebooting normally then, thanks ;) I prefer to double-ask with fragile things14:38
tvoss_sil2100, does not look like anything failed judging from the logs14:40
sil2100tvoss_: didn't paste in compiz logs, but it's the same as before - it's stopping on loading the opengl plugin14:41
tvoss_sil2100, ah, what was the error message again?14:42
sil2100tvoss_: no error message really, just this from compiz, as seen in this comment: https://bugs.launchpad.net/mir/+bug/1204939/comments/1114:44
ubot5Launchpad bug 1204939 in Mir "Unity doesn't start on ATI test machine (Mir fails to respond to drm_auth_magic request)" [Critical,Triaged]14:44
kgunnSaviq: confirmed with alan_g ....current mir worked on old image....so, makes me wonder if we updated android trees or move kernel14:45
kgunnmore so than i did14:45
Saviqkgunn, right14:45
kgunnSaviq: even better...alan_g says old mir fails on new image14:46
Saviqricmm, hey, wanted to ask, where should put the platform api part of QDesktopServices::openUrl()?14:46
ricmmSaviq: I need more context14:47
Saviqricmm, it'll effectively be just a DBus call to AppManager14:47
tvoss_sil2100, so it segfaults?14:47
Saviqricmm, that will ask it to open the url provided14:47
alan_gkgunn: Saviq - alf__ tried an old mir on the new image14:47
=== alan_g is now known as alan_g|tea
sil2100tvoss_: hm, hard to say this time, I would assume yes but it says nothing in the logs: jenkins.qa.ubuntu.com/job/autopilot-saucy-daily_release/1104/label=autopilot-ati/console14:48
sil2100tvoss_: scroll down to the bottom and try finding compiz init phase14:48
tvoss_sil2100, I get an error accessing https://jenkins.qa.ubuntu.com/job/autopilot-saucy-daily_release/1104/label=autopilot-ati/console14:49
sil2100hm, then maybe <- with VPN access14:49
=== alan_g|tea is now known as alan_g
=== dholbach_ is now known as dholbach
=== sfeole` is now known as sfeole
kgunnracarr_: ping16:17
kgunnso wrt "mir doesn't  work on n4"...the current belief is that its a um/km out of sync issue....16:19
kgunnlooks like at ~1:30 today there will be a 19.2 image available (where all will be resync'd)16:19
kgunnracarr_: so basically....save beating your head on the n4 mir thing until then....&16:20
kgunnhopefully it will just be corrected by the image creation...can you test that this afternoon ??16:20
racarr_kgunn: um/km?16:20
racarr_I definitely can! and that's16:20
kgunnuser mode/kernel mode16:20
racarr_good to hear16:20
racarr_oh ok16:20
racarr_yes that16:20
racarr_seems really feasible16:20
racarr_it was getting absurd on friday when I was trying to figure it out :p16:21
kgunnracarr_: yeah...sounds like it was android um as well as ubuntu um code that was outta whack16:21
kgunnat least ogra seems to be strongly thinking the same16:21
ogra_err, not exactly :)16:22
ogra_apparmor userspace and kernel got updated ... but the kernel didnt make it into the images yet, so something *could* be out of sync16:22
* ogra_ never said "strongly" ... it was just the best guess i could make on the situation you guys hit 16:23
racarr_ogra_: It might not be possible for it to be that...I think I tried16:24
racarr_to unload apparmor16:24
racarr_but im not confident I did it correctly so or the mismatch could have made that16:24
racarr_so I will test the image16:24
=== racarr_ is now known as racarr
ogra_with luck the fix made it into 19.116:24
ogra_it will definitely make 19.2 though16:24
ogra_(which starts building at 20:00 UTC ... so should be available in a few hours)16:25
ogra_racarr, to do it right you need to disable apparmor on the kernel cmdline16:25
ogra_(apparmor=0 iirc)16:26
racarrogra_: Ah ok16:26
racarrIt sounds pretty feasible16:26
kgunnogra_: "updated kernel"....so besides apparmor...there are other updates to the kernel ?16:31
ogra_kgunn, https://launchpad.net/ubuntu/saucy/+source/linux-mako/3.4.0-3.2016:32
ogra_and as i said above ... i was only guessing, not sure it is actually your issue16:35
=== chihchun is now known as chihchun_afk
ogra_kgunn, racarr: http://paste.ubuntu.com/6003505/ these are two scripts that help with manipulating the kernel cmdline on touch images16:51
ogra_so just try appending the apparmor=0 and see if it helps16:52
