| RAOF | racarr_: Enjoy your +1 on cursor-spike. | 00:44 |
|---|---|---|
| AlbertA | looks like the landing ci is stuck? | 01:37 |
| AlbertA | there's 4 MP's ready to land for the last 10 hours and they haven't landed yet | 01:37 |
| kgunn | ack, will see if someone in ci can kick the coke machine | 01:37 |
| duflu | AlbertA: Umm, happy final release eve... ? | 01:38 |
| duflu | That shouldn't block anything though | 01:38 |
| AlbertA | duflu: oh is that what it is? | 01:38 |
| duflu | https://wiki.ubuntu.com/TrustyTahr/ReleaseSchedule | 01:38 |
| duflu | Probably not related. People are usually working furiously on SRU-0/SRU-1 for their various projects around now | 01:39 |
| duflu | which means blocking is unexpected | 01:39 |
| AlbertA | kgunn: so for the non-blocking swapbuffers I believe the best course of action is to avoid | 01:40 |
| AlbertA | kgunn: starting the display buffer compositor threads during display off | 01:41 |
| kgunn | AlbertA: so does that just end up making it an order of operation issue, and can you guarantee display on prior to starting disp-buff-comp ? | 01:43 |
| AlbertA | kgunn: well currently the responsibility lies on whoever is changing the display configuration | 01:44 |
| kgunn | usc ? | 01:44 |
| AlbertA | kgunn: they need to make sure the compositor is stopped before changing the display configuration | 01:44 |
| AlbertA | kgunn: yes usc in this case | 01:44 |
| AlbertA | kgunn: the main issue will be that the compositor will have to have indication of the display power state somehow | 01:45 |
| AlbertA | kgunn: so either it's start method takes a hint parameter | 01:45 |
| AlbertA | kgunn: or the compositor interface exposes some method that makes sense for display power off | 01:46 |
| kgunn | AlbertA: right...i would think you'd want more of a handshake, hint "i need to know when you've actually stopped/started" | 01:46 |
| AlbertA | kgunn: well you want the compositor to only start the buffer consuming thread only | 01:46 |
| kgunn | mmm, yeah, maybe it does... | 01:47 |
| AlbertA | kgunn: the hard part is not exposing those internal implementation details | 01:47 |
| kgunn | well it is the "display" buffer compositor | 01:47 |
| kgunn | so it does imply it can "know" display stuff | 01:47 |
| kgunn | ok...still on irc on my other machine...trying xmir again :) | 01:48 |
| kgunn_ | RAOF, ok...ssh'd into my other machine which is now a black screen on xmir attempt | 01:50 |
| RAOF | kgunn_: Good, good. If you restart lightdm, does anything happen? | 01:51 |
| kgunn_ | one note, early today i reinstalled xserver-xorg-*, dri2 wondering if that would cleanly return any default conf files that might have been hosed....(e.g. apt-get install --reinstall..) is that true ? would it cleanly take care of conf files? | 01:51 |
| kgunn_ | RAOF, so i just restarted lightdm, nada | 01:51 |
| kgunn_ | just black screen | 01:51 |
| kgunn_ | lightdm.log says waiting for signal from xserver, Process 8238 terminated with signal 6 | 01:53 |
| RAOF | SIGABRT is unlikely to end in pleasantness. | 01:54 |
| RAOF | What sayeth /var/log/Xorg.0.log? | 01:54 |
| AlbertA | kgunn_: actually what I've said is already taken care of by https://code.launchpad.net/~afrantzis/mir/expose-display-buffer-only-for-power-mode-on/+merge/214758 | 01:55 |
| AlbertA | kgunn_: I wonder why that's not the case... | 01:55 |
| kgunn_ | RAOF, so it seems to go thru some display config checking & updates...then | 01:56 |
| kgunn_ | just before backtrace its [ 9484.653] (II) Putting surface on output 9 | 01:57 |
| kgunn_ | [ 9484.653] (EE) Backtrace: | 01:57 |
| kgunn_ | [ 9484.653] (EE) 0: /usr/bin/X (xorg_backtrace+0x48) [0x7fd14470ec48] | 01:57 |
| kgunn_ | [ 9484.653] (EE) 1: /usr/bin/X (0x7fd144566000+0x1ac939) [0x7fd144712939] | 01:57 |
| kgunn_ | [ 9484.653] (EE) 2: /lib/x86_64-linux-gnu/libpthread.so.0 (0x7fd143663000+0x10340) [0x7fd143673340] | 01:57 |
| kgunn_ | [ 9484.653] (EE) 3: /lib/x86_64-linux-gnu/libpthread.so.0 (pthread_mutex_lock+0x4) [0x7fd14366d414] | 01:57 |
