RAOF | <RAOF> "*Today*:..." <- > <@chris:cooperteam.net> *Today*:... (full message at <https://libera.ems.host/_matrix/media/v3/download/libera.chat/0e4baf092c7744e269425eeeb96df08b7196c622>) | 07:26 |
---|---|---|
Saviq | So isn't a amdgpu bug in the end? I suppose that's a better outcome, as we can hopefully fix, rather than work around? | 07:55 |
RAOF | Yeah. | 07:56 |
RAOF | It is something extremely strange. | 07:56 |
RAOF | (It might be a radeonsi bug instead 😉) | 07:56 |
RAOF | But, from further inspection, it looks like it's allocating a buffer that's scanoutable (and untiled), but it's rendering to it as if it's tiled. | 07:57 |
RAOF | So, not a kernel bug but a mesa bug? 🎉 | 07:57 |
RAOF | But, our current codepath doesn't hit this. | 07:57 |
RAOF | …but, while we're calling into the driver in a different order and from different files, the actual calls that we make seem to be the same? | 07:59 |
RAOF | So, today: | 08:00 |
RAOF | * Fighting with Mesa debug symbols (the debuginfod integration can't pull source files, apparently because the referenced source files *never exist*?) | 08:00 |
RAOF | * Chasing incorrect output on AMD. | 08:00 |
Saviq | <alan_g[m]> "Today:..." <- > <@alan_g:matrix.org> Today: | 11:41 |
Saviq | > * found a rabbit hole: https://github.com/Miriway/Miriway/issues/51 | 11:41 |
Saviq | Did you find an exit, BTW? Sophie and I've been banging at the GTK DBus delay a couple times, but with no real solution | 11:41 |
alan_g[m] | Saviq: Yes, down a different hole today (trying to understand the Python GTK abstractions well enough to break some of the encapsulation) | 11:43 |
Saviq | A'ight, today: | 16:15 |
Saviq | - dealt with a bunch of corner cases in checkbox-mir runs in the lab | 16:15 |
Saviq | - but have a working plan for Nvidia | 16:15 |
Saviq | - also have a plan for logind-based runs | 16:15 |
Saviq | 👋 until tmrw | 16:15 |
alan_g[m] | Today a new rabbit hole | 17:28 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!