04:09 zzoon[m]: airlied: do you have any branch of FFMpeg to test vulkan av1 decoding?
04:13 airlied: zzoon[m]: there was one in the khronos ffmpeg repo, but I'll push to my github one now
04:14 airlied: https://github.com/airlied/FFmpeg/tree/av1-decode-wip though I haven't tested it as much as I did with CTS
04:14 zzoon[m]: yeah that was not on top of the released spec.
04:15 zzoon[m]: oh thanks. I'll try with that branch
08:10 babylonian: I had plans for modern opengl es stack, little code needed, but I think I leave the scenes, disruptive situations, and overly obnoxious stalkers from those porn heros who harassed my territory. This was left to a situation that satisfies me, trashes teeth were knocked out, she comes again with her porn buddies to poke my life, they are all in wheelchair.
08:14 babylonian: There's lots of small improvements to the existing possible, but it's core methodical side should be known to you by now.
13:40 sadmoments: I finished with super results, all issues are resolved hardware parts and software wise I got all issues down, so I am just in super position, it's overwhelming to see I did the maximum, i.e more than I considered, my future is bright anyway, and all the future of computing is bright, I covered all the world's hw, so you should not worry about me just do not come to bully me on my territory with your porn heros saying your wank shit, your
13:40 sadmoments: sluts are banned among with you from all my territories, force will be likely used anyways against you since your actions were not only unfriendly but murderous and consistent to harm me, and I took more injuries as result.
14:25 mripard: tzimmermann: the ARGB8888 patch for udl has been committed to drm-misc-fixes, do you want to send a revert before it ends up in 6.8 final?
14:27 tzimmermann: mripard, when do you send out the PR? i don't want to send out a revert without talking to doug first.
14:27 mripard: no, I'll send it tomorrow
14:28 mripard: but that's kind of my point, it's controversial we're super close to the release anyway
14:28 javierm: it's still early there but I expect him to be around in an hour or two
14:29 javierm: but I agree with mripard that probably a revert would be better since at the very least needs more discussion
14:29 tzimmermann: right. but i missed that patch a week ago. so it's also my fault.
14:29 javierm: tzimmermann: same and I guess dianders wasn't aware of the previous discussions on this topic
14:30 dianders: tzimmermann / mripard: Just getting online. Revert is fine and we'll figure out what to do.
14:31 dianders: You want me to post it?
14:31 tzimmermann: dianders, thanks
14:31 mripard: there's no need to blame anyone, you can't be expected to review everything, and dianders followed the usual process but probably wasn't aware of the previous discussions
14:31 tzimmermann: dianders, yes please
14:31 mripard: so it's all in good faith, no worries
14:39 dianders: tzimmermann / mripard: Posted. Should show up as https://lore.kernel.org/r/20240306063721.1.I4a32475190334e1fa4eef4700ecd2787a43c94b5@changeid once it percolates through.
14:39 dianders: Feel free to land it yourself, or if I see a review tag I'll land it in a little bit. I'll also respond to the email now.
14:39 tzimmermann: thanks a lot
14:41 mripard: dianders: thanks
14:50 javierm: dianders: was trying to find the discussion but can't find in the mailing list... tl; dr jfalempe wanted to expose a "fake" ARGB that would just be XRGB and drop the alpha channel when copying the pixels to VRAM
14:51 javierm: dianders: this was because a display controller was very slow at scaning out and he wanted to save some bandwidth
14:52 javierm: and there was a heated irc debate with one of the linux asahi developers too
15:14 dianders: tzimmermann / mripard / javierm: Pushed the Revert. Sorry for the bother. The whole "don't break userspace" still feels like a reasonable argument for landing something in the kernel, but I'll also see if I can figure out how to make this work in our userspace...
15:15 tzimmermann: dianders, thanks for the revert. i was going to type a response to your mail
15:15 javierm: dianders: no bother at all, sorry that I missed your patch since I did have context about previous discussions
15:15 dianders: tzimmermann: Sounds good. We can continue the discussion there.
15:16 mripard: dianders: thanks for the quick turn-around :)
15:17 javierm: dianders: specially since I argued your point too :) but it seems I was in the minority that thought the kernel should expose/fake other formats than XR24
16:14 thellstrom: airlied:, sima: How is the "no regressions" policy handled in the the case where user-space is mishandling or not handling recoverable errors, like out-of-space for an operation that can be broken up to smaller chunks, or VM_BIND vm overcommit.
16:15 thellstrom: Let's say one kernel changes an internal structure size and returns a recoverable error for a case that was previously working.
16:35 jani: any idea how kbuild decides to recurse into top level subdirectories like drivers/ ?
16:37 jani: d'oh, it's in ./Kbuild nowadays instead of ./Makefile
18:27 zmike: merging https://gitlab.freedesktop.org/mesa/piglit/-/merge_requests/884 in a few hours if nobody comments otherwise
20:33 jenatali: https://mesa.pages.freedesktop.org/-/mesa/-/jobs/55960210/artifacts/results/summary/results/trace@freedreno-a630@gputest@tessmark-v2.trace.html - should there some kind of tolerance on the result here?
20:33 jenatali: It's a little frustrating having a change that sat in the merge queue for hours kicked back for 2 pixels that are not visibly different :(
20:58 zmike: yeah it sucks
20:58 zmike: was it for a nir change? or just a random driver flake
20:58 jenatali: This is a flake, passed on a rerun
20:59 zmike: :/
20:59 jenatali: It was run because the MR is a nir change, but in a pass that this driver doesn't run
20:59 airlied: thellstrom: you fix the userspaces first and wait 6m to a year usually
21:00 airlied: then the chances of someone seeing the regression is low enough
21:00 airlied: for xe you might not need as long depending on how released userspaces are
21:03 thellstrom: airlied: Thx!
21:10 daniels: jenatali: if it’s unstable then please disable it
23:25 DemiMarie: airlied: So that means that Qubes OS might have some problems, because we have new kernels with some very old userspace.