Post by Emil Velikov Post by Aaron Watry
We haven't landed the fixes to break the r600g dependency on AMDGPU yet.
I'm headed out of town for a long weekend and don't feel like risking the
push before being gone for five days.
I've got a v3 of Emil's patch 4/5 that removes the AMDGPU header dependency
from r600 and I'm good with the status of Jan's 3-patch series. I'm hoping
we can square that away early next week unless is gets resolved while I'm
I've double-checked and Jan's 1-3 (squashed 2+3) alongside my 4-5
resolve all the issues I could notice.
Pushed the lot and I'll parse through patchwork in a moment.
thanks for keeping tabs on this. I've been mostly sick/travelling past
2 weeks. I'll be fully online starting Wednesday.
The third patch of my series was separate, because it needed a bit more
baking (I posted it to shift the discussion away from deletion), but I
guess the few follow up bugs hit all the rough edges.
libelf is only needed for with-opencl configurations (so android should
be OK without this dep. see my response to Mauro). I'll post a patch
that fixes automake this week.
d5199c (Revert "amd/common: add missing libdrm include path") looks
incorrect. libamd_common still includes amdgpu.h so I'd say it needs
the CFLAGS. Moving the include to ac_gpu_info.c (in 81945) is enough to
remove libdrm_amdgpu dependency from r600g builds.
Jan Vesely <***@rutgers.edu>