[pkg-nvidia-devel] Bug#497077: [Patch 6/8] split libcuda1* packages from nvidia-glx* packages

Andreas Beckmann debian at abeckmann.de
Sat Jul 4 11:39:48 UTC 2009

tags 497077 + patch


here I propose a patch for the separation of libcuda1 from the
nvidia-glx packages I suggested some time ago.

This is useful for CUDA development and usage on headless systems (e.g.
using Tesla cards) that don't have or need X, GLX, ...

As names I choose libcuda1, libcuda1-ia32, libcuda-dev. I don't think we
need a vendor prefix nvidia- because I don't expect anyone else to
provide alternative implementations.

There are no dependencies from nvidia-glx on libcuda1 as it was also
done this way when nvidia-libvdpau was separated.

The libcuda.so symlink needs to go directly into the libcuda1 package,
not a -dev package because libcudart.so.1 (from nvidia-cuda-toolkit, not
packaged yet, but I plan to do an ITP soon) does dlopen("libcuda.so").

The patch contains also appropriate lintian-overrides and symbols files
for the new packages.

After applying the patch, debian/control has to be regenerated via
debian/rules clean.

The patch series 1/8 to 6/8 compiles successfully in pbuilder
environments sid/amd64, sid/i386, squeeze/amd64, lenny/amd64, lenny/i386

-------------- next part --------------
A non-text attachment was scrubbed...
Name: nvidia-graphics-drivers-6_of_8-split-libcuda1.v1.patch
Type: text/x-patch
Size: 15137 bytes
Desc: not available
URL: <http://lists.alioth.debian.org/pipermail/pkg-nvidia-devel/attachments/20090704/96153183/attachment-0001.bin>

More information about the Pkg-nvidia-devel mailing list