| kgunn_ | ...is this some stupid "hybris got installed" thing ? | 01:58 |
| duflu | kgunn_: Only if you have hybris installed :) | 01:58 |
| kgunn_ | hmmm, so libhybris isn;t, but ...libhybris-common1 is ? | 01:59 |
| duflu | kgunn_: Yeah apparently that much is safe | 01:59 |
| duflu | And required, for some reason | 01:59 |
| kgunn_ | ok...everytime i see pthread's i think libhybris...just cause of the tegra thing | 02:00 |
| RAOF | Harumph. | 02:02 |
| duflu | kgunn_: When that happens the path name will contain "hybris" so it's OK | 02:02 |
| RAOF | That's *all* the backtrace? | 02:02 |
| duflu | Also it will only apply to lib*GL | 02:02 |
| kgunn_ | the whole thing is https://pastebin.canonical.com/108605/ | 02:03 |
| RAOF | Hm. Died in mir_surface_is_valid? | 02:05 |
| RAOF | So I guess what's happening here is that XMir is getting a null MirSurface* and calling mir_surface_is_valid on it. | 02:10 |
| AlbertA | kgunn_: so it turned out I was not updating the mir platform libraries on my device...:) | 02:11 |
| AlbertA | kgunn_: volume keys work just fine now | 02:11 |
| kgunn_ | ah man! | 02:11 |
| duflu | RAOF: https://bugs.launchpad.net/mir/+bug/1248474 | 02:12 |
| ubot5 | Ubuntu bug 1248474 in Mir "mir_surface_is_valid(NULL) crashes instead of returning false" [Medium,Triaged] | 02:12 |
| RAOF | Another well-known cleanup bug! | 02:13 |
| kgunn_ | this is the part where duflu throws me "that look" | 02:14 |
| kgunn_ | i don't mean to question the experts...but, i'm on a mac/intel w/ integ gfx...wouldn't think it would be a consistent prob | 02:15 |
| kgunn_ | others would've complained ? | 02:15 |
| RAOF | I would have thought so, yes. | 02:16 |
| AlbertA | can anybody else download branches through bzr? Getting connection issues | 02:21 |
| RAOF | Also, how *can* we get a null MirSurface? The callback is called from Surface::created() thusly: callback(this, context); | 02:21 |
| kgunn_ | AlbertA, fails here too | 02:23 |
| kgunn_ | RAOF, so if its really a mir thing, then i should fail to launch a demo_server | 02:31 |
| kgunn_ | right ? | 02:31 |
| RAOF | kgunn_: Probably, yes. | 02:31 |
| === chihchun is now known as chihchun_afk | ||
| kgunn_ | RAOF, hmmm...got majorly distracted, but ok, server seems to launch, but client does fail w a seg fault | 03:13 |
| kgunn_ | and gdb backtrace showing same mir_surface_is_valid | 03:18 |
| kgunn_ | ok...heading to bed, maybe some more tomorrow | 03:21 |
| === chihchun_afk is now known as chihchun | ||
| RAOF | Bah! That's really annoying. | 06:33 |
| RAOF | Stupid race-condition testing! | 06:34 |
| === chihchun is now known as chihchun_afk | ||
| alf_ | greyback: Is this still needed? https://bugs.launchpad.net/mir/+bug/1226778 | 09:42 |
| ubot5 | Ubuntu bug 1226778 in Mir "[unity8 shell] mc::Compositor has no method to get running state" [High,Incomplete] | 09:42 |
| greyback | alf_: I think calling stop() twice doesn't cause a crash any more, so that bug is more a nice-to-have. | 09:43 |
| alf_ | greyback: ok, marking as "wishlist" then | 09:44 |
| === chihchun_afk is now known as chihchun | ||
| === pete-woods is now known as pete-woods-lunch | ||
| === pete-woods-lunch is now known as pete-woods | ||
| === dandrader is now known as dandrader|afk | ||
| kgunn | alf_: AlbertA greyback ...ok, unity-mir, qtubuntu rebuilt... | 14:56 |
| alf_ | kgunn: ok updating | 14:56 |
| * greyback too | 14:56 | |
| alf_ | kgunn: greyback: what do the updates fix/improve? | 14:59 |
| greyback | alf_: the qtubuntu change will have Qt client apps stop rendering as soon as they are off-screen (well, when they get lifecycle events) | 14:59 |
| alf_ | greyback: ah, great | 15:00 |
| greyback | alf_: I have instructions in a comment in https://code.launchpad.net/~gerboland/qtubuntu/surface-visible-hidden-side-channel/+merge/215884 to show how you can test it | 15:00 |
| greyback | I'd appreciate a third party check... | 15:00 |
| === dandrader|afk is now known as dandrader | ||
| kgunn | will update also ...(when/if i can stop the irc pinging) | 15:02 |
