13:44 <mntmn> hmm, having some new problems.

13:44 <mntmn> kmscube: ../src/gallium/drivers/etnaviv/etnaviv_uniforms.c:90: etna_uniforms_write: Assertion `cb->user_buffer && val * 4 < cb->buffer_size' failed.

13:44 <mntmn> when i comment out that assert, kmscube just shows a grey screen

13:44 <mntmn> this is with newest mesa from git

13:45 <mntmn> + patches 7603 8618 9214

13:46 <austriancoder> mntmn: lets see..

13:48 <mntmn> same with the patches removed.

13:48 <austriancoder> mntmn: k

13:51 <mntmn> guess i have to bisect :|

13:53 <austriancoder> mntmn: give me some minutes.. I am just starting my dev board

13:53 <mntmn> austriancoder: sorry, that wasn't meant at your direction

13:53 <mntmn> austriancoder: i'm just poking around, too

13:54 <austriancoder> mntmn: I have landed some bigger changes yesterday.. so it might have broken something

13:54 <mntmn> ah ok

13:55 <mntmn> i can already say that going back 100 commits fixes kmscube ;)

13:57 <austriancoder> mntmn: I think it the bad one might be: dba692c6187d2de284bb909f9da7fee483b064ff

13:57 <mntmn> thanks, going to test

14:01 <mntmn> austriancoder: confirmed

14:01 <mntmn> austriancoder: kmscube works for me on mesa master with that commit reverted

14:04 <austriancoder> mntmn: yep.. working on a fix

14:05 <mntmn> awesome

17:30 <mntmn> hmm hmm Xwayland windows now have garbage content under gnome3 (mutter) but look fine under sway.

17:30 <mntmn> looks like wrong pitch or some kind of tiling

17:32 <mntmn> there was an env var to disable supertiling, right? can't find it in the source atm

17:33 <lynxeye> mntmn: ETNA_MESA_DEBUG=no_supertile

17:35 <mntmn> lynxeye: thank you!

17:37 <mntmn> hmm no, that was not it.

17:39 <mntmn> lynxeye: any intuition? this is glxgears http://dump.mntmn.com/IMG_2044.mp4

17:40 <lynxeye> mntmn: That totally looks like mutter is sampling the client buffer as linear, while it's tiled.

17:40 <mntmn> alright

17:41 <mntmn> thanks

17:42 <lynxeye> Is mutter correctly importing the buffer with modifiers? Is Xwayland on mutter even using the zwp_linux_dmabuf protocol, which is a dependency for getting proper tiling information across the wayland connection?

17:44 <mntmn> it appears in the mutter sources, and i have it working in another installation. i'm gonna compare the versions or if i had made any patches

17:46 <mntmn> i'm in the process of puzzling together the disk image that ships with reform, which is built from scratch, so some versions have changed

18:57 <mntmn> looks like i first need to understand how gnome3 works. i don't yet understand how mutter is integrated in gnome-session and how xwayland is integrated, too

19:02 <mntmn> gnome-shell uses libmutter. libmutter has the xwayland integration.

21:23 <mntmn> now getting bizarre segfaults when resizing xwayland windows http://dump.mntmn.com/xwayland-resize-segfault.txt cc austriancoder

23:01 <mntmn> segfaults and crashes are gone with combination of wlroots 0.12.0, sway 1.5.1, mesa 20.3.3 and xserver/Xwayland master

23:22 <mntmn> daniels: any idea why xwayland surfaces would be sampled incorrectly on contemporary gnome-shell/mutter? (3.38.2 and .3) i see you added zwp_linux_dmabuf stuff 3 years ago.