/srv/irclogs.ubuntu.com/2014/04/01/#ubuntu-mir.txt

=== jono is now known as Guest60170
RAOFHey, would anyone have any objection to putting an event_time member in the MirEvent union?05:54
racarrRAOF: I tried to think of reasons to object and cant06:09
dufluRAOF: I have wanted that myself too. Just feared the necessary client ABI break. Maybe first thing next cycle?06:28
dufluAlthough we're still at a point where breaking the client ABI isn't a huge deal06:29
RAOFduflu: I'll queue it up; I'd like it for some testings.06:31
dufluVotes wanted: https://code.launchpad.net/~vanvugt/mir/frontend-server/+merge/20760206:58
RAOFduflu: Given that we agree that the code itself is bad, is renaming it worthwhile?07:14
dufluRAOF: Yes I think so. Less bad :)07:14
RAOFThat's my 2ยข; it's after 6 here, so tootleooo!07:14
dufluRAOF: Night07:14
=== chihchun_afk is now known as chihchun
=== chihchun is now known as chihchun_afk
alan_ganpok: has you "Needs Information" been answered? https://code.launchpad.net/~alan-griffiths/mir/introducing-SurfaceObserver/+merge/21343110:26
anpoknot really, but looking at the question it is not something that we can answer now..10:34
anpokusers will tell us10:35
anpokalf__: final opinions on https://code.launchpad.net/~andreas-pokorny/mir/no-initial-display-configuration-sent-to-hosting-server/+merge/213126?10:44
alf__anpok: I think creating the surfaces with the changed configuration is problematic10:47
alf__anpok: without applying the configuration10:47
alf__anpok: for example, what if the changed configuration says two outputs, but the host configuration is a single output, or outputs of different sizes etc?10:49
alan_galf__: there are already problems in that area. Try running a host as sidebyside and a nested render_surfaces with the defaults10:54
* alf__ tries10:56
anpokhm tried the new greeter without applying the policy10:58
anpokit still works10:58
anpokit also seems to still get a transparent surface10:58
alf__anpok: is this on android?11:01
anpokyes11:02
alf__anpok: it could be that the host configuration pixel format is RGBA (instead of RGBX)?11:02
anpokah yes the first (default) one it is..11:03
alf__anpok: alan_g: I think that since we are not using the initial display configuration for nested any more, we should change to use other information. For example a bool prefer_translucent_formats?11:07
alan_galf__, anpok: I don't know the "right" answer - we have a design choice as to how the control (and any negotiation) splits between the host and nested systems.11:09
anpokwe could also look at the other side - should focus changes apply the complete configuration?11:12
alf__anpok: alan_g: My understanding is that we have made the decision that the nested server won't touch the host configuration by default. This also means that we need to keep it in sync with it.11:14
anpokalf__: what if the nested seat only wants to use a single diplay, while the other display should be reserved for another user?11:14
alan_gMy understanding is that we have made the decision that the nested server won't touch the host configuration *on startup*11:15
anpokbut it is still alowed to make changes to that configuration11:15
alf__alan_g: anpok: right, on startup, that's what I actually meant11:15
anpoki.e. we want full screen applications run with different modes?11:15
anpokok11:15
anpokand that I consider to be only a band aid11:15
anpokshort lived..11:16
anpokbecause this discussion will take a few roundtrips and will lead to some rework..11:16
anpoki assume11:16
alf__anpok: band aid == proposed MP?11:16
anpokyes11:16
anpokjust what split greeter needs for now11:17
alf__anpok: alan_g: I am just worried that we are pushed into a lot of band aid solutions lately, and gathering tenchnical debt.11:17
alan_galf__: agreed. (Although I'm trying to fix the Surfaces debt ATM)11:18