| alf_ | kgunn: kdub: racarr_: stand up | 15:03 |
| alf_ | greyback: confirmed that printDate.qml stop requesting buffers, but still keeps printing to stdout with latest silo builds | 15:08 |
| greyback | alf_: that's perfect | 15:08 |
| greyback | alf_: the printing is from the GUI thread, which can continue working unblocked | 15:09 |
| greyback | so the rendering thread has shut down cleanly | 15:09 |
| greyback | (sorry I always fall into the Qt terminology. GUI thread = even handling thread) | 15:09 |
| greyback | event | 15:10 |
| alf_ | greyback: seems so, I am running with MIR_CLIENT_RPC_REPORT=log and see that the client stops requesting buffers | 15:10 |
| greyback | alf_: yes, sounds right | 15:11 |
| greyback | \o/ | 15:11 |
| greyback | \o\ | 15:11 |
| greyback | /o/ | 15:11 |
| alf_ | greyback: exercising? ;) | 15:18 |
| alf_ | _o_ | 15:18 |
| alf_ | \o/ | 15:19 |
| alf_ | _o_ | 15:19 |
| alf_ | \o/ | 15:19 |
| greyback | lol | 15:19 |
| greyback | just happy this is working :) | 15:19 |
| alf_ | greyback: kgunn: hmm, the falling-letters app freezes when you turn the screen of and back on again | 15:39 |
| alf_ | greyback: kgunn: but other apps (e.g. sudoku) work fine, so perhaps that app is at fault? | 15:40 |
| greyback | alf_: confirmed. will investigate | 15:42 |
| === dandrader is now known as dandrader|lunch | ||
| alf_ | greyback: the camera-app freezes too, but the clock doesn't, perhaps it has to do with whether there is some animation/app update going on when the screen is turned off (just wild guess) | 15:45 |
| kgunn | kdub: if you do happen to try xmir...i'd be interested to hear your results...and _after_ confirming it works, if you could try the mir/usc in ppa:ci-train-ppa-service/landing-003 | 15:46 |
| kgunn | that would be interesting | 15:46 |
| kdub | kgunn, i'm set up to try the demo server/shell pretty quickly, was that broken too? | 15:47 |
| kgunn | alf_: greyback ...is there a chance here the apps aren't expecting the expose event ? | 15:47 |
| greyback | alf_: that shouldn't be the case, most animations run on the unblocked thread. But yeah the render thread is blocked for dropping letters, need to figure out why | 15:47 |
| kgunn | kdub: demoshell/client was broken for me yes | 15:47 |
| kgunn | agian...it might be me | 15:48 |
| kdub | i'll try that out real quick.. a few minutes | 15:48 |
| kgunn | greyback: do you know if dropping letters and camera are pure qml ? (i would suspect dropping letters is for sure) | 15:48 |
| greyback | kgunn: it is pure qml | 15:49 |
| kgunn | mmm there goes that theory | 15:49 |
| greyback | so I must have something wrong in qtubuntu | 15:49 |
| kgunn | or bug in qt itself... | 15:50 |
| greyback | I'll assume I'm at fault first :) | 15:50 |
| racarr_ | even if the expose event wasnt working apps should be rendering out of control and not blocking right? | 15:52 |
| racarr_ | (maybe this has evolved since I briefly understood it in london ;)) | 15:52 |
| greyback | indeed. But I've managed to do this in qtubuntu before, it sometimes managed to lie to the renderer and renderer gets locked up | 15:53 |
| racarr_ | -.- | 15:55 |
| alf_ | greyback: in case it is helpful, here is the client rpc log for mir when running dropping-letters: http://paste.ubuntu.com/7262240/ | 15:57 |
| alf_ | greyback: Result id:0 responses are events | 15:57 |
| AlbertA | kgunn: just checked mir/devel on my haswell laptop, demo server/client works well here | 15:58 |
| greyback | alf_: is definitely my problem, render thread not shutting down correctly, need to determine why... | 16:00 |
| kgunn | ta AlbertA | 16:01 |
| kgunn | so when you play video, it keeps playing for about 2 secs when you hit the power button....and then immediately restarts for ~1 second after unlock | 16:25 |
| kgunn | even tho, you cant see it, but stops again, until you bring it into focus | 16:25 |
| kgunn | oops...ok, browser locked up | 16:26 |
| kgunn | greyback: alf_ ^ | 16:27 |