robotfuelit looks like 3d acceleration broke on nouveau starting on friday, for regular x as well.17:04
=== alan_g is now known as alan_g|EOD
racarrfinally have time to do powermangement17:24
racarrit seems maybe dpms_mode is a member of17:24
tvossracarr, ping18:11
=== alex_abreu is now known as alex-abreu
racarrtvoss: pong18:55
racarrWas lunching :)18:55
tvossracarr, cool, just wanted to say that I'm around to help out with the nexus 4 issue18:56
racarrtvoss: The rumor is that it could be a kernel space userspace mismatch for apparmor18:57
racarrso waiting on new images to build18:57
racarrwhere the rumor was they would build around 1:30 today18:58
tvossack, in ~1.5 hours from now, right?18:58
racarrill try ith apparmor=0 right now :)19:00
racarrHmm it doesn't work :(19:02
racarrogra_: ^ :(19:02
racarrI think I ill wait for the new images though to do anything more here19:02
kgunnracarr: can you elaborate on "doesn't work" ??19:43
racarrkgunn: I mean with apparmor=0 nothing changes19:44
kgunnracarr: meaning our theory is busted ?19:45
racarrkgunn: Probably!19:45
racarrbut it could be something else with the kernel19:45
racarror maybe something can someho load apparmor later, and something fishy19:45
racarrso I have decided to work on DPMS and wait for the new image19:46
racarrto know for sure19:46
kgunnracarr: for your next attempt...could you add a report of the egl error like so19:50
kgunnin android_framebuffer_window.cpp19:50
kgunn...or something like that19:51
racarrkgunn: ok19:57
kgunnracarr: thanks...19:57
racarrBased on some other errors I think either the framebufer isnt' allocating19:58
racarror some ioctl inside the driver to reference it is failing19:58
racarrso this might be useful19:58
racarrI may need to talk to alf before I can finish this DPMS stuff20:05
racarrI'm not sure we have theinfrastructure in place to apply display configuration changes without reallocating the scanout buffers20:05
racarrand don't want to step on things20:05
racarrThat could also just be what happens in the first implementation of DPMS but that kind of20:06
racarrdefeats the point I think20:07
kgunnracarr: kinda weird...i would20:08
kgunn've thot that20:08
kgunndps would've been more about cutting clocks to the display/gpu20:09
kgunnrather than messing with mem allocations....altho...maybe the system could save power on mem rails if it frees enoug ?20:10
racarrkgunn: Right, I just mean the way mir is set up20:11
racarrI was encoding the DPMS mode as part of the display configuration20:11
racarrwhich seems kind of correct20:12
racarrbut when you apply a display configuration it rebuilds everything20:12
racarrwhich is correct for most other things you might ant to change.20:12
racarrbut the DPMS settings dont change the layout or anything20:12
racarrso it just requires some reworking of the existing code20:12
racarrmaybe I can figure it out today :)it's almost done though so I think syncing iwth alf in the morning might be a good plan20:13
racarrand back to nexus 4 soon20:13
kgunnracarr: gotcha....you're just trying to avoid triggering an unnecessary tear down20:13
kgunnalf__: ^ in case you read scrollback :)20:14
racarrkgunn: Yeah :)20:18
racarrI feel like today is the day for client focus notifications.20:21
racarrBack in 20 to switch to nexus 420:22
racarrNexus time20:39
kgunnbregma: robert_ancell ....well...gooogle....that was rude20:41
robert_ancellwe're back20:41
racarrhttp://cdimages.ubuntu.com/ubuntu-touch/daily-preinstalled/pending/ hot of the press ;)20:43
robert_ancellolli_, do we have a hangout for the CFT meeting?20:51
* kgunn waits for word from racarr 20:54
racarrkgunn: Flash is alllmost done then takes a bit to install mir20:55
racarrfingers crossed20:55
racarrwow surface flinger took like 30 seconds to come up20:59
olli_robert_ancell, will add20:59
olli_robert_ancell, kgunn https://plus.google.com/hangouts/_/calendar/b2xpdmVyLnJpZXNAY2Fub25pY2FsLmNvbQ.miib2j0d57b9590jbd8dufi79c21:00
tvossracarr, any new insights?21:18
kgunntvoss: apparmor=0 didn't change the result21:21
racarrAlso it's still broken on the new image :(21:22
tvossracarr, okay, and you said that surfaceflinger is slow to start?21:22
