01:09 HdkR: Is there a test program for xdamage somewhere? Getting some weird spurious failure and I'm not sure why.
01:11 HdkR: Getting some `invalid Damage parameter` things and at a loss for how to track :D
02:25 alanc: there is an ancient & unmaintained https://gitlab.freedesktop.org/xorg/app/xshowdamage but it's been decades since I remember running it and don't know how much it will help
02:56 HdkR: weee, spicy
14:49 alyssa: dschuermann: so with loop_simplify, does it make sense to drop `continue` handling from backends?
14:50 alyssa: IIRC handling continues on AGX was kinda annoying, and Apple just requires llvm loopsimplify
14:50 alyssa: probably less interesting for Intel though.
14:52 alyssa: (read: could delete a crapton of code from src/asahi/compiler if we required loopsimplify form, and potentially improve code gen too)
14:53 alyssa: (and a smaller amount of code from Intel)
14:55 alyssa: (Intel has hardware CONTINUE instructions so .. lol.)
15:01 alyssa: I guess add a patch ripping out continue support from aco and see what happens? lol
15:01 alyssa: (need to also call loopsimplify in mesa/st for radeonsi I guess)
15:06 alyssa: could also clean up NIR if we have that as contract
15:06 alyssa: e.g. divergence analysis has special continue handling
15:10 alyssa: unless I'm missing the point entirely. It's been known to happen
22:31 dcbaker: zdobersek: I've got some tests on for the zink on turnip job for the a750, with the 25.3 branch that start passing somewhere between here: https://gitlab.freedesktop.org/mesa/mesa/-/commit/af0845c77cdce00999458f79560f59947c1228ab/pipelines?ref=staging%2F25.3 and https://gitlab.freedesktop.org/mesa/mesa/-/commit/5effc063d98fe04e6e5db0e09122972ae0e8135f/pipelines?ref=staging%2F25.3. The only turnip patch I see in that regions is this one:
22:31 dcbaker: https://gitlab.freedesktop.org/mesa/mesa/-/commit/5da94d1c6ec4734ff020207e09c1ad42469d6da9, any chance that your patch fixed those tests?
23:30 robclark: dcbaker: I suppose zdobersek would want to know which tests.. if you have link to a job
23:31 robclark: but also worth checking if those tests started passing in nightly full runs