[BUG]: Compiling the new `_nvml` files causes Cython to report a recursion depth error

[gflegar]

Is this a duplicate?

• I confirmed there appear to be no duplicate issues for this bug and that I agree to the Code of Conduct

Type of Bug

Compile-time Error

Component

cuda.bindings

Describe the bug

I'm getting this error with the internal Cython we have when compiling cuda/bindings/_internal/_nvml.pyx on v13.1.1 tag:

  File "cython/files/Cython/Compiler/Main.py", line 846, in main
    result = compile(sources, options)
             ^^^^^^^^^^^^^^^^^^^^^^^^^
...
File "cython/files/Cython/Compiler/Nodes.py", line 8629, in fresh_finally_clause
    node_copy = copy.deepcopy(node)
                ^^^^^^^^^^^^^^^^^^^
...
  File "/<embedded stdlib>/copy.py", line 221, in _deepcopy_dict
  File "/<embedded stdlib>/copy.py", line 136, in deepcopy
RecursionError: maximum recursion depth exceeded

(For now we're skipping compiling the _nvml files, as we do not use them, but still want to resolve this in case something starts depending on it at some point.)

How to Reproduce

Try to compile the file through Cython.

I'm not sure at this point if this is only an issue in our internal version of Cython, but wondering if Nvidia has seen this error and how you resolved it?

Expected behavior

N/A

Operating System

Custom linux-based OS

nvidia-smi output

N/A

[leofang]

Which Cython version are you using? You might be hitting this issue which is fixed in Cython 3.2 (which our build dependencies have migrated to): cython/cython#7122. If for any reason you need to use Cython <3.2, you can increase the system recursion limit in setup.py as blessed in that issue by a Cython maintainer.