https://bugs.winehq.org/show_bug.cgi?id=51541
--- Comment #3 from Zebediah Figura z.figura12@gmail.com --- (In reply to Sveinar Søpler from comment #2)
That would be dependent on libcuda being installed at configure-time tho won't it? Probably not too hard to achieve...
That's the standard for libraries. In general the idea is that you compile on the same distribution that you distribute the package for, with the assumption that the library will be located in the same place.
Though it's really worth examining whether we really want nvcuda in upstream Wine; and, if not, we should remove it from wine-staging.
There are many usecases for having this upstream tho. What is the best argument for scratching this altogether?
How many programs need CUDA specifically, and can't use OpenCL or some other library we already provide?