Napoleon extension gives unexpected error message when processing the docstring of a function written in Cython

[tovrstra]

Describe the bug
Our CI crashed with a rather cryptic error message when building the documentation, see https://travis-ci.com/github/theochem/iodata/jobs/370345521#L1842

Manual checks revealed this was related to the Napolean extension. (Commenting out napoleon from conf.py prevents the error message, but this is not a helpful solution.)

By adding a line print(obj) to the function _process_docstring in the file sphinx/ext/napoleon/__init__.py, the Cython extension appears to be the cause. The error message with the extra print lines then becomes:

Extension error:
Handler <function _process_docstring at 0x7f1a9f310cb0> for event 'autodoc-process-docstring' threw an exception (exception: module, class, method, function, traceback, frame, or code object was expected, got builtin_function_or_method)
<module 'iodata.overlap_accel' from '/home/toon/univ/code/iodata/iodata/overlap_accel.cpython-37m-x86_64-linux-gnu.so'>
<built-in function add_overlap>
<built-in function fac2>
make: *** [Makefile:21: html] Error 2

To Reproduce
Steps to reproduce the behavior:

git clone git@github.com:theochem/iodata.git
cd iodata
./setup.py build_ext -i
cd doc
make html

Expected behavior
A more understandable error message would be helpful. Ideally, Napoleon should just process docstrings of "builtin" functions. These are not necessarily built-in to Python. They can be functions from any extension module.

Your project
https://github.com/theochem/iodata/

Screenshots
not applicable

Environment info

• OS: Fedora 32
• Python version: 3.7.7
• Sphinx version: 3.2.0
• Sphinx extensions: sphinx.ext.napoleon
• Extra tools: none

Additional context
not applicable

[keewis]

I think this is a duplicate of #8074 and fixed in #8077. The reason is that inspect.getfile, which is used to provide a location for a warning, raises a TypeError for docstrings defined in C extensions (which I completely missed when writing #7690). The fix will be included in 3.2.1

[tovrstra]

@keewis. Thanks for explaining. It looks very similar indeed. I just tested with the 3.2.x development branch and the error disappears indeed.