Ilia Mirkin

I don't think I'm aware of how to access that FTP service, or what to do with the trace once I have it. Note that I won't be able to...

If you add Xnine tests or some other easy way for me to test, I'd be happy to try implementing this in nouveau.

Yeah, so unfortunately nouveau does a sort of shitty job with this. Setting half_pixel_center = 0 makes the frag coord returned also be 0,0. However nouveau only advertises frag coord...

Just hacked this up quickly: https://github.com/imirkin/mesa/commit/86ee2b612f320ba82e7d69e3bfc8bb79fa35350f.patch Doesn't seem to break the universe, but there are no Xnine tests that use the frag coord.

BTW, here's a theoretical fix for the depth bias issue that siro tried to hack around. Untested. https://github.com/imirkin/mesa/commit/cd4b868f77c629b77118d769c5781206a2c208d9.patch

Does this version instead help? Noticed a minor potential issue in the original: https://github.com/imirkin/mesa/commit/8ded8eadb748e6a9a6e4d3e934eeb791ad9292be.patch

This looks a lot like https://bugs.freedesktop.org/show_bug.cgi?id=91526 -- confirm that this happened on a build that had the patch in question. Perhaps there are more messed up formats.

Sounds like nine doesn't properly handle missing formats in drivers. FWIW the problem with BGRA4 was weird -- it did seem to _mostly_ work, but... in some cases, it didn't....

I've reintroduced BGRA4 on nvc0 in http://cgit.freedesktop.org/mesa/mesa/commit/?id=d68226087cf5f2f686d6c8f3377c5a1dec3d8bc4 . Figured out how to make the other trace work.

The problem here, BTW, is that priority is determined based on analyzing the list of `ws.ConditionalFormatting`. But this won't count the "x14" cfrules. So we have to re-parse it. There's...