nvidia-cuda-toolkit 6.5 transition

Andreas Beckmann anbe at debian.org
Tue Sep 8 18:21:25 UTC 2015

On 2015-09-08 19:32, Tomasz Rybak wrote:
> Thanks for information and patches.
> There is possibility to just disable building CUDAND bindings.
> Would it make more sense to just have something like this in d/rules:
> override_dh_auto_configure:
> ifeq ($(DEB_BUILD_ARCH_CPU), i386)
> 	./configure.py --boost-python-libname=boost_python-py \
> 		--boost-thread-libname=boost_thread \
> 		--cuda-root=/usr --cuda-enable-gl --no-cuda-enable
> -curand
> else
> 	./configure.py --boost-python-libname=boost_python-py \
> 		--boost-thread-libname=boost_thread \
> 		--cuda-root=/usr --cuda-enable-gl
> endif

simplified and using the correct variable:

	./configure.py --boost-python-libname=boost_python-py \
		--boost-thread-libname=boost_thread \
		--cuda-root=/usr --cuda-enable-gl \
		$(if $(filter $(DEB_HOST_ARCH),i386),--no-cuda-enable-curand)

> instead of disabling entire architecture?
> I'd also like to add Debian.NEWS entry describing this change.
> BTW - what's the procedure now? You upload new CUDA (6.5) to unstable

after we got a transition slot

> and all  packages using it should be rebuilt?
> I assume that rebuilding PyCUDA will also take care of rebuilding
> it against new GCC, new Boost, and will use Sphinx 1.3.

yes, whatever is in unstable at that point

> At the same time - there is new upstream release which I'd like
> to package.
> So if you don't mind, I can take care about PyCUDA.

That's fine.


More information about the pkg-nvidia-devel mailing list