racarryes that's true it was really slow to start. I haven't tested that again21:23
racarr> 30 seconds to the point where I started trying to  investigate if it was broken21:23
racarrthen it came up21:23
racarrIm building mir again now (actuallyd onloading 391mb of build dependencies to the phone woohoo)21:23
racarrthen probing the egl errors21:24
racarrsometimes there have been errors about not being able to find an egl config21:24
racarrkgunn: Wait where did that21:24
racarrexception come from21:24
racarrI saw it once and was never able to get it again (went away after reboot)21:24
kgunnracarr: its the exception that alan documented in the bug21:25
kgunnmy assumption was it was repeatable21:25
kgunn(or consistent)21:25
racarrkgunn:  not exactly I got that once21:27
racarrmany times ina  row21:27
racarrbut then when I rebooted it went back21:27
racarrto th same no errors except for the ioctl failures in strace21:27
racarrmaybe once I build mir I will get that exception this time? :/21:27
racarrI dont get it from running mir_demo_server on the fresh image21:27
racarralso for me/tmp/mir_socket is created21:28
racarrare stuck at half FPS21:33
racarr1 FPS21:34
racarr7 FPS21:34
racarr26 FPS21:34
racarr28 FPS21:34
racarr29 FPS21:34
racarr30 FPS21:34
racarr30 FPS21:34
racarrnot that anything shows up21:34
racarroh wo I left one open for a while21:34
racarrterminate called after throwing an instance of 'boost::exception_detail::clone_impl<boost::exception_detail::error_info_injector<std::runtime_error> >' what():  error posting with fb device21:34
robert_ancellrobotfuel, oh you are here already :)21:45
robert_ancellrobotfuel, we'd like to verify https://code.launchpad.net/~vanvugt/mir/bypass/+merge/180503 against the test machines before landing it - what is the best way to do that?21:46
robotfuelrobert_ancell: how do I test it?21:48
robert_ancellrobotfuel, Would a PPA fit in with the test system?21:48
robert_ancellOtherwise it's just building from source21:49
robert_ancellThe tests are the same as before - if you can run a session and run a GL app then it's good to go21:49
robotfuelrobert_ancell: yes ppa's fit as long as ps-jenkins can see them.21:49
robotfuelrobert_ancell: what ppa is it in?21:54
robert_ancellrobotfuel, I'll set one up and email you the details21:55
tvossracarr, kgunn which image is the latest that Mir is working with?21:55
kgunntvoss: unsure...i'm guessing circa 2 weeks ago (based on the bug info)22:00
kgunnmeaning 2 week old image22:00
racarrtvoss: Seems about right22:01
racarrmaybe even ~1 week22:01
kgunnracarr: should we bisect ?22:02
racarrkgunn: seems so...:(22:03
racarr^Cphablet@ubuntu-phablet:~/src/trunk/build/bin$ sudo ./mir_demo_standalone_rendeto_fb22:03
racarr__pthread_gettid -222:03
racarrterminate called after throwing an instance of 'boost::exception_detail::clone_impl<boost::exception_detail::error_info_injector<std::runtime_error> >' what():  Could not unblank display22:03
racarrwhy does22:03
racarrit keep moving22:03
racarrphablet@ubuntu-phablet:~/src/trunk/build/bin$ sudo ./mir_demo_standalone_render_surfaces22:03
racarr__pthread_gettid -222:03
racarrERROR: /home/phablet/src/trunk/src/server/graphics/android/hwc_common_device.cpp(64): Throw in function mir::graphics::android::HWCCommonDevice::HWCCommonDevice(const std::shared_ptr<hwc_composer_device_1>&, const std::shared_ptr<mir::graphics::android::HWCVsyncCoordinator>&)22:03
racarrDynamic exception type: boost::exception_detail::clone_impl<boost::exception_detail::error_info_injector<std::runtime_error> >22:03
racarrstd::exception::what: Could not unblank display22:04
racarr[boost::errinfo_errno_*] = 1, "Operation not permitted"22:04
racarrmore fascinating22:04
kgunnracarr: so basically new error code every time22:04
racarrit seems like new error code with new image22:04
racarron this image so far it's consistent22:04
racarrok so the mir from the archive ust silently fails22:04
racarrwhen I build mir it always fails with22:04
racarr"Could not unblank display"22:04
racarroperation not permitted22:04
racarrthe first demo ust doesnt print the errno22:05
racarralways = always on this image22:05
racarris root22:06
racarrnot root22:06
racarrof course now I reboot and its back to silently failing22:08
