/srv/irclogs.ubuntu.com/2015/02/12/#ubuntu-mir.txt

rsalvetianyone from the mir team around still?00:55
rsalvetikgunn: ^00:55
rsalvetiquite a lot of packaging changes landed with the new mir00:55
rsalvetibreaking the vivid touch build as we're depending on a package that got renamed00:55
rsalvetijust trying to understand the changes00:55
racarrrsalveti: Hi01:29
racarrrsalveti: I think im the only one around so ill give a best effort01:29
racarrrsalveti: Wait RAOF is around and he made the changes01:29
racarrRAOF: ^01:29
racarrerr01:29
rsalvetihe's not around01:30
rsalvetiit seems the backend can be dynamically loaded now01:30
racarrrsalveti: He is hes just not in this channel I pinged him on mir01:30
rsalvetihttps://code.launchpad.net/~mir-team/mir/server-platform-probing/+merge/24498201:30
tmpRAOFrsalveti: Sure.01:30
racarrrsalveti: Yes but thats basically what is it so01:30
rsalvetitmpRAOF: hey :-)01:30
racarryes so the soname version isnt required anymore01:30
tmpRAOFracarr: You need to highlight tmpRAOF now, as freenode ate my RAOF registration :)01:30
racarrand the libraries got renamed01:30
rsalvetitmpRAOF: our vivid build is broken because it is still depending on the older packages01:30
racarris the summary of my understanding01:30
racarrtmpRAOF: :)01:30
rsalveti-add_package install ubuntu-minimal mir-client-platform-android libmirplatform5driver-android ubuntu-touch01:31
rsalveti+add_package install ubuntu-minimal mir-client-platform-android mir-platform-graphics-android ubuntu-touch01:31
rsalvetiI changed this01:31
rsalvetibut we also had the update-alternatives for it01:31
rsalvetiand it seems it's not required anymore01:31
tmpRAOFCorrect.01:31
rsalvetitmpRAOF: how is mir finding out the backend to load now?01:31
tmpRAOFBy loading them all and finding the one that works.01:31
tmpRAOFThe only reason not to have mir-graphics-drivers01:31
tmpRAOFThe only reason not to have mir-graphics-drivers-mesa on the touch images is you'll waste some disc space.01:32
rsalvetitmpRAOF: right, that's something I'll check on the next build, in case we get that package installed for whatever reason01:32
rsalvetitmpRAOF: but in case both are installed at the same time, how to select which one to use?01:32
rsalvetinot sure if mesa will indeed fail on touch01:33
tmpRAOFIt will.01:33
rsalvetigreat, will remove the update-alternatives and trigger a new build then01:33
tmpRAOFUnless it can *actually* work, in which case it's fine to use :)01:34
racarrlol01:34
racarrin which it turns out mesa worked on nexus devices this whole time01:34
rsalveti:-)01:34
racarr;)01:34
rsalvetiwell, it might work with software rendering01:34
rsalvetinot sure how is the current support though01:34
tmpRAOFNo, our mesa platform doesn't support software rendering.01:35
tmpRAOFAlso, you'd need a kms driver.01:35
racarrMm I think we are covered...because of the modesetting bit01:35
racarrand then even if one did appear the platforms get "priorities"01:35
tmpRAOF(And once we *do* support software rendering, the mesa platform will only load with software rendering if there's nothing better; android will be better in that case)01:35
racarrand if an android driver is available (dont remember exactly how it works, but something like...can we load libhardware and open gralloc or whatever)01:36
rsalvetiright, awesome then01:36
racarrthen it gets priority "Best"01:36
tmpRAOFrsalveti: So, the only package you should need is mir-graphics-drivers-android (conversely, mir-graphics-drivers-mesa for desktop).01:36
rsalvetiright, that's already in our seeds01:36
tmpRAOFracarr: That's correct - android probes by trying to open gralloc.01:37
racarrinterestingly (not to this conversation, just to people interested generically in EGL implementations) mesa provides a DRI based implementation of gralloc01:39
racarrfor running android on intel free drivers01:39
racarrkdub and I discovered it before hybris appeared as a driver solution and it took an hour or so of like...does this do what we need? Wait no this is01:39
racarrthe exact opposite...01:39
