As per summary, maybe it can be changed to use the system ones instead?
These are 32-bit Qt4 libraries. In order to avoid installing them in the ebuild, these libraries would have to be added to app-emulation/emul-linux-x86-qtlibs.
CC-ing the amd64 arch team to see whether there is any interest/plans for adding Qt4 to emul-linux-x86-qtlibs.
The amd64 project wants the emul packages to be as minimal as possible (they are a PITA to update). So if a package provide its own libs, we are more than happy with this solution.
Leaving beside for a moment the security considerations [1] as well as the fact that they have been proven more than theory [2], can somebody tell me why this is WONTFIX just because AMD64 team does not have time to add Qt4? Skype has the same problem (so it would really make sense to have them in emul-, becaues be it minimal or not, it's just silly to leave it at that, but I do know that nobody is interested in either maintaining that or fixing multilib once and for all), and it solves it in a more sensible way with an USE flag regarging static Qt. So at a minimum, cuda could use the system copy on x86 and its own on amd64 until emul- gets Qt4 (until, not until eventually, because it'll either happen or we're pretty screwed in Gentoo...). [1] http://blog.flameeyes.eu/2009/01/02/bundling-libraries-for-despair-and-insecurity [2] http://blog.flameeyes.eu/2009/01/09/seems-like-my-ideas-werent-far-off
These are proprietary applications, they can in no way be considered secure, so security is a moot point. If at some point Qt4 gets added to the emul libs, I guess you can re-add us. In the mean time, I see no case where this would be required. Remember that the emul libs are only for the few 32-bit only proprietary apps.
dev-util/nvidia-cuda no longer exists, there are few separate ebuilds in dev-util now: nvidia-cuda-toolkit, nvidia-cuda-profiler, nvidia-cuda-sdk. i guess that this problem affects nvidia-cuda-profiler now
libpng version in /opt/cuda/lib/libQtGui.so.4: 1.2.29
nvidia-cuda-profiler will be removed from the tree when CUDA 2.2 is stabilized. Since I have just committed a change which makes the profiler installed by nvidia-cuda-toolkit-2.2 use system Qt4 libraries on x86, I consider this bug fixed. Thanks for the comments and for reporting the issue.
