02:30fdobridge: <airlied> oh avoiding peephole select is also something that fixes it
02:30fdobridge: <airlied> this is going to be something horrible
02:35fdobridge: <airlied> @georgeouzou okay pushed a workaround, hopefully should handle all tests 🙂
07:45daxe: karolherbst: thanks, I'm on mesa 23.1.3. Might just try playing around with debug prints and see if I can find anything interesting.
09:04karolherbst: you probably won't find much as the fault happens insider a shader and those things are super hard to track down currently. I have an EVoC student who will work on figuring out how to run shader trap handlers, so maybe in the near future we'll be able to dump what shader did fault at which instruction.
09:04karolherbst: daxe: but in your case it might also be simply a result of the previous errors
09:28daxe: when you say previous errors are you referring to the "Discord" line?
09:43karolherbst: well.. most of the errors before that
11:10karolherbst: anybody up for reviewing https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/23802 ?
21:15fdobridge: <georgeouzou> @airlied
21:15fdobridge: <georgeouzou> @airlied after the latest changes i get 4 fails in tessellation tests dEQP-VK.tessellation.geometry_interaction.point_size* (edited)
21:17fdobridge: <georgeouzou> If i enable nir_opt_peephole_select they pass, if not then nir_opt_if makes them fail.
21:33fdobridge: <airlied> great, so I guess need to figure out why peephole kills things