[wip] Add dmabuf feedback

[Consolatis]

Fixes

• DMABUF feedback support #1268

This turned out to be a bit more complex because the wlr_renderer_init_wl_display() call we used to do does not store or return the wlr_linux_dmabuf_v1 pointer anywhere (which we need to supply to wlr_scene_set_linux_dmabuf_v1()).

So instead of using wlr_renderer_init_wl_display() we just do it manually. This is also what the the sway scene graph patches do. We could also just use both but that causes duplicated globals warnings in wlroots which might cause further hard to detect / debug issues.

As for how to actually test this.. I have no idea.

Todo:

• remove logging
• find a way to actually test this

[yuiiio]

can test this with weston-simple-dmabuf-feedback.
and output is same as sway, so works fine.
(weston-simple-dmabuf-feedback needs revert this commit 34400d7d1686d53ceaf79ff1142349a1a171db68 cause assertion error, even with sway.
created sway issue swaywm/sway#7879 maybe bug in impl xdg-shell and wlroots or labwc has same logic)
but, not related this MR :)

[jp7677]

Accidentally I stumbled upon the dmabuf-wayland video output in mpv, usable by e.g. mpv -v --no-config --hwdec=vaapi --vo=dmabuf-wayland <video-file> (or without hwdec).
This output mode bails out without this PR. but works nicely with this PR applied. The mpv output looks normal as far as I can judge. Also the labwc verbose output shows no failing statements from this PR.
So I would also lean towards "This PR works as intended".

[ahesford]

The dmabuf-wayland output has always worked for me. (It mangles HDR colors, but that's expected because it does no color management.)

[jp7677]

ah yeah sorry. I though to be clever to just set the additions from this PR into comments for testing, but missed the fact that I had to also revert wlr_renderer_init_wl_shm back into wlr_renderer_init_wl_display. This resulted in mpv saying [vo/dmabuf-wayland/wayland] Compositor doesn't support the zwp_linux_dmabuf_v1 (ver. 4) protocol! which tricked me, but clearly my mistake. Next time it is git stash again...

Sorry for the noise.

[Consolatis]

I remember reading an anecdote about dmabuf feedback causing an overflow of the wayland connection buffer and thus wlroots killing the client. AFAIR there were talks about wlroots potentially sending too many tranches but I didn't follow up on that. That is another reason why I am a bit skeptical about this PR (I don't see it improving anything other than being able to say "look, we support a new feature" but it might actually cause issues that weren't there before).