Reported to openSUSE bug 879767
Just noting that NVIDIA itself maintains the repo. So report may need to go to NVIDIA also
But the packages are built by openSUSE.
If the bug cannot be reproduced with the same driver version from the nvidia homepage, there’s no need to report it to NVIDIA I suppose.
Indeed no need to report to nvidia
Huzzah! I have CUDA support back with the repo 331.67 driver(oS 13.1), thanks to a thread on the NVIDIA support site.
First, another package in the repo is needed - nvidia-uvm-gfxG03-kmp-desktop (this may need a reboot - I did several during this process). Then I ran these commands (as root)
cloyd:/home/gwb # mknod -m 0660 /dev/nvidia-uvm c $(awk '/nvidia-uvm/{print $1}' /proc/devices) 0
cloyd:/home/gwb # chgrp video /dev/nvidia-uvm
cloyd:/home/gwb # ls -l /dev/nvidia-uvm
crw-rw---- 1 root video 249, 0 May 25 09:13 /dev/nvidia-uvm
Note: this assumes the CUDA user is in group ‘video’; for real use, it should have the same ACL mods as the other nvidia devices. It also will NOT survive a boot, so must be executed after each.
After the above, einstein@home is happily computing on my GT-640.
On 2014-05-24 22:36, gogalthorp wrote:
>
> Just noting that NVIDIA itself maintains the repo. So report may need to
> go to NVIDIA also
No, the repo is maintained by openSUSE people; Nvidia solely hosts the
repo, because for legal reasons openSUSE can not.
There is a readme file somewhere in the repo that explains the
situation, or part of it. Nvidia explicitly says that they refuse to
listen to any bug report against the rpm packages. Any bug reporter has
to use the bundles and installer supplied directly by Nvidia, ie,
install the hard way, reproduce the problem (or not), then they can
report to NVidia.
They also say that for problems with the rpms, you have to report
directly to SUSE instead.
ftp://download.nvidia.com/opensuse/README
Notice that you must use FTP to get to it.
–
Cheers / Saludos,
Carlos E. R.
(from 13.1 x86_64 “Bottle” at Telcontar)