09:13 <kusma> jekstrand: Uuuh, I have some indication that ANV actually *alpha-blends* the clear color set in the render-pass clear values instead of just writing them verbatim when we use B10G10R10A2... Do you have any hint for how to debug this?

11:44 <kusma> Actually, after some debugging, I think I've concluded that what I'm seeing is actually a bug in the compositor I'm using, and not in mesa...

15:10 <jekstrand> :D

15:19 <daniels> even better, it's a bug in mesa/demos

15:20 <daniels> I really cannot recommend using them for anything ever

15:20 <zmike> not even benchmarking?!

15:20 <daniels> hey, I can only recommend on issues of reference correctness

15:20 <daniels> you're the benchmarking dude

15:21 <zmike> no, now I'm the correctness dude

15:23 <ajax> yeah i've never had glxinfo tell me anything i want to know

15:25 <ajax> in the sense of the old joke about if you play a windows 95 install cd backwards you get a satanic message, but what's worse if you play it forwards it installs windows

15:29 <daniels> GLX_ARB_evaluate_choices_which_brought_you_to_this_point

15:29 <ajax> i blame 3dfx, i really do

16:00 <kusma> daniels: And it will stay that way unless we fix things!

16:01 <kusma> Thanks for steering me in the right direction ;)

16:04 <kusma> Oh gosh, we even have two OpenVG tests in mesa-demos! And no OpenVG implementation any more!

16:04 <kusma> Three, even!

16:06 <daniels> haha

16:06 <daniels> yeah, eglut is borderline unfixable because of the way it's structured

16:07 <daniels> I made it a little less broken on Wayland, but you can't change much of it without changing the API, which also requires changing all the apps and ... easier to just point them somewhere less broken?

16:07 <kusma> You mean stuff like the event loop making the system wanting to kill the process all the time?

16:08 <daniels> oh, I'd not seen that bit, but that's also cool

16:08 <kusma> At least that's what I assume the periodic "program not responding, wanna take it out behind the shed"-dialogs are about

16:08 <daniels> oh

16:08 <daniels> that would be because it doesn't respond to ping events

16:09 <kusma> Sounds likely

17:09 <daniels> jesus, it still uses wl_shell as well

17:09 <daniels> that won't work on any wlroots compositors, doesn't work on Weston unless you explicitly enable the deprecated-wl-shell option (until we remove it which is soon) ... I'm surprised it works on Mutter tbh