The CUDA samples and some scientific libraries are longer provided by NVIDIA for x86-based linux distributions [1]: "(Linux) Support for 32-bit Applications on x86-based Linux Distributions Several portions of the CUDA Toolkit are no longer available for developing 32-bit applications on x86-based Linux distributions: [..] * CUDA Toolkit scientific libraries, including cuBLAS, cuSPARSE, cuFFT, cuRAND, and NPP [..] * CUDA Samples" I assume that installing the ebuild on an x86 platform will fail. Therefore I removed the ~x86 keyword before pushing my ebuild to the science overlay [2]. Unfortunately this changes didn't made it to portage. [1] http://developer.download.nvidia.com/compute/cuda/6_5/rel/docs/CUDA_Toolkit_Release_Notes.pdf [2] https://github.com/gentoo-science/sci/pull/257/files Reproducible: Always
+*nvidia-cuda-toolkit-6.5.19 (17 Nov 2014) + + 17 Nov 2014; Justin Lecher <jlec@gentoo.org> + nvidia-cuda-toolkit-6.5.14.ebuild, +nvidia-cuda-toolkit-6.5.19.ebuild: + Import Version Bump from sci overlay, thanks Marius Brehler for the work; Add + warning for reduced 32bit support, #529320 +
+*nvidia-cuda-sdk-6.5.19 (17 Nov 2014) + + 17 Nov 2014; Justin Lecher <jlec@gentoo.org> +nvidia-cuda-sdk-6.5.19.ebuild: + Import Version Bump from sci overlay, thanks Marius Brehler for the work; Add + warning for reduced 32bit support, #529320 +
(In reply to Marius Brehler from comment #0) > I assume that installing the ebuild on an x86 platform will fail. Therefore > I removed the ~x86 keyword before pushing my ebuild to the science overlay It doesn't fail, but it has reduced functionality. I add a warning and will observe what nivdia is doing in future. Thanks for your work.
Okay, if the ~x86 keyword is still supported, the SRC_URI to the x86 run file should be added again. That was my mistake, I removed it along with the keywords. Sorry for the circumstances.