[05:13] <RAOF> alf_: I'm in the process of wandering through graphics/gbm and wiring up buffer passing using dma-buf rather than flink. Just a heads-up.
[05:53] <RAOF> Hm. That was disturbingly quick, which suggests we don't test the buffers all that much :)
[07:10] <llstarks> RAOF, i've been trying to find a real world example of dma-buf use. how are you using it?
[07:11] <RAOF> llstarks: Passing prime fds rather than flinked gem buffer names.
[07:11] <llstarks> radeon?
[07:12] <RAOF> Radeon, Intel, and Nouveau all do it.
[07:23] <llstarks> i know
[07:23] <llstarks> waiting for maarten's fencing work and for nvidia to use the new shims
[07:24] <RAOF> Somehow I've managed to trigger a g++ 4.7 ICE.
[07:42] <llstarks> you broke c
[07:49] <tvoss> llstarks, RAOF is well-known for that :)
[07:50] <RAOF> I love the smell of C++11 in the morning
[07:52] <tvoss> RAOF, ;)
[07:57] <alf_> RAOF: ok, looking forward to that
[07:59] <RAOF> Hah. Looks like g++ 4.6 gives a useful error rather than ICEing.
[08:08] <RAOF> For those playing at home - you can ICE g++4.7 by having one of your Google Mock objects missing a pure virtual function.
[08:33] <RAOF> alf_: You can check out lp:~raof/mir/use-dma-buf; I *think* that's pretty much the server-side done.
[08:33] <RAOF> And with that, dinner.
[08:54] <alf_> RAOF: sure, checking
[08:57] <llstarks> it's cool seeing dma-buf userspace code
[09:02] <RAOF> llstarks: There's some in xf86-video-{intel,radeon,nouveau,modesetting} :)
[09:10] <alf_> RAOF: Looks good and runs well overall. With some slight cleanup it will be ready to go \o/
[09:10] <alf_> RAOF: plus answering the GEM_CLOSE/close(fd) questions
[09:10] <RAOF> alf_: Oh, you've tested it with actual clients? :)
[09:10] <alf_> RAOF: just the unaccelerated one, that doesn't use mesa
[09:11] <RAOF> I guess I could have done that; software-rendering should work, right? I haven't updated the mesa bits.
[09:11] <alf_> RAOF: and render_surfaces which is server side only
[10:21] <zAo^> Will GTK2 become Mir compliant?
[10:21] <zAo^> Or only QT5.1 & GTK3?
[16:57] <min2> hello
[19:12] <llstarks> i don't think of ddx drivers as userspace unless it's in the context of ums. for the past few months it's like dma-buf is present in every component it needs to be, but nothing calls it. i have to go out of my way to set up nouveau prime on my laptop just to see it in action. or am i missing something?
[21:39] <llstarks> can i run gnome-shell on mir?
[21:58] <RAOF> llstarks: Yes*
[22:07] <RAOF> llstarks: Where the * denotes - “in a nested X server”. Otherwise, the question is like “can I run gnome-shell on weston”, which is like “can I run gnome-shell on Unity” ☺
[22:17] <llstarks> so only unity runs natively in mir atm?
[22:19] <RAOF> llstarks: I think you're misunderstanding what Mir is.
[22:19] <AlanBell> which is a fault of the documentation to an extent
[22:20] <RAOF> True!
[22:20] <AlanBell> the spec is a somewhat challenging read to most people
[22:20] <RAOF> Mir is not a display server in the sense of X; it's more like a library for building display-server-compositor-shells, like wayland.
[22:21] <RAOF> It's somewhat hard for us to write a good summary of what Mir is, because it's obvious after you've spent several months designing it ;)
[22:22] <llstarks> that's just a problem of communication, which seems central to the case for mir at this point
[22:23] <RAOF> As far as I'm aware there *is* a build of the Ubuntu-touch Unity that's linked against Mir, though. That's the closest to "unity running on mir" that makes sense.
