/srv/irclogs.ubuntu.com/2015/07/07/#ubuntu-mir.txt

=== chihchun_afk is now known as chihchun
kenvandineracarr, pong02:34
=== chihchun is now known as chihchun_afk
duflu_racarr: ubuntu-device-flash touch --help03:44
=== duflu_ is now known as duflu
dufluThe non-obvious bit is that --help gives a different answer according to context03:58
racarrduflu: I guess that was the one that I found not to work actually04:02
racarr--recovery-image (which I copied from the PES wiki)04:02
racarrI still dont end up with adb in recovery04:03
racarrand have to fastboot flash04:03
racarrI dunno04:03
racarrthe USB issues make it hard to understand04:03
racarrwhat really happened04:03
racarrI have a cable/port combo that seems steayd now though04:03
=== chihchun_afk is now known as chihchun
=== chihchun is now known as chihchun_afk
=== chihchun_afk is now known as chihchun
anpokRAOF: hrhr05:07
RAOFanpok: Yeah, rad isn't it?05:08
* RAOF just had to finish it over the weekend.05:08
anpokhavent grasped it yet05:09
anpoki spent this weekend restarting the compile time widget tree thing.. and even uploaded it..05:11
=== chihchun is now known as chihchun_afk
=== chihchun_afk is now known as chihchun
RAOFOooh, InputSender. That's what I'm after!05:36
tjaaltonhm no ancell? wonder if the xmir patch is going upstream soon, 1.18 is finally about to freeze05:37
tjaaltonxserver05:37
=== mibofra is now known as Guest47074
=== Guest47074 is now known as mibofra
duflugreyback: Does unity8 always have GL_BLEND on (regardless of surface pixel format)?09:13
greyback_duflu: sorry, wifi dropped09:18
greyback_duflu: it doesn't always have blending enabled. It first draws opaques front to back with blending disabled but depth masking enabled. Then non-opaques are drawn back to front with blending enabled09:20
duflugreyback_: So "opaque" is XRGB as opposed to ARGB09:20
duflu?09:20
dufluDepending on the surface pixel format09:20
greyback_duflu: anything in the qml not needing an alpha channel09:21
greyback_the surface format for mir, we've probably chosen XRGB09:21
greyback_mainly because it'll always work. But we're working on more intelligently selecting the mir surface format, to improve perf09:22
duflugreyback_: I've got hints that Unity8 might be blending XRGB as if it was ARGB. Admittedly Mir demo servers have the same bug. I was just wondering if we're all affected09:22
greyback_duflu: is possible, unity8 does the same thing as mir in that respect09:22
duflualf_: Feel like chatting today?09:36
alf_duflu: coming09:37
mcphailgreyback_: do you think the blending contributes to this bug? : https://bugs.launchpad.net/ubuntu/+source/libsdl2/+bug/146014910:22
ubot5Launchpad bug 1460149 in libsdl2 (Ubuntu) "Visible corruption in SDL apps (Neverball, Neverputt) on Nexus 4 / Nexus 7." [High,Triaged]10:22
greyback_mcphail: not directly, but I suspect the sdl's surface may be using a mir surface format with an alpha channel, making unity8 blend it. Would have to dig to learn more10:28
mcphailgreyback_: just interested as to why the aberrant blending clears when you swipe the app away10:28
greyback_/usr/lib/gcc-cross/arm-linux-gnueabihf/4.9/../../../../arm-linux-gnueabihf/bin/ld: ../../gmock/libs/gtest/libgtest.a(gtest-all.cc.o): relocation R_ARM_THM_MOVW_ABS_NC against `__stack_chk_guard' can not be used when making a shared object; recompile with -fPIC12:44
greyback_../../gmock/libs/gtest/libgtest.a: error adding symbols: Bad value12:44
greyback_^^ I get that building mir in my arm Xcompile chroot. Would libgtest.a not be compiled with -fPIC?12:45
=== chihchun is now known as chihchun_afk
kdubgreyback_, libgtest.a seems to be compiled with fPIC from the mir build system13:39
kdubalf_, was curious, is it the mg::DisplayBuffer/mg::Display system that has the gl stuff baked in the most? or is there somewhere else13:40
greyback_kdub: yeah. Am still confused13:40
kdubgreyback_, using sbuild?13:41
greyback_kdub: yes13:41
* kdub did that a day or two ago, seemed okay, after updating with -udacr13:41
alf_kdub: Excellent timing :) I have started a document with a tentative plan about removing gl dependencies (including where these dependencies are in the code). Will post a link of a first draft before EOD today.13:42
kdubalf_, cool, I just felt compelled to point out that mg::DisplayBuffer::post_renderables_if_optimizable is really not posting anymore13:47
kduband mg::DisplayBuffer is mostly about content, which seems good for abstracting the gl bits out13:48
=== chihchun_afk is now known as chihchun
=== Olli_ is now known as olli
=== dandrader is now known as dandrader|afk
=== mibofra is now known as Guest83710