tmpRAOFrsalveti: So, you should be able to drop mir-client-platform-android and keep just mir-graphics-drivers-android. We guarantee not to rename that one, and it'll always depend on a full set of packages required to get a fully functional Mir drivers on android.01:39
racarrit took me an hour or so maybe kdub understood because he knew gralloc lol01:39
rsalvetitmpRAOF: yeah, that's the change I did01:40
rsalvetiwill trigger a new image in a few and let you konw01:40
tmpRAOFSweet.01:41
TrevinhoMhmmh... I've just updated the gdk-GL mir backend to match upstream changes... and well, with then now after few seconds I run the test-gl I'm getting a kernel panic on MIR! :o02:36
Trevinhogdk changes are basically:02:40
Trevinhohttps://bugzilla.gnome.org/show_bug.cgi?id=741946 plus the relative changes to the mir backend...02:40
ubot5Gnome bug 741946 in Widget: GtkGLArea "OpenGL context should allow for GL attribute selection" [Normal,Resolved: fixed]02:40
Trevinhobasically things have been moved to gl 3.2 APIs by default... But Not sure how I can help to fix this... I'd say it's mostly a driver issue....02:42
duflu_Trevinho: What requires GL 3.2?03:01
Trevinhoduflu_: gdkgl03:01
=== duflu_ is now known as duflu
dufluThat's unfortunate. We still have Ubuntu users limited to GL 1.x which I aim to support with Mir. It would be a pity if they just can't run GTK03:02
Trevinhoduflu: they can run gtk, but not gtk-gl apps03:02
dufluAh OK03:03
Trevinhoduflu: they supported legacy gl till few days ago, then they removed that as it was overcomplicating stuff for them03:03
dufluTrevinho: P.S. Sleep03:03
Trevinhoduflu: so basically now in gtk you can create a gl context inside your app and make it interact with the rest03:04
dufluTrevinho: I know. I've written toolkits with GLAreas :)03:04
Trevinhoduflu: I was about to go, then I got a friendly ping :)03:04
Trevinhoduflu: yeah, stuff like that... arrived just a little late...03:04
TrevinhoI mean, instead of putting such the efforts in clutter, maybe....03:05
tmpRAOFTrevinho: Yeah, kernel panics are very much Not Our Fault™03:08
Trevinhoyes, I was wondering that...03:08
TrevinhoI've not them in X, but I guess it's due to the fact it's using a different driver03:09
tmpRAOFIt's not really using a different driver. Unless you're using the nvidia binary drivers for X and nouveau for Mir (or fglrx/radeon)03:10
Trevinhomh, no I'm using radeon currently... But I thought it was using an egl driver at mesa level, isn't it?03:10
tmpRAOFYeah; GTK doesn't use EGL on X as well?03:11
Trevinhoglx03:12
tmpRAOFThere certainly are differences in the uses, though. If you could capture the panic it might point to something we obviously do differently and possibly wrong.03:12
tmpRAOFHuh. Fair enough.03:12
Trevinhowell, I currently I've no clue how to debug that kind of panic... 😢03:12
Trevinhoprobably I need to ping someone with hands at lower levels than me :)03:14
* Trevinho also can't build current mir without http://pastebin.ubuntu.com/10182012/03:14
tmpRAOFYeah, we've got some weirdness with our headers at the moment.03:14
Trevinhomaybe it works with pre-compiled headers, but I guess I've disabled them03:15
rsalvetiGet:335 http://ftpmaster.internal/ubuntu/ vivid/main mir-platform-graphics-mesa i386 0.11.0+15.04.20150209.1-0ubuntu1 [91.6 kB]04:48
rsalvetiGet:336 http://ftpmaster.internal/ubuntu/ vivid/main mir-platform-graphics-android i386 0.11.0+15.04.20150209.1-0ubuntu1 [82.4 kB]04:48
rsalvetitmpRAOF: both were installed, will check the image in a few04:48
rsalvetinot sure why we're still getting the mesa package04:48
=== chihchun_afk is now known as chihchun
rsalvetilibmirserver29 depends on it04:49
rsalvetias it's |, it seems it will always get the first04:49
rsalvetieven when my seeds is forcing the android one04:50
tmpRAOFrsalveti: Garh, sorry.05:02
tmpRAOFrsalveti: I fixed that for the client platform earlier, but at that point the server loading bit hadn't landed.05:03
rsalvetiright05:03
tmpRAOFrsalveti: For what it's worth: https://code.launchpad.net/~mir-team/mir/server-doesnt-depend-on-drivers/+merge/24944905:07
rsalvetitmpRAOF: great, thanks05:09