racarrkgunn: After reboot that error is now gone forever seemingly and it's back to failing silently22:14
kgunnracarr: what the what?...so consistently fails with no permission to blank display ....then fails silent ? (meaning just a hang, no exception thrown)22:15
kgunnracarr: well...bisecting will likely be better use of cycles i think22:17
racarrno hang22:17
racarrjust nothing on screen22:17
kgunnracarr: got it...(sorry misuse of the word hang)22:17
kgunnracarr: what happens if you try a demo client in that state ?22:18
racarrjust the same half FPS stuff as before22:19
racarrthe files in /usr/include/android/hardware22:19
racarrthe headers22:19
racarrare significantly different than the files in 3rd-party/android-deps22:19
kgunnracarr: but at least, the client can get an egl context and render to the fb22:19
racarrkgunn: Seemingly it can22:20
racarrbut nothing happens22:20
racarrI assume there are the same strace errors there were on the last image with the ioctl failing22:20
kgunneeeewwww....header changes huh22:20
racarrmaybe libhardware itself changed22:20
racarror maybe this is normal and the files in android-deps are magical I don't really know the22:21
racarrstory around them22:21
kgunnracarr: actually this is what i was fearing...that egl/hwc/fb glue we can't see possibly changes22:21
tvosskgunn, we can see all of the glue22:22
kgunntvoss: you have egl & hwc implementations ?22:22
racarrlibhardware itself is open afaik22:23
tvosskgunn, racarr exactly22:23
robotfuelrobert_ancell: this is blocking nexuiz from running on nvidia and radeon https://bugs.launchpad.net/ubuntu/+source/xorg-server/+bug/121403022:23
ubot5Error: launchpad bug 1214030 not found22:23
racarrI'm not sure im interpreting things right22:24
robert_ancellrobotfuel, ta22:24
racarrbut it seems like if these headers are out of sync, they have been for ages22:24
racarrso it seems to not matter22:24
racarrwhat do the different headers mean then22:24
tvossracarr, are we building against libhardware?22:24
robotfuelrobert_ancell: do I need to make that bug public or can you see it?22:25
robert_ancellrobotfuel, I can see it - any reason not to make it public?22:25
racarrtvoss: Oi, mir is22:25
racarrbut it has it's own headers in 3rd-party/android-deps22:26
kgunntvoss: racarr ....sure, there is an open source impl...but what about vendor blobs ?22:26
robotfuelrobert_ancell: not that I am aware of, I guess apport does that by default.22:26
tvossracarr, so which headers do we pull into the build?22:26
racarrtvoss: For mir we use the headers in 3rd-party/android-deps22:26
kgunnmost vendors replace hwcomposer.cpp22:26
racarrseemingly, because I just copied the headers from /usr/include in to there22:26
racarrand it doesn't build anymore ;)22:27
tvosskgunn, no worries, all good. It's only about symbols22:27
tvossracarr, okay, and then we link against what? libhardware?22:27
tvossracarr, is that yes or no?22:28
robotfuelrobert_ancell: there is also this bug that is blocking us from getting results on nvidia22:28
ubot5Launchpad bug 1214066 in xorg (Ubuntu) "phoronix benchmarks on nouveau is 4 times slower today than it was last thursday" [Undecided,New]22:28
robotfuelrobert_ancell: it's a regression in regular x as well22:28
racarrtvoss: Yes22:28
racarrtvoss: We link against system libhardware but explicitly use the22:28
racarrin tree headers22:28
tvossracarr, here we go ...22:29
robotfuelrobert_ancell: benchmarks are timing out on nvidia because acceleration is non-existent or slow22:29
tvossracarr, do we _link_ against libhardware on the ubuntu side or are we using hybris?22:30
racarrtvoss: Err,as far as I can see we are linking against the22:31
racarrhybris libhardware22:31
tvossracarr, okay, even worse than22:31
racarrthere is no hwcomposter.h in /usr/include22:31
tvosskgunn, when is kdub back?22:33
racarrthere is some series of macros preventing me from immediately seeing why this compile is faling XD but it's something about version defines/compatibility macros22:33
racarrthis compile is when I tried to use the system headers22:33
racarrI think22:33
racarrif I could get a new22:34
racarrit would work22:34
kgunntvoss: sept222:35
tvosskgunn, really?22:35
