Sadness.

I had to package the 12.8 version of cuda-bindings, because we're on cuda 12.8.
However, it dlopens libcuda.so and the driver (on my system at least) tells it that it uses cuda 13.1.
Hence, it errors out :/

https://github.com/NixOS/nixpkgs/pull/482638

(cuda-bindings is supposed to be required in the new torch update, not sure how critical it is though)

Here's the stack trace when I re-enable the failing test:

______________________ ERROR collecting test_nvjitlink.py ______________________
tests/test_nvjitlink.py:62: in <module>
    not check_nvjitlink_usable(), reason="nvJitLink not usable, maybe not installed or too old (<12.3)"
        ^^^^^^^^^^^^^^^^^^^^^^^^
tests/test_nvjitlink.py:58: in check_nvjitlink_usable
    return inner_nvjitlink._inspect_function_pointer("__nvJitLinkVersion") != 0
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
cuda/bindings/_internal/nvjitlink.pyx:254: in cuda.bindings._internal.nvjitlink._inspect_function_pointer
    ???
cuda/bindings/_internal/nvjitlink.pyx:257: in cuda.bindings._internal.nvjitlink._inspect_function_pointer
    ???
cuda/bindings/_internal/nvjitlink.pyx:205: in cuda.bindings._internal.nvjitlink._inspect_function_pointers
    ???
cuda/bindings/_internal/nvjitlink.pyx:99: in cuda.bindings._internal.nvjitlink._check_or_init_nvjitlink
    ???
cuda/bindings/_internal/nvjitlink.pyx:57: in cuda.bindings._internal.nvjitlink.load_library
    ???
cuda/bindings/_internal/utils.pyx:137: in cuda.bindings._internal.utils.get_nvjitlink_dso_version_suffix
    ???
E   cuda.bindings._internal.utils.NotSupportedError: only CUDA 12 driver is supported
------------------------------- Captured stdout --------------------------------
+++++++++++++++ 13010