04:05 karolherbst: imanho: the code should be close to your own functions though.. but maybe nvidia is cheating and prevents dumping code :D
04:52 anholt: swapped jetson back to kernel 5.15-rc4, which I'm pretty sure is what I was using last time, with no luck. also tried smashing in the firmware from the chroot that worked before, still no luck.
05:49 imirkin: karolherbst: anholt: have a look at https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14797 - want to push it tomorrow, object soon if you dislike.
05:50 anholt: imirkin: I'd recommend moving copy_image.functional to a skip for normal dev work, it's really slow.
05:50 imirkin: anholt: yeah, i guess i should move the timeouts to a skips file
05:50 anholt: post-xfb is not one i think I've seen before
05:51 anholt: draw_indirect is slow on other drivers, though I wouldn't have said "to the point of 60s timeout"
05:51 imirkin: but this is more like recording the results for now, and then we can sort stuff later
05:51 imirkin: yeah, the xfb ones are real
05:51 imirkin: i need to look into that.
05:55 imirkin: i'm wondering if it's worth trying to do something about the texgather thing for ES 3.1. annoyingly they want to be able to select components, which the hw doesn't do
05:58 imirkin: oh, and i bet KHR-GL33 doesn't run copy_image even though it probably could be made to
14:49 karolherbst: imirkin: actually nice to have those files in mesa... myabe I should set something up for other gens as well...
14:50 karolherbst: now I have a spare CPU and board
17:05 karolherbst: imirkin: I also see that not all DVI actually has EDID as well...
17:06 imirkin: mmmmm
17:06 imirkin: well, DVI-A which is VGA, deifnitely not
17:06 karolherbst: right
17:06 imirkin: but pretty sure all DVI-D displays should have it
17:06 imirkin: not 100% sure. but pretty sure.
17:06 karolherbst: apparently DDC1 doesn't do EDID
17:06 imirkin: LVDS can have it missing i think
17:07 imirkin: as can eDP. generally there's one in the vbios.
17:08 karolherbst: or ACPI or wherever one might to hide it
17:08 imirkin: right yea
19:09 Lyude: https://gitlab.freedesktop.org/drm/nouveau/-/issues/140#note_1238436 imirkin any chance you have any machines that are old enough to reproduce this?
19:09 imirkin: Lyude: i mean ... i have a GF108 plugged in right now
19:10 imirkin: but i never had any laptops with nvidia
19:10 imirkin: Lyude: that said i'm running an older kernel
19:10 imirkin: i could try a newer one
19:10 imirkin: did you not test on fermi though?
19:10 Lyude: imirkin: I'm certain I did, because I remember testing it - finding out it was broken - and then fixing before submit
19:11 Lyude: so I'm not sure what happened
19:11 Lyude: being able to see a decoded backtrace from the freeze would probably be enough to figure out what's happening though
19:11 imirkin: yea ok
19:12 imirkin: i'm on 5.6, which is from before
19:12 imirkin: my setup is slightly esoteric though, might not be able to trigger the issue
19:12 imirkin: since i have rotated monitors
19:12 imirkin: i effectively never get "page flipping"
19:12 Lyude: another LVDS bug too, that's weird
19:13 imirkin: Lyude: my concern is that the atomic_crc check _might_ end up doing work?
19:13 imirkin: not sure
19:13 Lyude: imirkin: ah, Ill have to look around and see if I've got access to one of these then if you can't reproduce it. I've already got a friend of mine trying to setup access to a fermi laptop for a different backlight bug on nouveau so maybe I can look at it at the same time
19:13 Lyude:will also just look through the code again to see if she can spot anything silly
19:14 imirkin: yeah, i mean it's all guarded behind the crc flags being set
19:14 imirkin: which i don't think they would be
19:14 imirkin: but for all i know something jams them to ~0 somewhere
19:14 imirkin: (i mean clr.crc / set.crc)
19:14 Lyude: mhm
19:32 karolherbst: Lyude: I could try really hard, but I don't think I hit any issues testing the fermis I've gotten
19:38 imirkin: karolherbst: i'm guessing page-flipping might be required
19:38 imirkin: the complaing wasn't that it always happened
19:38 imirkin: but rather only under certain circumstances
19:38 imirkin: like playing movies
19:39 karolherbst: yeah... I guess
21:44 karolherbst: imirkin: about to push the of by one patch, wanna add your r-by tag as well or should I just push?
21:44 imirkin: the nv30 one?
21:44 imirkin: (well, nv44 one)
21:44 imirkin: link? i can put r-b in the thing
21:45 karolherbst: https://patchwork.freedesktop.org/patch/470449/
21:45 imirkin: oh, that off-by-one
21:45 karolherbst: yeah
21:45 imirkin: you landed the nv44 one a long while back
21:45 imirkin: yes, r-b: me
21:48 karolherbst: (should push the patchwork stuff at some point as well)
21:54 imirkin: karolherbst: want me to respond with r-b?
21:54 imirkin: or you're good?
21:55 karolherbst: not sure if people care :D
21:55 imirkin: pretty sure people don't
21:55 karolherbst: I already pushed but if you want to reply that's fine as well
21:55 imirkin: esp for an obvious thing like that
21:55 karolherbst: yeah..
21:55 imirkin: just need proper S-o-b chain
21:55 karolherbst: Lyude: did you already push https://patchwork.freedesktop.org/patch/468141/ ? I guess not?
21:58 karolherbst: not sure if it's even worth caring though.. but I guess it is..
22:47 Lyude: karolherbst: nope, not that one