06:46 <steev> you mean you guys don't do all your kernel hacking while paddleboarding?

06:59 <HdkR> Hacking on the kernel is what I call complaining on IRC :P

07:33 <Xyaon> Could there be any USB related modules that are not enabled by default in the 6.3 distro_defconfig asides from the EC driver, that are enabled in the aarch64-laptops config?

07:36 <Xyaon> Even with the 6.3's devicetree my 5.x builds recognize usb devices while 6.2 ones don't

07:37 <steev> maybe?

07:37 <steev> check what modules are loaded

07:57 <Xyaon> I can't see anything usb related with lsmod

09:52 <jhovold> bamse, robclark: I use the same exports as rob, so including LIBGL_DRIVERS_PATH

09:53 <jhovold> I get "DRI driver not from this Mesa build ('23.1.1 (git-...' vs '23.1.1')" when I run startx

09:54 <jhovold> due to the commit I mentioned above that added the versioning check

09:55 <jhovold> or rather, after that commit, the real problem seems to be that LD_LIBRARY_PATH is ignored

09:56 <jhovold> when running X as I do

10:28 <jhovold> ftr, enabling glvnd made no difference

13:16 <jhovold> robclark, bamse: it's the Xorg setuid wrapper that causes LD_LIBRARY_PATH to be ignored, removing it is enough to allow the mesa-run script to work as expected

14:16 <quinine> I noticed that after upgrade my x13s kernel to latest steev lenovo-x13s-linux-6.3.y branch, the headset jack is not work properly. but strange thing is that even if I downgrade to an older kernel, it still doesn't work properly.

14:16 <bamse> jhovold: time to move to wayland ;)

14:43 <jhovold> bamse: yeah, perhaps it's time soon

14:45 <jhovold> robclark: found a good reproducer now, firefox crashes every time I open https://hlsjs.video-dev.org/demo/

14:46 <jhovold> trying to register on gitlab to open an issue, but not receiving any confirmation mail...

14:47 <robclark> ok.. cool.. check spam folder maybe?

14:49 <jhovold> nope, nothing in there either

14:52 <robclark> hmm, you can ask in #freedesktop which is kinda the admin channel for fd.o

14:53 <jhovold> ok, i'll try that

15:13 <bamse> jhovold: i switched to wayland because robclark said that he had switched and didn't test x a lot (at the time at least)...so it's probably a good idea that you and others are using x, so we get the test coverage...

15:27 <jhovold> robclark: i've created in an issue here: https://gitlab.freedesktop.org/mesa/mesa/-/issues/9110

15:27 <robclark> thx

15:28 <jhovold> bamse: yeah, i have not really seen any issues besides the firefox video playback one

16:00 <bamse> jhovold: and that's without any v4l interaction?

16:04 <bamse> jhovold: looked at your report, that's what i hit immediatly if i point firefox at any webgl app...and at random times when not trying webgl

16:06 <jhovold> bamse: yeah, i'm pretty sure v4l is not involved here, but I can double check tomorrow

16:07 <bamse> jhovold: no need, i should have read your report first

16:08 <jhovold> ah, good. sounds like robclark knows how to fix it

16:09 <bamse> jhovold: are you running your desktop environment under the LD_PRELOAD etc? or just the apps where you want acceleration?

16:10 <robclark> bamse: hmm, could be some ads using webgl?

16:11 <bamse> jhovold: can you open https://webglsamples.org/aquarium/aquarium.html ?

16:11 <bamse> robclark: webgl ads sounds like the future :)

16:12 <bamse> robclark: but i fear that it's common practice to use my gpu cycles without me knowing it?

16:12 <robclark> well, even non webgl ads tend to (indirectly, ie. via animations, etc) use your gpu cycles

16:13 <robclark> hopefully not too many of them are bitcoin mining

16:13 <robclark> (but since it is your cpu/gpu cycles they could probably just do that with javascript)

16:26 <bamse> robclark: that's why i run the ad-aware devfreq governor...

16:29 <bamse> robclark: speaking of crashes though, where you able to look at the fd6_texture_state() assert we spoke about a while ago?

16:29 <bamse> or, has there been any progress on that front?

16:29 <robclark> heh, ad-aware devfreq governor

16:30 <robclark> hmm, remind me what assert that was? And maybe file gitlab issue if you are still seeing it

16:31 <bamse> assert(state->view_rsc_seqno[i] == seqno);

16:31 <bamse> yeah, it kills sway for me with 23.1 once every 2-3 days

16:32 <robclark> hmm, right.. also don't build mesa with asserts enabled

16:32 <bamse> hmm, thought i did remove that

16:33 <bamse> -Db_ndebug=true ?

16:33 <robclark> right

16:33 <robclark> bamse: ok.. I think I fixed it.. https://gitlab.freedesktop.org/mesa/mesa/-/issues/8816

16:35 <bamse> 87978c393340 ("freedreno/a6xx: Allow z24s8 format casts") is included in mesa-23.1.0, which is what i'm running...

16:37 <robclark> hmm, then build mesa with debug syms and get a backtrace? I guess it is a similar sort of issue where we end recursing into setting sampler view via blit we can't do on 2d

17:02 <agl7> Good Evening! :-)

17:32 <jhovold> bamse: that webgl sample crashes the same way as the video player here, same suballocation issue it seems

17:33 <jhovold> I can load the page with my patched mesa without it crashing, but I don't see any fish

17:33 <jhovold> so likely the same issue

17:34 <jhovold> bamse: and yes, I run everything under LD_LIBRARY_PATH etc

18:27 <steev> quinine: i have to admit, i don't check the headphone jack often, i'll try to find my headphones again and see

18:31 <steev> quinine: headphones work here, using what i just pushed; the headphones i have are the old apple ones though (no idea if the mic on them work but i'm able to play a youtube video and hear the music, as well as play an oga file with canberra-gtk-play (at the same time even)