| kgunn | ...i abused it several times before lockup | 16:27 |
| kgunn | (and only the browser, shell & other apps still respond) | 16:28 |
| === dandrader|lunch is now known as dandrader | ||
| greyback_ | kgunn: I've something that I think will fix it. Have just pushed it to the qtubuntu branch. Am building package | 16:32 |
| greyback_ | (locally) | 16:32 |
| kgunn | ack | 16:35 |
| kgunn | dude...watching the "i think we're alone now" trailer over and over is strarting to creep me out | 16:35 |
| racarr_ | ... | 16:39 |
| racarr_ | :p | 16:39 |
| greyback_ | kgunn: could you please test http://people.canonical.com/~gerboland/qtubuntu-android_0.54+14.04.20140402-0ubuntu1_armhf.deb | 16:39 |
| greyback_ | and see if it fixes the blocking issue | 16:40 |
| josharenson | Is there an easy way to change the background color of mir_server_demo_shell? I think I found a bug, and changing the color to, say red, would make identifying it a bit easier. | 16:42 |
| kgunn | greyback_: i will...stepping out for a quick bite | 16:42 |
| kdub | josharenson, search for glClearColor in the examples/ code... i beleive its set to grey there | 16:42 |
| josharenson | kdub: ack | 16:42 |
| greyback_ | kgunn: ok. I'm heading out for dinner, so will be away a few hours. Please email with the verdict | 16:42 |
| kgunn | lunch date running late...lemme see if i can do quickly | 16:45 |
| greyback_ | man the new browser is good | 16:49 |
| kgunn | greyback_: nope.....same experience | 16:49 |
| greyback_ | kgunn: with dropping letters even? | 16:50 |
| kgunn | sorry...just video | 16:50 |
| kgunn | lemme check dropping letters | 16:50 |
| kgunn | cool!...time out worked on video...same little hiccups | 16:51 |
| greyback_ | you play video from the video scope right? Why not working for me... | 16:52 |
| kgunn | yep, i'm playing the "i think we're alone now" | 16:52 |
| greyback_ | ** (process:2021): WARNING **: Unable to dispatch url 'https://videosearch.ubuntu.com/v0/redirect?s=TPzN98RZt0eNHhrYvplj9d3_WtKAbOSC3wun2Q%3D%3D&u=http%3A%2F%2Fvodo.net%2Fitwan':GDBus.Error:com.canonical.URLDispatcher.BadURL: URL 'https://videosearch.ubuntu.com/v0/redirect?s=TPzN98RZt0eNHhrYvplj9d3_WtKAbOSC3wun2Q%3D%3D&u=http%3A%2F%2Fvodo.net%2Fitwan' is not handleable by the URL Dispatcher | 16:52 |
| kgunn | no hang on dropping letters.... | 16:52 |
| greyback_ | ok, improvement | 16:53 |
| greyback_ | though I notice dropping letters not pausing when we switch away from it | 16:53 |
| * greyback_ has to go | 17:00 | |
| === alex_abreu is now known as alex-abreu | ||
| josharenson | if MIR_BYPASS is unset, then bypass is enabled? | 17:09 |
| kdub | 1) that's becoming MIR_SERVER_BYPASS very soon and 2 yes, default is on | 17:20 |
| kdub | its also becoming a --help option soon, so it will be clearer wh at the default is | 17:20 |
| dandrader | "what(): error during hwc prepare(). rc = ffffffff" | 18:16 |
| dandrader | is that known on the N4? ^ | 18:16 |
| kdub | dandrader, not known | 18:28 |
| dandrader | oh, that's surprising | 18:30 |
| dandrader | happens when I turn off/on the display, pressing the power button | 18:31 |
| dandrader | backtrace is useless as usual | 18:31 |
| kgunn | camako: ^ might be an intersting one | 18:47 |
| kgunn | what josharenson found | 18:47 |
| kdub | dandrader, yeah, I know there's some synchronization issues with the compositor/display power on/off being worked on | 18:47 |
| josharenson | kgunn: ? | 18:48 |
| josharenson | kdub: that sounds like what I see... flickering/tearing when starting fullscreen app | 18:48 |
| camako | kgunn: abt bkgrnd color? | 18:48 |
| kgunn | sorry josharenson, meant dandrader | 18:48 |
| josharenson | ah | 18:49 |
| kgunn | camako: sorry, meant the hwc prepare issue | 18:49 |
| kdub | josharenson, well, bypass is only on the mesa platform, is that where you're seeing it? | 18:49 |
| kgunn | as far as digging in code | 18:49 |
| camako | kgunn: o i see | 18:49 |
| josharenson | kdub: nope | 18:49 |
| camako | kgunn: perfect.. right up my alley :-) | 18:49 |
