04:49 <riverdc> the dmabuf section in the wayland book says to call 'eglMakeCurrent to make current the EGL context for your surface and eglSwapBuffers to send an up-to-date buffer to the compositor and commit the surface'

04:49 <riverdc> do I understand correctly that clients don't need to call wl_surface_commit in addition to eglSwapBuffers?

08:24 <emersion> correct

10:58 <damjan> this might be off-topic but I didn't know better place, can I run 2 wayland compositors on a Radeon Vega on its 2 display outputs (HDMI and DP) … well, actually it's a Ryzen APU with Radeon iGPU. if not, is there some "spliter/mux" compositor, that would allow me to run a sub-compositor (a DE) on one output, and just a single app on the other

10:58 <damjan> display?

11:19 <ramcq> damjan: you should in theory be able to set them up as completely independent screens, with their own login sessions and their own compositors. unless you need them to be under the same session somehow...?!

11:55 <damjan> no, I don't care about the session

11:56 <MrCooper> it's not possible to just run separate compositors on different outputs of the same GPU, this would indeed require some kind of nested compositor

13:01 <ramcq> oh sorry I misparsed as there being two GPUs

13:31 <damjan> MrCooper: I kind of expected that, but hoped things might have progressed :). what is the actual singleton there, hardware-wise?

14:03 <MrCooper> damjan: it's more of a software architecture issue than a HW one; a bare metal display server requires DRM master, of which there can be only one at a time per DRM device (ignoring DRM leasing), of which there is one per GPU

14:15 <damjan> MrCooper: thanks

14:26 <MrCooper> actually there are HW issues as well, e.g. global resources which can be used with any output, but not multiple at the same time; or global resource limits which mean that while all outputs individually support the maximum resolution & refresh rate of the connected displays, it's not possible for all of them at the same time

14:28 <MrCooper> so there would need to be some kind of cooperation between the compositors to avoid surprising failures

14:36 <damjan> which would also require intricate knowledge of the hardware and its global resources

14:44 <damjan> let me try to find a suitable global compositor then. I wonder what are the proper keywords. a "system compositor", "mux compositor" …

14:44 <MrCooper> FWIW I don't know of any such compositor yet, you might have to write one :)

14:47 <damjan> I already have a long queue of programs I should write, but are far too above my competence :D