=== LockeAnarchist- is now known as LockeAnarchist [05:56] RAOF: Can you remove the "devel" series now? I think we're used to "lp:mir" now. [06:14] duflu: Maybe? Let me check. [06:15] I can't... You own it :) [06:16] Turns out that I can :) [07:19] Oh, wow. [07:19] Yeah, that renderer frig is a really terrible cludge. [07:32] Ok CI. Review that branch before I wake up tomorrow morning. I dare you! [10:54] alf_: PidFakingSessionCoordinator? [10:57] alan_g: ok === dandrader is now known as dandrader|afk === dandrader|afk is now known as dandrader [14:06] alf_: in Washington we talked about adding "emergency cleanup" handling to mir::Server. Have you thought about that yet? [14:07] alan_g: no [14:09] OK. Unless you're keen to do it I might look at that. (Once I've figured out what is happening to ClientCredsTestFixture.session_authorizer_receives_pid_of_connecting_clients) [14:12] alan_g: It's not in my short-term plans, so feel free to pick it up if you have the cycles. [14:12] alan_g: currently focusing on main loop and a couple of assigned bugs [14:45] kdub: which part is too heavy, in real_hwc_wrapper? [14:47] would you prefer having the same interface has hwc, without a structure and conversions taking pace? [14:47] +l [14:49] anpok, somewhat, although maybe we could just log what's going on in terms of the api [14:49] hmm oh! [14:50] i missed adding the logs [14:50] like the point of that class is to be light and easy to pop logging in and out at the hwc api level [14:51] and if we do type conversions in there, it becomes thicker and thicker. Like, I could be sending a std::list in the prepare and set functions and converting the types in the 'RealHwcWrapper', but [14:52] thats the job of other classes to do the tricky work of converting the mir types to the raw hwc stuff [14:52] like, originally, I just mocked out the hwc type, but that proved zany in the test code (list checking in particular) [14:53] and the wrapper provided cleanups in terms of the code that executes the logic, as well as in the test code for HwcDevice [14:55] anpok, I guess my main suggestion is [14:55] since we have the need to requery this information from the hwc device, we should have a class that does the some sort of things that the factory does [14:56] and the DisplayDevice implementers can use that to get subsequent display configurations [14:56] which cuts down on duplication, and isolates the tricky business of gathering that information [14:57] the resource factory? [14:57] because the hwc api designers didn't really think it through and make it easy for us [14:58] hm i thought the duplication would go away as soon buffer creation would rely on the display configuration outputs provided by the display device.. [14:58] i thought display device would be the indirection to use for that purpose .. [14:59] .. differing between fb, hwc-1.0/fb and hwc-1.1 [14:59] well, there's more distinctions to be made [15:00] like hwc 1.3 supports opaque pixel formats and such [15:00] so hwc 1.1-1.3 have the same fundamental stuff going on in the business of posting displays, but the initial arrangements vary [15:13] kdub: so you mean I should be moving assembling of mir display structures to resource factory? [15:26] ok misread what you said above [15:29] right, like the code in framebuffers (in the anonymous namespace) may as well be its own helper class that the Framebuffers can use to make the fb-buffers [15:29] and some other part of the system (even, mga::Display) can use to generate the configuration === alan_g is now known as alan_g|tea [15:30] the DisplayDevice implementations have enough to juggle in the business of simply posting the layer list === alan_g|tea is now known as alan_g === dandrader is now known as dandrader|lunch === dandrader|lunch is now known as dandrader [17:50] I just built qtmir on my desktop. a test binary refuses to run because it's looks for libmirserver.so.25 whereas libmirserver.so.26 is the one installed [17:50] I wonder what's out of sync [18:07] oh, it's the installed /usr/lib/x86_64-linux-gnu/qt5/plugins/platforms/libqpa-mirserver.so