On 03/12/2012 09:30 PM, Dimitri Jakov wrote:
Should be fixed now. Please update and report your experience. Thanks!
Dimitri
In the mirrors there is only new nvidia-cuda-toolkit(-devel) packages,
but still require something not detected:
werewolf:~/in# rpm -U nvidia-cuda-toolkit-4.1.28-3.mga2.nonfree.x86_64.rpm
error: Failed dependencies:
libcuda.so.1()(64bit) is needed by
nvidia-cuda-toolkit-4.1.28-3.mga2.nonfree.x86_64
nvidia-cuda-toolkit = 4.1.28-1.mga2.nonfree is needed by (installed)
nvidia-cuda-toolkit-devel-4.1.28-1.mga2.nonfree.x86_64
werewolf:~/in# rpm -qf /usr/lib64/nvidia-current/libcuda.so.1
nvidia-current-cuda-opencl-295.20-2.mga2.nonfree
werewolf:~/in# rpm -q --provides
nvidia-current-cuda-opencl-295.20-2.mga2.nonfree
nvidia-current-cuda = 295.20-2.mga2.nonfree
nvidia-current-cuda-opencl = 295.20-2.mga2.nonfree
nvidia-current-cuda-opencl(x86-64) = 295.20-2.mga2.nonfree
The latter does not provide libcuda.so.1()(64bit). Should it be automatic
or not, it should show anyways in --provides ?
Perhaps the library location in /usr/lib{64}/nvidia-current makes rpm skip
the automatic provides for libs ?
TIA
--
J.A. Magallon <jamagallon()ono!com> \ Winter is coming...
