17:56 <marex> mntmn: I have a feeling the gl1.0 depth is broken, so you were right about the neverball depth problem

17:56 <mntmn> marex: oh ok!

17:57 <mntmn> depth, not alpha?

18:01 <marex> mntmn: export MESA_GL_VERSION_OVERRIDE=2.1 ; export MESA_GLES_VERSION_OVERRIDE=2.1 ; export PIGLIT_PLATFORM=gbm ; export WAFFLE_PLATFORM=gbm ; piglit run -ovs1 tests/all -t depth results/all

18:02 <mntmn> ok

18:02 <marex> spec/!opengl 1.0/gl-1.0-drawpixels-depth-test: fail

18:02 <marex> spec/!opengl 1.1/depthstencil-default_fb-clear: fail

18:03 <marex> I need to look into those tests, but it looks like depth

18:03 <mntmn> interesting

18:03 <marex> that's on GC2000 btw

18:03 <marex> also, I have a feeling that somehow, point coord / tex coord is influenced by depth or something

18:03 <mntmn> that sounds strange

18:04 <marex> i.e. when I look at the stars in neverball right at the beginning in the menu, they are scaled

18:04 <marex> the quad is the same size, always, but the texture on it is scaled

18:04 <marex> (it shouldn't be, it is not with LIBGL_ALWAYS_SOFTWARE)

18:15 <marex> mntmn: WAFFLE_PLATFORM=gbm PIGLIT_PLATFORM=gbm /usr/lib/piglit/bin/gl-1.0-drawpixels-depth-test

18:16 <marex> mntmn: if you have time, try that, it will likely fail

18:17 <mntmn> i believe i do not have piglit set up atm

18:18 <marex> mntmn: then dont worry about it :)

18:19 <mntmn> ok :3

18:57 <mntmn> meanwhile, mutter (wayland compositor version) is currently broken with etnaviv/current mesa/gc7000l

18:58 <mntmn> when running on kms directly. works on top of sway as nested compositor.

18:59 <mntmn> but that's mutter in debian, will build from source

19:13 <marex> mntmn: broken how ?

19:27 <mntmn> mutter 40 (from git) is borked in the same way

19:27 <mntmn> video coming up

19:29 <marex> mntmn: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8174

19:29 <marex> mntmn: I dunno if this changes anything for you, but I noticed this bugfix is necessary, it broke around 20.3

19:30 <mntmn> marex: http://dump.mntmn.com/IMG_1328.MOV

19:31 <mntmn> marex: ok, gonna try

19:33 <mntmn> marex: it's actually hard to see in my video, because it flickers at 60hz i guess

19:33 <marex> mntmn: that mouse multi-cursor is quite crazy :)

19:34 <mntmn> the cursor is updating only when a lot of other stuff around it is also updating, and in general there's a big flicker going on between the last 2 buffer states

19:35 <mntmn> it's like not all draws/updates are flushed or something

19:35 <marex> mntmn: I think I noticed the same thing with chromium on WL then

19:36 <mntmn> marex: hey i think your patch fixed it

19:36 <mntmn> wtf

19:37 <marex> mntmn: it did ?

19:37 <mntmn> i'm using gnome3 right now without any of this problem

19:37 <marex> cool

19:37 <mntmn> either your patch fixed it _or_ i forgot to do an install when i compiled mesa earlier

19:37 <mntmn> i will double check now

19:37 <marex> better do that

19:41 <mntmn> marex: i can say with 100% certainty that your patch fixes this problem

19:41 <mntmn> amazing.

19:52 <marex> mntmn: that's nice to hear

19:52 <marex> austriancoder: ^

19:54 <mntmn> this is cool because it means we can ship gnome3 with reform

19:54 <marex> mntmn: maybe comment on that MR, that it fixes stuff ...

19:54 <mntmn> i did!

19:55 <mntmn> https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/8174#note_742303 ^^

19:55 <marex> mntmn: thank you

19:56 <mntmn> hmm gnome feels a bit faster these days... maybe i'll switch to it for a while

19:58 <mntmn> for the record, it is gnome-shell etc Version: 3.38.2-1 and mutter 40.alpha.1

20:00 <marex> mntmn: maybe you should run deqp functional tests with every mesa update, then you would notice these new issues :)

20:00 <mntmn> yeah

20:01 <mntmn> i should set up a reform/imx8mq permanently as a gitlab ci worker

20:03 <marex> that would indeed be real nice