rsalvetiracarr: which version of hwcomposer.h are you originally using?22:38
rsalvetitvoss: indeed, that's not part of the android-platform-headers22:38
rsalvetiI can add it there, but I first want to check how different it is from our current hwcomposer.h from 4.2.222:38
tvossracarr, ^22:39
racarrrsalveti: see in lp:mir 3rd-party/android-deps/hardware22:39
racarrPhone sorry, will be back in 5 min with more info XD22:39
racarrso...um. I guess I dont know22:42
racarrhow to identify what version of hwcomposer.h it is22:42
racarror why we use these in tree headers22:42
racarris it legacy only? or is there another reason22:42
rsalvetiracarr: http://phablet.ubuntu.com/gitweb?p=CyanogenMod/android_hardware_libhardware.git;a=tree;f=include/hardware;h=139b8bb682030607ffea03beca691ef877a5e794;hb=refs/heads/phablet-saucy22:43
rsalvetithis is the one we're using at our android images now22:43
kgunnhwcomposer_defs.h changed as well22:44
kgunnracarr: ^22:44
racarrrsalveti: Ok thanks. I am going to see if I can get mir to compile22:44
racarragainst these headers22:44
rsalvetiracarr: tvoss: http://paste.ubuntu.com/6004566/22:45
rsalveticomparing the headers, nothing really critical22:46
rsalvetionly difference is removal of HWC_BLIT22:46
racarrrsalveti: Err, ok so these22:46
racarraren't the headers22:46
racarrthat are in /usr/include/android22:46
racarron the device22:46
racarrthe diff there is much larger22:46
rsalvetiracarr: there's no hwcomposer.h in there22:47
rsalvetithat's why I'm comparing with what got changed in the android side22:47
rsalvetito see if the headers inside mir would need any sort of update22:47
rsalvetiracarr: what was the last known working image?22:48
racarrI had a hardware.h with a much larger difference22:48
racarrand no it wont build22:48
racarrrsalveti: No one knows. 2 weeks ago should be safe though22:48
kgunnrsalveti: yeah...but that addition of ifdef qcom hw adds an int right in the middle of a hwc struct22:48
kgunnthat's used for hwframebuffer22:48
rsalvetiright, can't we get someone to trace that based on the ones available in http://cdimage.ubuntu.com/ubuntu-touch/ ?22:49
rsalvetiwe got one from 12 there, would be nice to see if it works there22:49
kgunnracarr: can you test http://cdimage.ubuntu.com/ubuntu-touch/daily-preinstalled/20130812/22:50
rsalvetikgunn: right, but I don't think QCOM_HARDWARE is defined at all for nexus 422:50
tvossracarr, rsalveti off now, quite late here. kgunn, racarr can you guys send me a handover mail?22:51
kgunntvoss: you bet22:51
kgunnrsalveti: really...? that would surprise me...isn't n4 qcom ?22:51
racarrtvoss: Ok22:52
rsalvetikgunn: it is, but iirc that define was cm specific for other qcom based devices22:52
racarrrsalveti: http://pastebin.ubuntu.com/6004587/22:52
rsalvetibut would need to double check, and you can easily rebuild against the updated header to test as well22:52
racarrthis is the diff I get when copying /usr/include/android22:52
racarrin to the android-deps22:52
racarrso where are22:52
racarrTHESE files coming from?22:52
racarrusing the files from github with the small diff didnt ork22:53
racarrinteresting trivia: Cmake does not detect changes to the android-deps folder ;)22:55
rsalvetiracarr: these are from libhybris, which originally came from our phablet git repos22:55
rsalvetibut they were always different, that's why I'm not sure if this is indeed the cause22:55
racarrthats hat the changelogs seemed to say to me too22:56
rsalvetitracing the last working image would be a better idea I guess22:56
racarrbut I am double checking because hat else could it be22:56
racarrkgunn: http://cdimage.ubuntu.com/ubuntu-touch/daily-preinstalled/20130812/22:56
racarrstill ant me to test that one?22:56
kgunnracarr: i think that would help22:57
racarrok the thing is installing the PPA on this image will bring in other updates I guess so I have to23:08
racarrbuild mir again..23:08
racarrat this rate it will take like 1 hour per image23:12
RAOFAh, vgdb. You're the best.23:20
kgunnracarr: sorry dude23:25
kgunnracarr: i do think you're on to something with the header deltas....23:26
kgunngonna hop off to eat dinner..23:27

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!