I migrated some code to a Pants project, and it has some weird structure. The issue is that running lint shows some messages that don't have source references, and I don't have direct references in the code to those modules. Does anyone know? These are reported by pylint, but I suppose that somehow pylint gets to run on something that is not in the source code?

*************

<?>:16:4: W0622: Redefining built-in 'exit' (redefined-builtin)

*************

<?>:4:4: W0611: Unused import _pytest.mark (unused-import)

*************

<?>:5:4: W0611: Unused import _pytest.recwarn (unused-import)

Huh....I would suggest trying to narrow down the problem. For example, if you're running

./pants lint ::

, try

./pants lint dir1::

(tip to use

./pants lint --only=pylint

to skip other linters/formatters)

Great suggestion, thanks. I found the issue, some idiot named a local module pytest.py..... Unfortunately, I think that name is needed, but if I'm lucky it might work with a separate lockfile that doesn't include pytest.

oh hm, where is the file located relative to your "source roots"? Because normally that file would be something like

my_project.utils.pytest

, rather than the top-level symbol module

pytest

Like I said, the project structure is a mess. There are some folders that are used for plugins, and these are special. But yes, I should look at fiddling with the source roots too.

Cool, I suspect your best options are either a) rename the file to something like

my_pytest.py

or b) nest it under a directory like

utils

so that it's

utils.pytest

rather than

pytest

(source roots)

I think the name is used with reflection, to locate it among all plugin modules. It's a bad system, and I start to doubt I can mangle it to work with Pants. Not with a big refactoring. The folder containing pytest.py is not even listed as a source root.

I’m surprised/impressed this works even outside of Pants TBH! Shadowing names like that seems like a recipe for disaster… 🙂