Bugfix: fixes a CLI argument bug for the conda-debug command

[travishathaway]

Addresses the following issue:

• conda debug fails when called on a path without trailing slash on Windows #4404

Although at first glance this issue seem pretty straight forward, it was not. The bug unearthed a dormant code path that was revealed by some interesting behavior in Python.

After some debugging, it became obvious that this line wasn't doing what it intended to do:

conda-build/conda_build/cli/main_debug.py

Line 63 in 4d1d522

if not any(os.path.splitext(_args.recipe_or_package_file_path)[1] in ext for ext in CONDA_PACKAGE_EXTENSIONS):

If I'm correct, the intent of the above line was to filter out everything out that was not a "conda package file", so in theory, directories would be diverted to this code path. In reality, it also allows strings like, /tmp/some-dir/ to evaluate as True, which means directories have always been sent to the next else block (currently the working and expected behavior).

Why is that so? It all has to do with the fact that '' in 'somestring evaluates to True. Check out the example for details:

>>> exts = {'.conda', '.tar.bz2'}
>>> path1 = '/tmp/some-dir-not-trailing-slash'
>>> os.path.splitext(path1)[1] in exts
False
>>> path1 = '/tmp/some-dir-with-slash/'
>>> os.path.splitext(path2)[1] in exts
True

With all that being said, I think it is fair to assume that the code executed under this if statement was never actually working in the first place because when it is run, it generates the error that we see. Therefore, I made the decision to entirely remove the execution of this code.

I'm not entirely sure why that was ever there in the first place, but it's definitely been broken for a bit. Additionally, the conda_build.cli.main_debug module had zero test coverage when I began, so it's very easy to imagine that this broke a while back, and we were unaware of it.

If you know more about this command, please comment, so we can find the best way to fix it!

[kenodegard]

I'm not thrilled about adding a new submodule but can see the value in standardizing validators. Perhaps we should identify a few other places where using validators will help simplify the code before we move ahead with this change?

[kenodegard]

Looking good, just noticed that there's more we can remove since you removed the unused if clause.

[kenodegard]

@travishathaway you may also want to rebase/squash 15f9605 to change the author/committer

[kenodegard]

Everything looks good, just want to make sure we aren't changing any user-facing messages that don't need to be changed.

[travishathaway]

Putting this on hold until I can resolve a comment from another one of our projects:
conda/conda#11463 (comment)

(I just need to implement fsspec instead of pyfakefs)

[travishathaway]

Putting this on hold until I can resolve a comment from another one of our projects: conda/conda#11463 (comment)

(I just need to implement fsspec instead of pyfakefs)

The pytest fixture tmpdir actually solved this problem for me.

[jezdez]

Nice!