| josharenson | kdub: reason I wanted to change bkgnd color is i see the flickering when unity is running... want to isloate unity/mir | 18:49 |
| camako | dandrader: does it happen every time? | 18:50 |
| josharenson | kdub: working on other things now, but will file a bug when i investigate in a bit (my guess is it will be a dupe of something filed already... I'll read up) | 18:50 |
| dandrader | camako, started happening consistently after I restarted unity8 manually. | 18:51 |
| kdub | greyback and alf_ , they might know something more than me | 18:51 |
| camako | dandrader: let me poke around a bit | 18:52 |
| dandrader | camako, btw, I was using the "qt as mir compositor" branch of unity8, if that makes any difference... | 19:20 |
| camako | dandrader: I see | 19:21 |
| AlbertA | we are hitting some sort of timeout issue in CI for landings | 19:26 |
| AlbertA | https://jenkins.qa.ubuntu.com/job/mir-mediumtests-runner-mako/1116/console | 19:27 |
| AlbertA | https://jenkins.qa.ubuntu.com/job/mir-mediumtests-runner-mako/1117/console | 19:27 |
| AlbertA | kdub: ^ just pinged CI channel they are checking | 19:34 |
| kdub | oh, I just was poking through that, and retriggered the landing | 19:34 |
| AlbertA | yeah me too :) | 19:35 |
| kdub | sometimes that's a transient problem, other times, its a permanent problem | 19:35 |
| AlbertA | kdub: I guess the lander is getting hit hard today | 19:37 |
| kdub | AlbertA, yeah, those 4mps have been waiting a while | 19:39 |
| kgunn | camako: i think dandrader means this... https://docs.google.com/a/canonical.com/document/d/163nyfh_G90nzQnRdI7IYgrMH_0VdmesBju5jpb4wse0/edit | 19:39 |
| camako | kgunn: Thanks for the doc... I am looking through code atm... | 19:41 |
| AlbertA | dandrader: and you sure unity8 is started in nested mode? sounds like it's not | 19:41 |
| dandrader | AlbertA, hmm, that could be it! | 19:42 |
| * dandrader tries | 19:43 | |
| dandrader | AlbertA, you got it. | 19:46 |
| AlbertA | dandrader: cool | 19:49 |
| === chihchun is now known as chihchun_afk | ||
| kdub | i'm being brave | 20:47 |
| kdub | and cleaning up GLRenderer test | 20:47 |
| racarr_ | IN TO THE DEPTHS! | 20:49 |
| anpok | yeah thats an interesting bunch of tests | 20:49 |
| kdub | it was maybe the second test ever written | 20:51 |
| racarr_ | Ok...mostly finished...new ozone-mir (lots of upstream changes, quite helpful ones...just...all the code has to change) | 20:54 |
| racarr_ | feeling kind of exhausted though, so going to switch to fixing this bug I found in introduce-scene-observer (or think might exist, rather the first task is writing a test) | 20:54 |
| racarr_ | and then fiddle with xinput for a while. | 20:54 |
| racarr_ | actually first im going to eat ice cream. haha. brb. | 20:55 |
| kgunn | AlbertA: camako....good grief this took me too long....could you review https://wiki.ubuntu.com/Mir/NonBlockingSwapTesting#preview | 21:59 |
| camako | kgunn: sure... how soon do you need it? | 22:00 |
| kgunn | just look it over for 5 min...for any big idea that maybe i missed | 22:02 |
| kgunn | pretty sure the notes are good...its editable :) so we can fix if it strikes you in 10 minutes not 5 | 22:03 |
| camako | kgunn: looks good to me | 22:10 |
| camako | kgunn: but what do I know... | 22:11 |
| kgunn | camako: thank you mir-expert :P | 22:11 |
| camako | kgunn: at least wording is good and I was able to follow the text... Dunno abt the commands mentioned therein.. | 22:12 |
| camako | kgunn: mir-expert... :-) | 22:12 |
| kgunn | greyback: earlier you said dropping letters isn't stopping...but it is, i think you just witness that a few more letters drop before sigstop | 22:20 |
| greyback | kgunn: right, but there are times that it's not sigstopped, yet it is not in front. Leaving the spread open for instance. | 22:21 |
| greyback | kgunn: oh, would you please kick qtubuntu to build the latest revision | 22:22 |
| kgunn | greyback: you bet | 22:22 |
| greyback | thanks | 22:22 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!