alan_ganpok: are you able to drive a discussion about nesting and display configuration? E.g. Write down some key scenarios?11:19
anpokok11:20
alan_gIn that case, how about circulating those and scheduling a hangout?11:21
alan_galf__: We ought to start putting debt items where we (and especially kgunn) will be reminded of them. I know some are in the bugs.11:22
anpokalan_g: google drive document?11:23
alf__alan_g: anpok: [technical-debt] bug tag?11:23
anpokfor the scenarios to discuss..11:23
alan_gWhatever works11:25
anpokalf__: so yes both ways turn out to provide a transparent surface format, so no need to call initial policy inside that (temporary) fix11:26
alf__anpok: both ways?11:26
anpokyes, with and without applying the initial policy in the greeter and unity8 nested shells11:28
alf__anpok: that will not work on mesa though, because it uses an rgbx format for the host server11:31
=== alan_g is now known as alan_g|lunch
anpokcannot test, thought the array of supported formats might also be led by a transparent format..11:53
=== chihchun_afk is now known as chihchun
kgunngreyback: alf__ got to thinking in my slumber...it might not matter what mir does wrt egl, since qt is interfacing directly, we can effect the u-s-c12:02
kgunnbut i don't even think we'll be able to effect the session compositor12:03
kgunnesp with our effort to plug in qt sg12:03
anpokalf__: you are right12:03
greybackkgunn: you lost me. Can you rephrase please?12:04
kgunngreyback: egl is a hw vendor specific implementation (unfortunately eglswapbuffers is up for interpretation)12:07
kgunnso whomever is interfacing with egl directly can be effected by those choices, and blocking might be occuring anyway12:07
kgunnor "other errors"....context lost could be one12:08
greybackkgunn: I see what you mean12:08
kgunngreyback: i was thinking mir could arbitrate for qt, but that's only really true for the compositor...not the the apps12:09
kgunnright ?12:09
kgunnand even the compositor...we're looking at using qt sg12:09
anpoki thought we can decide about the behavior of eglswapbuffers in every system mir will be shipped?12:13
kgunnanpok: we can decide as long as we're the ones dealing with swapbuffers, but if the app is native (which think of toolkits as native gl apps)..they're dealing with egl directly12:14
greybackI guess there's no way to prevent that12:15
alf__kgunn: We are implementing the EGL platform though, so we do have at least some control of what's happening, although perhaps not full control.12:16
kgunnalf__: the egl platform, but not the egl driver per se12:16
kgunnright ?12:16
greybackand since it is vendor specific, there's little we can do other than document that we recommend people use Mir's client API, which does what we want12:16
kgunnalf__: anpok greyback happy to be proven wrong...12:17
=== chihchun is now known as chihchun_afk
kgunnmorning duties, bbiab12:18
=== alan_g|lunch is now known as alan_g
=== alan_g is now known as alan_g|tea
=== alan_g|tea is now known as alan_g
=== dandrader is now known as dandrader|afk
=== dandrader|afk is now known as dandrader
dandradershould I feel bad about casting a shell::Surface to a scene::Surface?16:06
alan_gdandrader: maybe16:07
dandraderon the QPA, MirSurfaceItem has a shell::Surface and it uses the shell:Surface::lock_compositor_buffer()  hack16:07
alan_gThe usual rule is that "if you destroy type information you can restore it".16:07
dandraderbut that hack is to be removed in favor of the new scene::Renderable::buffer16:08
dandraderwhich is implemented only by scene::Surface16:09
alan_gI think it would probably be better to have a scene::Surface in the first instance16:09
alan_gBut I've not looked at what that involves16:09
dandraderthe qt compositor QPA is getting surfaces from shell::SessionListener::surface_created16:09
dandraderwhich gives shell surfaces, not scene surfaces16:10
dandradernaturally16:10
alan_gdandrader: yeah, that hasn't been fixed (yet)16:10
