16:58gfxstrand[d]: Uh... I think nouveau might have serious issues with duplicate BO mappings. I'll investigate more in a couple hours but this branch should NOT be blowing up like it is. 😬
21:42gfxstrand[d]: Actually, nothing in this patch series should be causing duplicate mappings. 🤔 I'm going to add stuff which does duplicate mappings but they shouldn't actually happen yet.
21:42gfxstrand[d]: Maybe I'm not cleaning up a map?
21:46gfxstrand[d]: Looks like it's just failing to create contexts because stuff was getting held open by maps I didn't clean up. I don't think it has anything to do with duplicate mappings. :frog_sweat:
21:49valentineburley[d]: Unrelated and not urgent, but do you have an idea why `dEQP-VK.ycbcr.plane_view` might be failing for plane1 on NVK with the new pipe formats? I first suspected my swizzles to be wrong but that didn't seem to be the cause
21:51gfxstrand[d]: Nothing jumps to mind. Do you have a patch or MR that I can try?
21:53valentineburley[d]: https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/30821
21:53valentineburley[d]: and by plane1 I meant on the 2 plane formats, so R10X6G10X6 and same for 12bit
21:56gfxstrand[d]: Swizzles were wrong. See comments.
21:57valentineburley[d]: That is what I had initially, and that was fairing even worse
21:59valentineburley[d]: I also messed with the swizzles in vk_format.c
22:01gfxstrand[d]: Hrm...
22:02gfxstrand[d]: Okay, let me pull it
22:45gfxstrand[d]: I think you may have found a preexisting bug
22:55gfxstrand[d]: Looks like swizzles are wrong in shaders for some reason
22:56gfxstrand[d]: Yeah, I think it's because you messed with the swizzles in the YCbCr table
23:09gfxstrand[d]: Yeah, I think that's it. Commented.
23:19gfxstrand[d]:fixes the barycentric tests because she's tired of their impact on CTS run times.
23:24gfxstrand[d]: I'm gonna make some tacos while that's running.
23:37mohamexiety[d]: those swizzles and tables were so confusing back then
23:39gfxstrand[d]: They really are
23:39gfxstrand[d]: gfxstrand[d]: I think getting rid of those spurious fails is going to get my CTS runs back under 50 min again. :frog_party: