See #379, https://github.com/doitsujin/dxvk/issues/613#issuecomment-434619658. As far as I know, this is not a DXVK bug but some issue with various Nvidia libraries.
Actually, nevermind, this seems to be a separate issue. Looks like a RADV bug to me since it doesn't happen with AMDVLK.
That said, the apitrace is pretty much broken, not sure how much I'd trust it.
I have only noticed HUD problems on the trace, everything else is rendered properly, even the terrain bug on RADV, I used apitrace's dlls directly from its github page, it's there any fork that works better for this case?
Not really; sometimes it just bugs out and you can't do much about it.
Can you test whether amdvlk actually fixes the bug in-game on your end?
it seems that it only happens on RADV, AMDVLK and -PRO don't have the bug, where should I report bugs to mesa?
It's already reported:
https://gitlab.freedesktop.org/mesa/mesa/issues/851
Another bump likely wouldn't hurt though.
Wasn't that yet another problem or actually the nvenc/nvcuda/... bug?
What I see when replaying the apitrace in this report is that the terrain geometry is glitching randomly.
Just Cause 3 has some flickering terrain, changing settings in game makes no difference. This also happen on both LLVM and ACO compilers
The trace (recorded on WIN10) replays with some HUD errors for me but it reproduces the terrain bug
Software information
Just Cause 3, max settings
System information
Apitrace file(s)
Log files (using Wine's DXGI)