=== Guest83710 is now known as mibofra
=== chihchun is now known as chihchun_afk
=== dandrader|afk is now known as dandrader
kenvandineracarr, you pinged me last night?17:21
=== dandrader is now known as dandrader|bbl
racarrkenvandine: Oh hey sorry yeah I think I've since resolved my question17:57
racarrI was trying to narrow down what may be triggering the text input failure...because17:57
racarrjust uh17:58
racarrLike making an autopilot keyboard object from python17:58
racarrtyping17:58
racarrdoing stuff17:58
racarrtyping doing stuff17:58
racarrseemingly infinitely never triggers it17:58
racarrs/seemingly infinitely/for a few minutes/17:58
racarrSo I was trying to find a collection of tests to trigger it but I have a few now...17:58
racarrits hard to find the common element the17:58
racarrlol17:58
kenvandineyeah17:58
kenvandinesucks17:58
racarrthe autopilot tests aren't easy to read17:58
kenvandinenope17:58
racarrbecause theres so much inheritance of implementation17:58
racarrIts like17:59
racarrI dunno17:59
racarrI tried tracing base classes up and17:59
racarrbutttt17:59
racarrhard to isolate anything17:59
racarrso my new strategy is instrument USC :)17:59
kenvandinei don't think it has anything to do with the system-settings autopilot tests17:59
kenvandinewell17:59
racarrwell...it may have something to do with autopilot though17:59
kenvandinewe're using the keyboard input provided from the uitk i think17:59
kenvandineyeah18:00
racarrlike I said with just the fake uinput thing18:00
racarrI can't trigger it18:00
kenvandineit's triggered by something autopilot is doing18:00
kenvandinealso... i think it's more than just the fake keyboard18:00
kenvandinethere is a fake mouse too for touch events18:00
kenvandinealso uinput18:00
racarrthat stops working?18:00
racarryes so I thought18:00
kenvandineyes18:00
racarrthis was all related18:00
racarrto uinput18:00
racarrwhich according to the bug is what is being used in autopilot now18:00
racarrand it seems that the problem is18:01
kenvandineyeah18:01
racarrthe unity8 session loses focus18:01
kenvandineand mir starts discarding the events18:01
racarrloses keyboard focus18:01
racarrmm18:01
kenvandinefrom what dandrader said18:01
racarrYeah I confirmed that unless im only seeing part of the bug18:01
racarrI haven't seen any unity8 crashes for example so far18:02
racarrand don't understand where that comes in18:02
kenvandineyeah...18:02
kenvandinethat wasn't related18:02
kenvandinei don't think18:02
racarrok :)18:02
kenvandinewell, we saw the same unity8 crash everytime we had run into this18:02
racarryeah that was my other question.18:02
racarrah18:02
kenvandinebut the crash stopped18:02
kenvandineand this kept happening :)18:02
kenvandineso unrelated18:02
kenvandineit was a distraction from the real problem18:02
kenvandineracarr, it affects at least webbrowser-app and ubuntu-system-settings18:03
kenvandinebut likely more too18:03
kenvandinewe just have long autopilot tests18:03
kenvandinei think we've had the problem intermittently for a while, we used to have random failures where it seemed to stop getting the uinput events18:04
kenvandinetouch stuff18:04
greybackcrazy idea: could there be a test pressing down a meta key, and crashing before that key is released?18:04
kenvandinebut re-running it would pass18:04
kenvandinenow it's consistent18:04
kenvandinei haven't had a passing CI run in settings in 2 weeks18:04
kenvandinegreyback, i don't think so18:05
greybackworth a shot :)18:05
kenvandinei'm pretty sure i've hit it when tests that didn't even use the keyboard18:05
racarrgreyback: That's a good idea but I think it wouldn't explain the focus behavior18:05
kenvandinejust the fake mouse events stop18:05
racarrwhich im 99% sure isn't a red herring18:05
racarrwait the fake mouse events stop oO18:06
kenvandinemakes sense though18:06
kenvandinebecause of focus18:06
=== racarr is now known as racarr|lunch
=== dandrader|bbl is now known as dandrader
racarr|lunchyay just catching up on MLs...anpoks libinput changes are landing :)19:32
=== racarr|lunch is now known as racarr
anpokhum .. time to say goodbye19:36
anpok.. to those hp/webos usb cables19:37
racarrkenvandine: After upgrading arale to r50 from19:54
racarrubuntu-touch/rc-proposed/meizu.en19:54
racarrThe test I was using autopilot3 run ubuntu_system_settings.tests.test_search.SearchTestCases.test_search_filter_results19:55
racarrdoesnt fail anymore after like 25 times19:55
kenvandineoh?19:55
racarrI was on either 48 or 49 yesterday19:55
kenvandineany idea what change fixed it?19:55
racarrI Was wondering 1. Could you suggest some other tests. 2. Can you try upgrading as well?19:55
racarrNo...lol...19:56