dufluTrevinho: I just tried out building Mir on trusty. There are lots of hacks required. More than you suggest. How/where are you building mir?06:03
Saviqhey guys, could someone please have a look at bug #1417773 and see if the attached traces suggest a Mir thread blocked somewhere?08:45
ubot5bug 1417773 in unity8 (Ubuntu RTM) "Unity8 completely frozen (unable to unlock, receive calls, etc)" [Undecided,New] https://launchpad.net/bugs/141777308:45
* duflu looks at the bug09:49
dufluSaviq: Fun. Looks like some kind of deadlock in Mir's GlibMainLoop with Mir's framedropping logic09:53
Saviqduflu, that sounds fun indeed09:54
* duflu shies away from owning problems he violently tried to stop from us having09:54
dufluAlso it's time to make dinner09:54
duflufrom us? us from.09:55
dufluSaviq: I know we did try to decide both ends of that deadlock so that deadlocks can't happen. But sounds like we're not there yet09:57
dufluSaviq: RTM running Mir 0.10? That sounds odd given https://launchpad.net/ubuntu-rtm/+source/mir09:58
dufluOh, also a stack from 0.1109:59
Saviqduflu, my gdb output was from vivid indeed09:59
dufluSaviq: Was the actual deadlock on vivid?10:00
Saviqduflu, yes, my gdb output was from vivid 0.10 and 0.11 today10:00
tmpRAOFI don't think that analysis is correct?10:00
tmpRAOFIt looks more like RPC failure?10:00
Saviqduflu, might be I misattributed the lock incorrectly10:00
dufluSaviq: OK, that needs clarification. Deadlocked on Mir 0.8 but retraced using Mir 0.10/11 symbols?10:00
tmpRAOF(Note one of the threads has blocked in recvmsg)10:01
Saviqduflu, no no10:01
Saviqduflu, the original bug was from rtm10:01
Saviqduflu, and rsalveti's trace should be from there10:01
Saviqduflu, my deadlocks were both on vivid, and retraced then and there, while the device was deadlocked10:01
dufluSaviq: rsalveti's looks quite different (dbus deadlock instead)10:02
dufluDo we have two bugs in one?10:02
Saviqduflu, very much possible10:02
Saviqduflu, there was a bunch of random lockups reported10:03
dufluSaviq: That's a huge can of worms. Time to make dinner10:04
Saviqduflu, enjoy10:04
tmpRAOFSaviq: Did you break and then continue in this backtrace?10:06
SaviqtmpRAOF, I just attached gdb and got the trace there, didn't continue10:06
tmpRAOFSaviq: So, the fact that it was blocked in recvmsg suggests that something in the RPC stack was being bogus, and the deadlock was U8 waiting for some RPC to finish.10:07
SaviqtmpRAOF, if you have any pointers as to what else should I check (continue + break) the next time I see that, please comment on the bug10:08
tmpRAOFSaviq: It might be helpful to work out what unity-system-compositor is doing when this happens.10:08
=== chihchun is now known as chihchun_afk
=== dandrader_ is now known as dandrader
=== alan_g is now known as alan_g|lunch
=== alan_g|lunch is now known as alan_g
=== snadg3 is now known as snadge
=== dandrader is now known as dandrader|lunch
greyback_AlbertA: hey, I've got QtCreator working reasonably well using this patch on top of your use-mir-surface-types: http://pastebin.ubuntu.com/10190374/16:32
greyback_thought I should share the changes I've made, in case it helps16:33
AlbertAgreyback_: cool...I've changed the branch a bit too....16:34
AlbertA:)16:34
greyback_AlbertA: ack, will pull and see16:34
greyback_AlbertA: moveResize deletes & re-creates the mir surface?! Cheeky ;)16:41
AlbertAgreyback_: works for menus lol16:42
racarrin which we become what we once hated :(16:48
racarr:p16:48
greyback_:)16:54
AlbertAracarr: that's why we are going to support client resize :) it's good for the mir team to feel the pain :)16:57
=== dandrader|lunch is now known as dandrader
=== alan_g is now known as alan_g|EOD
=== sturmflut_ is now known as sturmflut
mlankhorstyeah no other way :P20:49
mlankhorstsoon you'll find you want to support client hit testing for cursors too20:49
mlankhorstand  then realize you want to give sub windows a cursor to make it feel more responsive20:50
adrian47Hello is there anyone who can help with:  tiny.cc/i8lytx  (porting ubuntu touch)21:19

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