alan_gI'm still fixing scene::Surface et al. scene::Session comes next16:11
dandraderalan_g,  hmmm, so in the meantime I can just do the cast from shell::Surface to scene::Surface because later mir interfaces will change in a way that I won't have to do it anymore?16:11
alan_gdandrader: there ought to be a scene::Session and a scene::SessionObserver and we'll obsolete the shell stuff16:12
dandraderalan_g, ok, sound good16:12
alan_gAt least that's the current "vision"16:12
alan_g*that shell stuff (there's still a role for shell)16:13
=== alan_g is now known as alan_g|EOD
=== dandrader is now known as dandrader|lunch
dansufHello once again. I still have got this error in mir that it fails to bind buffer to texture and also in logcat W/Adreno200-EGL( 1576): <qeglDrvAPI_eglCreateImageKHR:4526>: EGL_BAD_PARAMETER19:10
dansufIs there any possibility to see more exactly what is the problem?19:11
dansufSurfaceflinger also doesn't work so it may be also something device-specific, but I still don't know what.19:13
kgunndansuf: judging from this...19:40
kgunnhttps://bugs.launchpad.net/mir/+bug/127204119:41
ubot5Ubuntu bug 1272041 in mir (Ubuntu) "mir nested server failure: what(): error binding buffer to texture" [Critical,Fix released]19:41
kgunnmost likely a color buffer mismatch ?19:41
dansufkgunn: i don't have this unsupported buffer format line19:41
kgunnAlbertA: ok..so, i'm a little confused...19:43
AlbertAkgunn: ?19:43
kgunnso looking at qtubuntu...i 1/2 expected to find some code19:44
kgunnthat said something like qt->mir.swap_buffers, that would in turn call the eglSwapBuffers....19:44
kgunnbut instead...i only find19:44
kgunnqtubuntu/src/platforms/base/context.cc19:45
kgunncalling it direct...19:45
kgunnam i searching wrong?19:45
dansufkgunn: there's this bad_match like in the bug report, bad_context in eglSwapInterval and also19:45
dansufW/Adreno200-ES20( 3349): <get_simple_queries:1360>: GL_INVALID_ENUM19:45
dansufkgunn: this one appeared amongs other errors when running lightdm19:46
AlbertAkgunn: no I think that's expected19:47
AlbertAkgunn: because then the egl will foward the call to the native window type19:47
AlbertAkgunn: which is where the mir platform will get hooked on19:47
kgunndansuf: so in all this instance are you using the mir qt plugin ?19:47
dansufkgunn: I'm not sure but if you mean if the variable QT_QPA_PLATFORM contains ubuntumirclient then yes19:49
=== dandrader|lunch is now known as dandrader
kgunndansuf: can you share your exact device name once more ? i recal you said Sony....but model #?19:54
dansufkgunn: live with walkman, a'ka coconut19:55
dansufkgunn: but I'm not using official cyanogenmod repos as they dropped support for the device19:55
kgunndansuf: so where do you get your gpu drivers from ?19:58
dansufkgunn: I got gpu drivers from adreno's site19:58
kgunndansuf: so binary i assume ? your not building from source?19:59
dansufkgunn: generally, w-flo who ported desire z also did an update of his drivers and in a commit wrote it helped him with some issues19:59
dansufkgunn: yes, a binary19:59
kgunndansuf: so this is only a guess...but you might be careful of binaries compiled on differing kernel & android versions vs what you have...e.g. do you know which android version those were compiled against ?20:00
dandraderHave you guys seems this error before? http://paste.ubuntu.com/7191321/ this is a symbol from libboost_program_options.so.20:01
kgunnAlbertA: so even tho it looks like qtubuntu calls eglSwapBuffers directly...this call actually gets rerouted ? (i don't see how)20:01
AlbertAkgunn: right because mir is the one providing the egl native display20:02
dandraderso it seems libmirplatformgraphics.so should have a depencendy on it but ldd says there's none..20:02
dansufkgunn: adreno says its for Android 4.2 Jelly Bean MR1 and also in logcat there's some info that suggests it was compiled for mako20:04
