Is it intended to exlude gcc-4.7 from building nvidia-cuda-toolkit ? I don't find the hint about this in the bugs. I am stuck at dev-util/nvidia-cuda-toolkit-4.2 because of this ... Is there a plan to unmask gcc-4.7.x in the future? I am using this compiler for two months now without issues - but these alike.
Try setting USE=multislot for sys-devel/gcc. Otherwise, this bug looks invalid: you're not stuck at all and the deps are probably correct.
dev-util/nvidia-cuda-toolkit is a binary package so we can not do much about its dependencies. My guess is that dev-util/nvidia-cuda-toolkit-5* will work with gcc:4.7, however as gcc:4.7 is not even unmasked yet, I would leave this bug open for later.
@Christoph, it installs using gcc-4.7.2 (without multi-gcc), but this message appears: * Messages for package dev-util/nvidia-cuda-toolkit-5.0.35-r1: * gcc >= will not work with CUDA * Make sure you set an earlier version of gcc with gcc-config * or append --compiler-bindir= pointing to a gcc bindir like * /usr/*pc-linux-gnu/gcc-bin/gcc * to the nvcc compiler flags >>> Auto-cleaning packages... Wanting to have another gcc to work with?
This warning is a sanity check by us. Edit your /usr/bin/cuda-config, so that it support gcc:4.7 by adding 4.7 to SUPPORT_GCC_VERSIONS_BY_CUDA. And then emerge a package depending using cuda to see if it actually works. @Justin: It seems the warning should use ${a[0]} and not ${a[1]}. Are you a fortran guy? ;-)
There was some bad logic mistake in that section. +*nvidia-cuda-toolkit-5.0.35-r2 (22 Jan 2013) +*nvidia-cuda-toolkit-4.2.9-r2 (22 Jan 2013) + + 22 Jan 2013; Justin Lecher <jlec@gentoo.org> + +nvidia-cuda-toolkit-4.2.9-r2.ebuild, +nvidia-cuda-toolkit-5.0.35-r2.ebuild: + Correct postinst msg logic; fine grade deletion of files +