00:02 mareko: Sachiel: just "python3 external/fetch_sources.py" and "cmake . -G Ninja -DCMAKE_BUILD_TYPE=RelWithDebInfo" and that fails
00:03 Sachiel: oh, there's a good chance that doing that is broken. Try "cmake -B build ..." instead
00:11 mareko: thanks, that fixed it
00:18 zmike: if you're trying to build GTF make sure you're passing the right cmake option for that
00:23 mareko: zmike: is -DGLCTS_GTF_TARGET=gl enough?
00:23 zmike: looks right
00:23 mareko: that just builds GL CTS, not sure about GTF
00:23 zmike: though I'm on my phone at the pub
00:23 zmike: yeah
00:23 zmike: it's part of glcts
00:24 zmike: you run it the same way assuming you ran the fetch_kc script
00:24 zmike: with your khronos credentials
00:24 mareko: yes
00:24 zmike: then you should be good
00:24 mareko: I guess I need a different caselist though
00:27 zmike: there's separate mustpass lists
00:27 zmike: iirc it's in with the regular lists
00:35 mareko: external/openglcts/modules/gl_cts/data/mustpass/gl/khronos_mustpass/4.6.1.x/gl46-gtf-master.txt must be it
00:36 zmike: looks right
02:40 zmike:needs to stop doing mesa after pub
02:46 mareko: doing mesa after pub is never a bad idea
02:53 zmike: haha
07:24 pq: _DOOM_, https://dri.freedesktop.org/docs/drm/userspace-api/dma-buf-alloc-exchange.html has a terse glossary of "plane" etc.
11:50 kode54: cool
11:50 kode54: and by that, I mean, huh?
11:50 kode54: amdgpu hangs in some UE5 games possibly bisected to a zmike commit
11:50 kode54: funny thing is, I haven't had any whole compositor hangs lately
12:17 anholt: daniels: looks like mesa jobs are not surfaced on lava.collabora.dev?
12:21 anholt: was hoping to go digging and figure out why our duts feel overloaded, when they don't look that way from a gitlab perspective.
12:29 daniels: anholt: correct, they're ACLed because they contain the JWT as part of the job definition - LAVA upstream are working on secrets support which would solve that problem, so we were just leaving them to it since it didn't seem super urgent
12:30 daniels: wanna shoulder-surf in the break?
12:31 anholt: daniels: yeah, let's take a look
14:02 anholt: hakzsam: thanks for your persistence with the cts!
14:09 daniels: ++
14:09 daniels: and for tuning DEQP_FRACTION down so we don't blow out the runtime :)
15:44 wv: How is an overlay plane exposed to userspace? I have 2 hardware planes on my imx53. When I launch an application (kmscube let's say), it gets to the primary plane. But how can I get another application (gstreamer kmssink fi) to the overlay plane?
15:44 wv: in old legacy devices, we had a fb0 and fb1 exposed
17:39 mareko: why not call the compiler NCO
17:43 zmike: well there was a pub at xdc...
18:25 mareko: or let's rename ACO to ACK
18:29 zmike: 🤔
18:48 panera: hi, not sure if this is the right place to ask, but adaptive_sync is blocked for firefox in /usr/share/drirc.d/00-mesa-defaults.conf but when I fullscreen a youtube video adaptive sync still kicks in (giving flickering).
18:57 panera: I'm on KDE wayland.
19:07 zf: according to validate_var_decl(), nir expects geometry input to be an array of arrays of scalars
19:08 zf: why are there two arrays there?
19:12 zf: eh, never mind, I think I see the answer
19:19 cheako: This is a hot path for anv and I wonder if this is a good patch, I don't know much about c++? https://www.irccloud.com/pastebin/NnS25OPi/
19:24 cheako: There is also a big patch to anv aimed at the same issue: https://gitlab.freedesktop.org/mesa/mesa/-/issues/9999#note_2131807