02:29Lynne: having CTS is a great idea, until you find out the hard way it covers only a minimum part, and everything outside of that specific part gets broken all the time
02:54HdkR: Dang those game devs, pushing the limits of validation.
06:41Lynne: I'm not a game dev, I'm running bog standard compute code
06:41Lynne: no one's really doing codecs inside shaders but me
06:42Lynne: and probably because no one cares but me, what I want to do broken on every single implementation in a different way
06:43HdkR: Bink used to, not sure if it still does.
06:43Lynne: single invocation, single workgroup, single everything, yet when I read a value I just wrote to a non-shared normal non-array variable, I get garbage
06:45Lynne: is there an envvar that generates the most naive foolproof NIR and shader binary to at least narrow it down to a driver?
06:49dj-death: Lynne: on Intel VK_ENABLE_PIPELINE_CACHE=0 INTEL_DEBUG=cs
06:50Lynne: it just dumps the disassembly?
07:01dj-death: Lynne: NIR at 3 different times if I remember and then assembly
07:11Lynne: not that useful, my code's not simple
07:11Lynne: doing entropy encoding, so it has to be bitexact
09:49emersion: jadahl: do you think you'd have time to update this PR? https://github.com/KhronosGroup/EGL-Registry/pull/205
11:09demarchi: sima, airlied_, jani: https://gitlab.freedesktop.org/drm/maintainer-tools/-/merge_requests/94 ... changing the Link: trailer as requested by Linus
13:00mlankhorst: jfalempe: Is the panic screen supposed to show on all screens or just first one?
13:08tzimmermann: jfalempe, thanks for reviewing https://patchwork.freedesktop.org/series/156210/ do you have any comments on patch 7 before i send out an update to the series?
13:09jfalempe: mlankhorst: it's supposed to show on all screens. (from drm, it's drawn on all primary planes of all GPU's that supports it).
13:10jfalempe: tzimmermann: it's fine, looks like I forget the last one of the serie, sorry.
13:10tzimmermann: jfalempe, no worries :)
13:33mlankhorst: jfalempe: might be crashing after the first screen then
13:34mlankhorst: https://lankhorst.se/etc/panicc.jpg
13:36mlankhorst: It's a dual panel DP-MST screen, only draws on the right side (DP-0.8) not left (DP-0.9)
13:41jfalempe: mlankhorst: interesting setup, but yes I think we should see 2 panics on each half of the screen.
13:50mripard: jfalempe: you should post that one on reddit for karma :)
13:51jfalempe: mripard: this one looks classic, I prefer this one: https://www.reddit.com/r/linux/comments/1o5byvt/no_one_told_me_kernel_panics_could_be_diagonal/
13:52mlankhorst: Strrride
14:51mlankhorst: I used to debug with a usb-to-serial output, wonder if it would still work for the panic
16:01lucaceresoli: mripard: if that makes sense with your priorities, I'd like to ping about these two small series fixing drm-misc-next regressions:
16:01lucaceresoli: https://lore.kernel.org/dri-devel/20251014-drm-bridge-alloc-imx-ipuv3-v1-0-a1bb1dcbff50@bootlin.com/
16:01lucaceresoli: (leftover conversion to devm_drm-bridge_alloc)
16:02lucaceresoli: https://lore.kernel.org/dri-devel/20251017-drm-bridge-alloc-getput-bridge-connector-fix-hdmi_cec-v2-0-667abf6d47c0@bootlin.com/
16:02lucaceresoli: (actual bug in my previous patch, and the fixed version is even simpler than the buggy one)
16:02lucaceresoli: lumag: the 2nd series perhaps for you as you had reviewed the initial patch
18:14jadahl: emersion: yea, i'll get back to that soon
19:22alyssa: jenatali1: Oh foo.
19:22alyssa: My "just typeof" util_dynarray_append plan won't work
19:22alyssa: we have code in NIR like:
19:22alyssa: `util_dynarray_append(&dynarray, uint16_t, 0)==
19:22alyssa: er
19:22alyssa: `util_dynarray_append(&dynarray, uint16_t, 0)`
19:22alyssa: but sizeof(typeof(0)) != sizeof(uint16_t)!
19:23jenatali1: Oh, literals, sure
19:23alyssa: not sure if I want to press on anyway or not
19:24jenatali1: Seems useful to have an automatic variant of it even if you can't convert everything
19:24alyssa: Yeah, maybe
19:24jenatali1: Not sure if you can coccinelle a conversion that ignores literals?
19:25alyssa: I mean I'd just sed first and then coccinelle back the stuff that's safe
19:25alyssa: (_append_typed or something)
19:25alyssa: Honestly I'm a little tempted to add a more ergonomic container to src/uitl/ altogether
19:25alyssa: because util_dynarray sure ain't it.
19:26jenatali1: Yeah, one that generates typed functions maybe?
19:26jenatali1: That's one place where C++ really shines
19:26alyssa: Yeah..
19:28alyssa: yeah, what I actually want is templates :/
19:29HdkR: Oh no, tainted by C++
19:30alyssa: HdkR: this is your fault, you know ;)
19:30HdkR: I'll take full responsibility and pay alimony
19:32alyssa: rofl
19:32alyssa: I had a thought so googled it and found lots of people had the same thought
19:33alyssa: https://github.com/NickHackman/C_Vector/blob/master/src/vector.h
19:34alyssa: maybe we should import something like thaqt
19:46alyssa: regardless I'll post the dynarray patch, it's a win now
20:05alyssa: jenatali1: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/38038 there we go
20:05alyssa: this is gonna accumulate conflicts fast >.,
20:07jenatali1: alyssa: consider it acked but I just started leave, my wife had surgery this morning so I need to stay home for a few weeks
20:07alyssa: jenatali1: oh no! best wishes, hope her recovery is smooth
20:08jenatali1: Thanks!
20:13simon-perretta-img: alyssa: Would `util_dynarray_append(&dynarray, (uint16_t){0})` not do the trick?
20:14alyssa: simon-perretta-img: I mean sure.
20:14alyssa: Is that actually better than what I did that in MR? shrug.
20:14alyssa: Harder to coccinelle/sed, that's for sure
20:15simon-perretta-img: Fair
21:18isaacryu: Hey, wheres the best place to ask a question about mesa NIR?
21:22HdkR: isaacryu: Likely here.