00:41 <zmike> anholt: you may be interested in reviewing !18689 if you get time

16:47 <zmike> ajax: I put up the MR for the resolve target thing, but I'm unsure whether it's actually needed (for zink) since I've got a deeper series working through kopper to handle linking up swapchains+msaa attachments

16:48 <zmike> and if I'm not going to use it, then it seems improbable anyone else will

16:55 <ajax> i kind of don't understand why the frontend allocates the msaa surface

16:55 <zmike> as opposed to?

16:56 <ajax> well, not the winsys. the display server never sees the ms buffer

16:57 <zmike> I'd assume it's so that postprocessing works

16:57 <zmike> because of how hud is implemented

17:33 <zmike> ajax: great news: I think I've finally solved https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16168

17:35 <zmike> or at least I've finally managed to repro it locally and I fixed it here

17:52 <ajax> that... looks pretty credible

17:52 <ajax> give me a minute to finish typing a thing and i'll review a bit closer

17:54 <zmike> it was just missing the GALLIUM_DRIVER check in drisw

17:54 <zmike> oh ffs what now

17:54 <zmike> wtf

18:05 <daniels> [narrator: he had not]

18:09 <zmike> aha!

18:10 <zmike> yesssssssssssssssssss

18:10 <zmike> #nailedit

18:10 <zmike> line 666 too

18:10 <zmike> fitting

18:38 <zmike> hm

18:38 <zmike> somehow this fixed glx-tfp for virgl?

18:43 <zmike> ah

18:45 <ajax> istr that test is bunk to begin with

18:45 <zmike> nah it's unrelated

18:47 <zmike> okay THIS time I got it

18:47 <zmike> for real

18:52 <Sachiel> such an incredible capability for self delusion

19:29 <zmike> ci is G R E E N

19:29 <zmike> it's a miracle

19:30 <Sachiel> or it's mold

19:31 <zmike> or both

19:32 <Sachiel> mold, the miracle of life

19:32 <zmike> love to see it in action

20:30 <zmike> ajax: here's a weird one: https://gitlab.freedesktop.org/mesa/mesa/-/issues/7165

20:30 <zmike> I think what's happening is the app is calling swapbuffers in sequence such that it's trying to swap to an uninitialized buffer

20:31 <zmike> not exactly sure what the expected behavior is here

20:31 <zmike> is doing nothing a viable option?

20:32 <zmike> wondering if maybe that's what I should've done in flush_frontbuffer instead of trying to copy the last frame over

20:43 <ajax> it's calling swapbuffers twice in sequence without drawing anything between?

20:44 <zmike> so it seems

20:44 <zmike> I'm remotely debugging by printf

20:45 <ajax> o_O

20:46 <zmike> some kind of super race condition

20:48 <ajax> well. that should acquire a new back buffer, draw nothing into it, and present it.

20:48 <ajax> i don't think doing nothing is an option since it'd mess with buffer age

20:49 <zmike> hm

20:51 <daniels> correct