04:55 wlb: wayland Merge request !305 opened by yshui (yshui) protocol: fix inconsistent typing of wl_output.transform https://gitlab.freedesktop.org/wayland/wayland/-/merge_requests/305
09:21 MrCooper: ofourdan: mutter 44.0 does send a scanout tranche to weston-simple-dmabuf-feedback on my Intel laptop
09:22 ofourdan: same as me then
09:22 ofourdan: and fwiw, glxgears -fullscreen does not get direct scanout either (contrary to what I thought)
09:23 ofourdan: same with weston - but sway does send a scanout tranche
09:24 jadahl: ofourdan: could the non-matching visual in the ssd parenting be the cause of that?
09:24 emersion: does the window size match exactly the output size?
09:24 ofourdan: jadahl: maybe?
09:24 ofourdan: emersion: yup
09:25 jadahl: ofourdan: if you revert the revert you can test if it makes any difference
09:25 ofourdan: although, there is a a difference indeed
09:32 ofourdan: err, I am completely lost now… I reverted the revert, and now I do get a scanout tranche in weston-simple-dmabuf-feedback, except that should make no diffrence considering weston-simple-dmabuf-feedback is wayladn native :(
09:32 ofourdan: also, glxgears -fullscreen still does not get direct scanout
09:33 MrCooper: ofourdan: does for me, doesn't for you, right?
09:35 MrCooper: ofourdan: glxgears doesn't even with your Xwayland MR?
09:35 ofourdan: it seems inconsistent, I was sure it did yesterday, but today, no matter what, I never get direct scanout (testing with the repaint flag in looking glass)
09:36 ofourdan: yeah, I cannpt get glxgears -fullscreen to get dorect scanout, regardles of my patches, or even the reparenting revert-revert
09:36 ofourdan: I am puzzled…
09:37 MrCooper: there are many factors which can prevent direct scanout, any one of them is enough
09:38 MrCooper: the stars have to be aligned perfectly
09:43 ofourdan: yeah, could be a race somewhere
09:44 ofourdan: it *is* a race
09:44 MrCooper: FWIW, glxgears -fullscreen explicitly disables decorations, so mutter's reparenting issues don't affect it
09:44 ofourdan: I just retried and I got direct scanout
09:46 ofourdan: and same with the revert-revert + my xwayland branch
09:47 ofourdan: so it's just too inconsistent to tell, but my branch does not make it any worse at least :)
09:48 ofourdan: maybe I should ask the O-R to retry my branch on amdgpu?
09:53 MrCooper: yeah, why not
09:56 MrCooper: ofourdan: one issue I see is that Xwayland might use incomplete dma-buf feedback if not all feedback events are received/processed in one go
09:56 ofourdan: yeah
09:57 ofourdan: I thought of that
09:58 MrCooper: so maybe glxgears can race to query the modifiers before Xwayland has processed the scanout tranche
09:58 ofourdan: but the result with glxgears are inconsistent regardless of my branch
09:59 MrCooper: that's the thing with a race, sometimes you win, sometimes you lose
09:59 ofourdan: and so is weston-simple-dmabuf-feedback, which is wayland native
09:59 ofourdan: so I suspect there is a race in mutter, at least
10:01 ofourdan: err, actually
10:03 ofourdan: we recycle the buffers when the buffer format chnages, and that also applies once the full change is received (the done event) so I think we should be fine wrt the race you describe
10:04 MrCooper: right, just noticed the same
10:06 MrCooper: still seems possible for a modifier query to return incomplete results though
12:23 wlb: wayland-protocols Issue #137 opened by Xaver Hugl (Zamundaaa) Dealing with tabbed applications and virtual desktops https://gitlab.freedesktop.org/wayland/wayland-protocols/-/issues/137
13:08 Ermine: Hello, sorry for possibly dumb question, but I see xdg-shell protocol under unstable/ . Is it some leftover that should be ignored?
13:09 emersion: these were previous versions, Ermine
13:09 emersion: they have been superseded by stable/
13:09 Ermine: emersion: ok, thank you