Fix #299: Error message about bindings not working

[gmarkall]

To avoid a ModuleNotFoundError being raised when there is no cuda namespace package installed, we check for a spec for it first before checking for cuda.bindings.

[copy-pr-bot]

Auto-sync is disabled for ready for review pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.

[gmarkall]

/ok to test

[gmarkall]

A check on my machine when cuda and cuda.bindings do not exist:

$ python -m numba.runtests numba.cuda.tests
Traceback (most recent call last):
  File "<frozen runpy>", line 198, in _run_module_as_main
  File "<frozen runpy>", line 88, in _run_code
  File "/home/gmarkall/miniforge3/envs/test-numba-cuda-0151-wheel/lib/python3.12/site-packages/numba/runtests.py", line 9, in <module>
    sys.exit(0 if _main(sys.argv) else 1)
                  ^^^^^^^^^^^^^^^
  File "/home/gmarkall/miniforge3/envs/test-numba-cuda-0151-wheel/lib/python3.12/site-packages/numba/testing/_runtests.py", line 25, in _main
    return run_tests(argv, defaultTest='numba.tests',
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/gmarkall/miniforge3/envs/test-numba-cuda-0151-wheel/lib/python3.12/site-packages/numba/testing/__init__.py", line 54, in run_tests
    prog = NumbaTestProgram(argv=argv,
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/gmarkall/miniforge3/envs/test-numba-cuda-0151-wheel/lib/python3.12/site-packages/numba/testing/main.py", line 204, in __init__
    super(NumbaTestProgram, self).__init__(*args, **kwargs)
  File "/home/gmarkall/miniforge3/envs/test-numba-cuda-0151-wheel/lib/python3.12/unittest/main.py", line 105, in __init__
    self.runTests()
  File "/home/gmarkall/miniforge3/envs/test-numba-cuda-0151-wheel/lib/python3.12/site-packages/numba/testing/main.py", line 376, in runTests
    run_tests_real()
  File "/home/gmarkall/miniforge3/envs/test-numba-cuda-0151-wheel/lib/python3.12/site-packages/numba/testing/main.py", line 361, in run_tests_real
    super(NumbaTestProgram, self).runTests()
  File "/home/gmarkall/miniforge3/envs/test-numba-cuda-0151-wheel/lib/python3.12/unittest/main.py", line 281, in runTests
    self.result = testRunner.run(self.test)
                  ^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/gmarkall/miniforge3/envs/test-numba-cuda-0151-wheel/lib/python3.12/site-packages/numba/testing/main.py", line 167, in run
    run.sort(key=cuda_sensitive_mtime)
  File "/home/gmarkall/miniforge3/envs/test-numba-cuda-0151-wheel/lib/python3.12/site-packages/numba/testing/main.py", line 82, in cuda_sensitive_mtime
    from numba.cuda.testing import CUDATestCase
  File "/home/gmarkall/numbadev/numba-cuda/numba_cuda/numba/cuda/__init__.py", line 37, in <module>
    raise ImportError(
ImportError: CUDA bindings not found. Please pip install the cuda-bindings package. Alternatively, install numba-cuda[cuXY], where XY is the required CUDA version, to install the binding automatically. If no CUDA bindings are desired, set the env var NUMBA_CUDA_USE_NVIDIA_BINDING=0 to enable ctypes bindings.

[gmarkall]

I do wonder if there could be other sources of exceptions that we might not catch, from the info in https://docs.python.org/3/library/importlib.html#importlib.util.find_spec - for example, if cuda is a module in the environment (and not a package) perhaps ModuleNotFoundError could get raised by find_spec("cuda"). This might be more of a corner case / less likely to arise though.