[00:04] <RAOF> Why hello, odd bug.
[00:05] <RAOF> Why isn't disconnect() being responded to?
[00:07] <RAOF> Oh, goodie. It's somehow a race?
[00:11] <RAOF> AAAaaaah.
[00:13] <RAOF> It's a race between mir_render_surface_release and mir_connection_release.
[01:27] <RAOF> Hm.
[01:27] <RAOF> Why do we ignore logical_width and logical_height for bufferstreams in mir_surface_spec_add_render_surface?
[01:27]  * duflu shrugs
[01:29] <RAOF> Ah.
[01:29] <RAOF> The reason why the tests don't *currently* deadlock in that race is because they race allocate_buffer instead.
[01:39] <duflu> RAOF: I think I'm overallocated for features this cycle. Do you want to take one of my major items off my hands?
[01:39] <RAOF> duflu: Whatta ya got?
[01:40] <duflu> (fbdev + osmesa || SimpleDRM kernel work)
[01:40] <duflu> The first one is just a Mir driver
[01:46] <RAOF> The second one is driving an already roughly-agreed-to patch through lkml.
[01:49] <duflu> Indeed, but it's been driving for a few years
[01:49] <duflu> Not sure how long to wait
[01:49] <duflu> Perhaps if we do nothing the feature will exist :)
[01:50] <RAOF> Patch set v5 got submitted by a drm guy in September.
[01:50] <duflu> To me fbdev would be fun to learn driver development. And it would work on any kernel... assuming osmesa works
[01:50] <duflu> But that's just me
[01:51] <RAOF> Why would fbdev involve learning driver development?
[01:52] <duflu> RAOF: Mir driver development
[01:52] <RAOF> Ah, right.
[01:53] <duflu> I know the basics... I was building something similar back in the early 90s before I got distracted
[01:56] <duflu> RAOF: I just figured you know all the parts. If we were to build a fallback driver to solve all the fallback problems before 17.04 you would be able to do it
[01:57] <duflu> And working in any VM out of the box would be awesome
[01:58] <duflu> It's assigned to me, but before then I need to finish client-side vsync and probably now do GL-based display cloning
[01:58] <duflu> So there's a reasonable risk I won't have all three features done before 17.04
[01:59] <RAOF> Seems reasonable.
[01:59] <RAOF> I'll see where SimpleDRM is at.
[01:59] <RAOF> Because if it's done, that's the cheapest option :)
[02:05] <RAOF> And, indeed, looks like simpledrm has landed.
[02:11] <duflu> Not in Linus' tree. Which one comes before that for DRM work?
[02:12] <RAOF> In whatever tree all the patches to drm-devel are applied to.
[02:12] <RAOF> I'm actually not sure what tree that is.
[02:14] <duflu> You always need to know who the second in command is, per component
[02:14] <duflu> The kernel docs say who
[02:18] <RAOF> Yeah, it's drm-misc. I thought so.
[02:18] <RAOF> Can't see it in there, though.
[02:20] <RAOF> Which is odd, because I see unrelated patches that patch simpledrm.
[03:36] <RAOF> Hey, cool! rr will now record mir_acceptance_tests!
[03:43] <duflu> And I thought Unix already took all the two letter names
[06:04] <duflu> RAOF: Is it intentional or incidental that streams use SurfaceIDs and SurfaceObservers?
[06:04] <RAOF> Incidental, mostly.
[06:04] <duflu> Oh I see. We started with just surface and it morphed incrementally into streams
[06:05] <RAOF> For the former, it might have made sense to have just a single “object ID” namespace.
[06:05] <RAOF> For the latter, SurfaceObserver basically makes sense with Streams, too. :)
[06:05] <duflu> Indeed
[17:44] <applemuncy_1> Greetings : )  Is this the right place to ask about mir_android_diagnostic testing and failures?
[17:58] <greyback> applemuncy_1: yep