[BUG]: Doc build generates lots of ignored exceptions

[cpcloud]

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

Silent Failure

Component

cuda.pathfinder

Describe the bug

It seems that doc build are generating a number of exceptions that are ignored.

Here's a run on main: https://github.com/NVIDIA/cuda-python/actions/runs/17918838071/job/50949130232

How to Reproduce

N/A

Expected behavior

Docs build without any exceptions.

Operating System

Linux

nvidia-smi output

No response

[leofang]

xref https://github.com/NVIDIA/cuda-python/actions/runs/17918838071/job/50949130232#step:20:927

[rwgk]

@cpcloud is this what you saw (and had in mind)?

2025-09-22T14:49:43.4618681Z WARNING: autodoc: failed to import attribute 'cudaAtomicOperationCapability.cudaAtomicCapabilitySigned' from module 'cuda.bindings.runtime'; the following exception was raised:
2025-09-22T14:49:43.4619446Z Traceback (most recent call last):
2025-09-22T14:49:43.4620042Z   File "/home/runner/miniconda3/envs/cuda-python-docs/lib/python3.12/site-packages/sphinx/util/inspect.py", line 397, in safe_getattr
2025-09-22T14:49:43.4620639Z     return getattr(obj, name, *defargs)
2025-09-22T14:49:43.4620893Z            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-09-22T14:49:43.4621385Z AttributeError: type object 'cudaAtomicOperationCapability' has no attribute 'cudaAtomicCapabilitySigned'
2025-09-22T14:49:43.4621810Z 
2025-09-22T14:49:43.4621992Z The above exception was the direct cause of the following exception:
2025-09-22T14:49:43.4622262Z 
2025-09-22T14:49:43.4622350Z Traceback (most recent call last):
2025-09-22T14:49:43.4622980Z   File "/home/runner/miniconda3/envs/cuda-python-docs/lib/python3.12/site-packages/sphinx/ext/autodoc/importer.py", line 207, in import_object
2025-09-22T14:49:43.4623780Z     obj = attrgetter(obj, mangled_name)
2025-09-22T14:49:43.4624027Z           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-09-22T14:49:43.4624611Z   File "/home/runner/miniconda3/envs/cuda-python-docs/lib/python3.12/site-packages/sphinx/ext/autodoc/__init__.py", line 343, in get_attr
2025-09-22T14:49:43.4625400Z     return autodoc_attrgetter(self.env.app, obj, name, *defargs)
2025-09-22T14:49:43.4625744Z            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-09-22T14:49:43.4626391Z   File "/home/runner/miniconda3/envs/cuda-python-docs/lib/python3.12/site-packages/sphinx/ext/autodoc/__init__.py", line 2835, in autodoc_attrgetter
2025-09-22T14:49:43.4627040Z     return safe_getattr(obj, name, *defargs)
2025-09-22T14:49:43.4627305Z            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2025-09-22T14:49:43.4627876Z   File "/home/runner/miniconda3/envs/cuda-python-docs/lib/python3.12/site-packages/sphinx/util/inspect.py", line 413, in safe_getattr
2025-09-22T14:49:43.4628463Z     raise AttributeError(name) from exc
2025-09-22T14:49:43.4628749Z AttributeError: cudaAtomicCapabilitySigned
2025-09-22T14:49:43.4629032Z  [autodoc.import_object]

These are under

cd ../../cuda_bindings/docs

I looked systematically, this is conclusive: Only the the cuda_bindings doc build generates Tracebacks.

The other three doc builds (cuda_python, cuda_core, cuda_pathfinder) don't have any Tracebacks.

I'll change the Component in the issue description and the Label to cuda.bindings

@leofang Do you want to leave this assigned to me? (Happy to look into it if that's a priority.)

[leofang]

@rwgk see the above link to the CI log that I left earlier. It points to warnings raised during pathfinder doc build. Don't try to fix cuda-bindings warnings, they are very convolved (#152).

[cpcloud]

I opened this issue precisely because of #152. I don't see any place where we definitively state that we can ignore these errors because of some specific reason, or conversely where there's an actual bug that needs to be fixed.

If we can ignore them, then we should suppress them to avoid the noise. My hunch is that we cannot ignore these. It's almost never the case that exceptions during doc builds are intentional.

[rwgk]

@rwgk see the above link to the CI log that I left earlier. It points to warnings raised during pathfinder doc build. Don't try to fix cuda-bindings warnings, they are very convolved (#152).

Ah, got it now, I missed that link before:

/home/runner/work/cuda-python/cuda-python/cuda_pathfinder/docs/source/release/1.2.0-notes.rst:4: WARNING: duplicate object description of cuda.pathfinder, other instance in release/1.1.0-notes, use :no-index: for one of them
/home/runner/work/cuda-python/cuda-python/cuda_pathfinder/docs/source/release/1.2.1-notes.rst:4: WARNING: duplicate object description of cuda.pathfinder, other instance in release/1.2.0-notes, use :no-index: for one of them
/home/runner/work/cuda-python/cuda-python/cuda_pathfinder/docs/source/release/1.2.2-notes.rst:4: WARNING: duplicate object description of cuda.pathfinder, other instance in release/1.2.1-notes, use :no-index: for one of them
/home/runner/work/cuda-python/cuda-python/cuda_pathfinder/docs/source/contribute.rst:14: WARNING: Bullet list ends without a blank line; unexpected unindent. [docutils]
/home/runner/work/cuda-python/cuda-python/cuda_pathfinder/docs/source/release/1.2.3-notes.rst:4: WARNING: duplicate object description of cuda.pathfinder, other instance in release/1.0.0-notes, use :no-index: for one of them

I'll change the Component/Label back to cuda.pathfinder