kenvandinei mostly care about wily, since that's what CI runs on19:56
kenvandineracarr, try running the full suite19:57
racarrkenvandine: Running :)19:58
racarrI don't think I've ever put wily on my uh19:58
racarrarale...is it19:58
racarrubuntu-touch/devel/meizu.en?19:58
kenvandinenot sure off hand, i have it on my mako19:58
* kenvandine checks channel19:59
kenvandineubuntu-touch/devel-proposed/ubuntu19:59
kenvandineracarr, that's the channel i have on my mako19:59
racarrok thanks :) anyway full suite running now20:00
kenvandineracarr, ubuntu-touch/devel-proposed/meizu.en20:00
racarrchanging laundry...uh20:00
kenvandineis for arale20:00
kenvandineracarr, thx20:00
racarrthen I guess I can switch to latest wily20:00
racarrand see whats going on20:00
racarrI was just trying to test with an20:00
racarrinstrumented USC to try and understand20:00
racarrbut then before I was like ok first20:00
racarrclean image and reproduce20:00
racarrand now I can't reprouce anymore lol20:00
kenvandinemight be a good sign :)20:01
racarrprobably but not understanding is frustrating...20:01
kenvandinei'm updating my mako and will try the tests too20:03
racarrkenvandine: Aghhh I managed to trigger it...20:16
racarrI guess it was just being elusive20:16
racarr:/20:16
kenvandineok20:16
racarris there a way to work with the citrain on vivid + overlay without20:34
racarrwriting apt pins, etc...?20:34
anpoki think you no longer have to20:36
anpokracarr: the current phablet tools should already do that20:37
anpokat least after i used an unbroken cable.. a current image from the right channel.. it finally worked20:37
kgunnyeah robru said you shouldn't have to pin with overlay20:46
racarrhmm20:46
racarrI have a plausible theory about the text input bug now!20:55
racarr1. An autopilot test eventually fails because that happens with the timeouts and stuff causing mirscreencast to be invoked20:55
racarr2. Bug in surfaceless session management with mirscreencast20:56
racarrconnecting and disconnecting20:56
racarrcauses20:56
racarrfocus to be cleared when20:56
racarrmirscreencast closes20:56
racarrinstead of20:56
racarrunity820:57
racarrI kind of see a plausible path in the code but its hard to tell whats happening because20:57
racarrum20:57
racarras far as I can tell there are two window managers in USC20:57
racarrlol20:57
racarrWindowManager seems to take care of like focusing sessions when sessions are removed and stuff like that20:58
racarrbut then so does SessionSwitcher oO20:58
racarranyway...I feel good about this theory but I'm still struglging to test anything20:58
racarrI needed to install the mir 0.14 ppa to build USC trunk with my instrumentation on vivid+overlay of course20:58
racarrbut mir 0.14 ppa did not work on vivid+overlay for me20:58
racarr(u8 crashes after 3-4 seconds)20:58
racarrso now I'm trying to test with wily...20:58
racarrin particular usc::WindowManager l115 auto const next_session = session_coordinator->successor_of({});21:00
racarrI can't see why that would be the right value...21:00
racarrhmm and this code is in mir now in system_compositor_window_manager21:02
racarrwith some modifications21:02
racarrtoo bad about no Alan...21:02
racarrInteresting ok21:03
racarrI guess the SurfaceReadyObserver21:03
racarrmay be bale to prevent screencast from stealing focus...21:03
racarr(a new introduction to the logic when it was imported)21:03
racarrum so both vivid and wily+overlay21:34
racarrthe mir 0.14 ppa21:34
racarrunity8 crashes within seconds on arale21:34
racarrhas anyone else tested it?21:34
anpokthe ppa is incomplete.21:35
anpoki was waiting for platform api and qtubuntu to land21:35
racarroh21:36
racarranpok: When is that expected?21:37
anpokit just happened for wily21:38
anpokit now needs to be synced to vivid+overlay... but at least the lp branches are updated.. so I guess I can build it..21:39
racarranpok: :D That would help me21:39
kenvandineRan 135 tests in 2750.871s22:32
kenvandineOK22:32
kenvandineracarr, ^^ on my mako with wily22:32
kenvandinethat hasn't happened in 2 weeks22:32
kenvandinemaybe it was a fluke... or maybe something's been fixed?22:33
racarrkenvandine: So uh...yeah I also got22:40
racarrwell right I ran that one test like 30 times and22:40
racarrthen saw dozens in the test suite complete22:40
racarrbefore failing22:40
racarrso its possible the race has just gotten22:40
racarrway tinier lol...22:40
racarrthough, right I still haven't tested wily! my result was from vivid + overlay22:41
RAOFracarr: With your input sending reworkything, were you thinking of having mf::Surface have a send_event method?23:53

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