[22:24] <AlanBell> llstarks: so far, the source enables you to get a scrolling picture of the mir space station in a green box
[22:24] <llstarks> i saw that
[22:24] <llstarks> very sexy
[22:24] <AlanBell> it stops on ctrl+c but as I understand it you can't capture other keys yet
[22:25] <AlanBell> actually I am not sure if mir stops then kills the client or the other way round
[22:25] <RAOF> Well, and also X-on-mir, but apparently I haven't pushed that anywhere usefully public yet.
[22:25] <AlanBell> I think clients tell mir what to put on the screen, and mir talks OpenGL to the GPU
[22:26] <AlanBell> mir kind of manages requests from multiple clients, it shares out the GPU access kind of like a hypervisor does with a CPU
[22:26] <llstarks> usefully public should be what you and tvoss are cooking with nvidia. you guys are about to kill 2 birds (optimus and mir enablement) with one stone and i must commend that.
[22:27] <llstarks> counting on nvidia to fix their x stack would never work
[22:27] <RAOF> AlanBell: Mir hands out a buffer, the client does what it wants with it (which will generally involve rendering to it with some variant of GL), the client says “I'm done with this frame”, and then Mir takes the buffer and displays it.
[22:27] <AlanBell> RAOF: a 2d buffer?
[22:27] <AlanBell> oh, rendered to the buffer with GL
[22:28] <RAOF> AlanBell: A GPU buffer, although clients can request a buffer that's CPU-accessible.
[22:28] <RAOF> It is, not *entirely* surprisingly, quite a lot like wayland in that respect.
[22:33] <racarr> Finally close to getting rid of the last frontend/shell unit test coupling to shell::ApplicationSession and ms::Surface/ProxySurface (except of course the application session unit test ;))
[22:33] <racarr> have lost so much time over the last week to having to change 10 tests everytime I mess with the application session or surface interface
[23:12] <wolfslord> How should I install the custom Mesa in order for Mir to work?
[23:14] <AlanBell> wolfslord: I installed mesa from the PPA and build mir from source on Raring
[23:15] <AlanBell> oh, Mir is now in the PPA with lightdm and xorg-server with xmir
[23:15] <wolfslord> I'm using raring now under VirtualBox and tried to install the ppa but it gives errors when I do 'apt-get update'
[23:16] <AlanBell> but the PPA version of xorg is based on the prior version in raring so that might not install
[23:16] <AlanBell> 2:1.13.2-0ubuntu2+xmir1 is in the ppa but 2:1.13.2-0ubuntu3 is in the main repos
[23:17] <wolfslord> Also how should I install mesa. I see from the ppa that there are a lot of packages. Is there a meta one or something?
[23:17] <AlanBell> I am kinda doubtful that it will support the virtualbox 3d stack?
[23:18] <AlanBell> wolfslord: add the ppa with sudo apt-add-repository ppa:mir-team/staging
[23:19] <AlanBell> then sudo apt-get install mir will probably bring in the rest
[23:20] <RAOF> Note that xorg-server failed to build, because mesa was buggy at the point it tried to build.
[23:21]  * AlanBell installs 166MB of assorted stuffs
[23:21] <AlanBell> rolling releases = 100MB/day \o/
[23:22] <AlanBell> !mir
[23:22] <AlanBell> ^^ does that look accurate?
[23:22] <racarr> Anyone have time to review: https://code.launchpad.net/~robertcarr/mir/reduce-coupling-in-application-mediator-tests/+merge/152304 would be nice to land it this afternoon so I can submit another branch based on top of it that people can review in the morning without using pipes
[23:22] <racarr> guessing the next branch will be more contrversial than this one ;)
[23:23] <RAOF> Heh.
[23:25] <kdub> with blobs, there's always something fun to figure out
[23:28] <robert_ancell> AlanBell, that looks good. Thanks
[23:29] <AlanBell> it was main inclusion request, but that wasn't used much so we changed it