dansufkgunn: and before w-flo updated his drivers he needed to make changes in mir's source to make it working20:05
kgunndansuf: hmmm sounds fishy (no pun intended :)20:05
kgunndandrader: are you changing mir config? or is it bombing out even on the default ?20:06
dandraderkgunn, you mean the DefaultServerConfiguration class?20:08
kgunngotta run...20:10
kgunnkdub: ^ can you assist dandrader20:10
kdubyou mean dansuf?20:11
kgunnkdub: dandrader20:11
kgunndansuf: i think we're on to android 4.4.2...dunno if that might be an issue20:11
kdubokay :) lets see20:11
dansufkgunn: we're on cm10.1 which i believe is 4.220:12
dansufkgunn: 4.4 is kitkat, the newest20:12
kdubdandrader, there was a MP that added a boost dependency to mirplatform graphics20:12
dandraderkdub, still under review?20:13
dansufkgunn: anyway, thank you20:13
kdubdandrader, merged yesterday20:14
kdubto devel20:14
dandraderhmm20:14
* dandrader checks20:14
dandraderbecause I just rebased my stuff on top of latest devel...20:14
kdubdandrader, although, that looks more like a build time problem that was being averted20:14
kdubyours looks a bit different20:15
kdubdandrader, but I don't see the packages listing program-options as a dependency though too20:19
kdubdandrader, does installing program-options fix the error?20:20
dandraderkdub, no it was already installed. I'm now trying this http://paste.ubuntu.com/7191409/20:21
dandraderkdub, but I have  a feeling I might have messed up my installation somehow....20:21
dandraderbecause the problem only happens when running unity8. mir examples work fine....20:23
kdubdandrader, i'll see if I can reproduce the problem today20:23
racarrdandrader: I ran in to this21:03
racarrI worked around it by LD_PRELOADING libmirserver21:03
racarrbecuase it wasnt ust boost option stuf (LD PRELOADING boost options give21:03
racarrother errors)21:04
racarrhoping the21:04
racarrinflight MP would fix it21:04
racarrI cant figure out what caused it to happen though so dont really understand21:04
racarrmaybe libmirplatform graphics needs to be explicitly linked against libmirserver...I dont think it is? but since we load it dynamically its been fine21:05
racarrand that stopped working because <some explanation goes here>21:05
dandraderracarr, true. After I added boost options I get miroptions missing symbols. after I also add miroptions I get missing _ZN3mir21default_server_socketE and so on21:05
dandraderwill try your workaround now21:06
racarrI cant remember in the end if I ended up LD_PRELOADING libmirserver or linking in CMake but either should work21:06
dandraderracarr, yeah LD_PRELOADING libmirserver works!21:06
racarrdandrader: Ok ill send an email to the list or make an MP or something in ust a few thanks.21:08
racarrI originally thought maybe it was ust an isolated issue on my phone21:09
racarrbecause of something dumb I did :p21:09
racarrabout 9/10 times its that.21:09
racarr17 conflicts encountered.21:11
racarr...lovely21:11
kdubI'll look into that program options today, getting my stuff out the door21:16
=== popey_ is now known as popey
kdubdandrader, racarr is there a bug about that program options stuff? (if not, ill file one)21:55
dandraderkdub, not that I'm aware of21:56
kdubokay, i'll file21:56
dandraderkdub, thanks!22:01
kdubno problem22:01
kdubhey racarr, if you can still repro that symbol problem, I posted a fix here https://code.launchpad.net/~kdub/mir/fix-1301040/+merge/21373922:30
racarrkdub: Not sure if I can reproduce it still...will check in a sec...should be able to22:31
racarrI dont think that will fix it though because22:31
racarrif you LD_PRELOAD boost options22:31
racarryou then go on to get missing symbols in mir options22:31
racarri you link mir options as well22:31
racarryou go on to get missing symbols in default server conf22:32
kdubracarr, ack, i'll keep seeing if i can get a build22:57

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