00:03 zmike: tarceri: awesome thanks
09:53 karolherbst: I was under the impression that some driver(s) might have broken UAPI in the past to fix security problems and I'm kinda trying to figure out what the details there were. Anybody know anything? Or am I simply remembering wrongly here
10:28 MrCooper: doesn't ring any bells offhand
10:30 karolherbst: panthor did restrict high priority queues, but there were no users.. but I think intel also restricted perf counters at some point? or was that CPU only?
10:30 karolherbst: but anyway
10:48 pq: if no-one notices, it's fine? :-)
10:51 karolherbst: basically :)
10:53 HdkR: I'll notice! But not out of spite :D
10:54 HdkR: Mostly because it means I might need to change stuff in my ioctl translation layer.
10:57 pq: but will you scream? or ask quitly and fix it up? ;-)
10:57 pq: *quietly
10:58 HdkR: A little of A and little of B.
10:59 HdkR: Keeps life exciting.
11:31 jani: tzimmermann: mlankhorst: ack for merging this one via drm-intel? there's just a small change for drm buddy: https://lore.kernel.org/r/20250829174601.2163064-3-jani.nikula@intel.com
12:00 tzimmermann: jani, of course
12:00 tzimmermann: a-b /me
12:03 jani: tzimmermann: thanks!
12:22 alyssa: karolherbst: asahi with upstream uapi doesn't even define high prio queus
12:22 alyssa: because there are no users and I didn't want to write rust bindings for the permission check (:
12:38 glehmann: there are users though?
12:38 glehmann: some compositors try to use high prio queues
14:38 alyssa: ddie
14:38 alyssa: oops
16:47 kisak: Does https://github.com/ValveSoftware/steam-for-linux/issues/12295 make sense to anyone? Ubuntu 24.04, mesa transition from 25.1.9 to 25.2.2 has a busted DRI3 support. User claims bumping Xwayland from a newer distro release helps. Is this from Xwayland on Ubuntu 24.04 not supporting explicit sync?
16:49 karolherbst: we've removed dri2 support, so maybe that's related? not sure if that happened in 25.2 tho
16:49 karolherbst: hopefully it's not
16:50 karolherbst: "No DRI3 support detected" oof..
16:50 karolherbst: sounds like it might be dri2 related indeed
16:51 daniels: yes, Xwl needs upgrading
17:52 kisak: tjaalton: It might be worth a making a note that you'll hit similar in the Ubuntu Questing -> Noble backport cycle. Xwayland is going to want to be >= 2024.1 to avoid the same regression.
17:54 kisak: that and a matching xorgproto bump shouldn't be too painful.
20:29 anholt: we don't have a NIR pass at the deref level that detects var[0] = vec4(var[0].x, var[0].y, c, d) and trims .xy off the write mask, do we?
20:35 anholt: maybe I can just make copy_prop_vars's value_equals_store_src() be a bit more clever.
20:35 karolherbst: no, but I'd be kinda interested in that as well
20:35 karolherbst: maybe