Do not error if use_2to3 is set to a false value

[plumdog]

Summary of changes

I think this slight change fixes an issue where setting use_2to3=False would fail fast when really it ought to be ignored.

I think this results in behaviour closer to what I had in mind with #2769.

Pull Request Checklist

• Changes have tests - no because I can't see where one would go
• News fragment added in changelog.d/.

[jimmyhli]

should 58.0.2 be yanked? I assume this will go into a 58.0.3

[webknjaz]

In general, I think it's a move in the right direction but could you also add a regression test for this scenario? I'd like to see both positive and negative test cases added.

[webknjaz]

FTR somebody needs to evaluate this as a possible fix for #2776.

[jaraco]

should 58.0.2 be yanked? I assume this will go into a 58.0.3

What value would there be in yanking 58.0.2?

[jaraco]

In general, I think it's a move in the right direction but could you also add a regression test for this scenario? I'd like to see both positive and negative test cases added.

I'm happy to skip regression tests for these tweaks.

[jimmyhli]

should 58.0.2 be yanked? I assume this will go into a 58.0.3

What value would there be in yanking 58.0.2?

No, there wouldn't be, if a 58.0.3 is released. I think some were suggesting that 58.0.2 be yanked so the latest just points to 58.0.1 :)

Thanks for the fix!

[potiuk]

We have some packages as dependencies, which have 2_to_3 set to true (specifically FlaskOpenId which has not been updated since 2016, so this is likely it won't be updated. https://github.com/mitsuhiko/flask-openid/ Any chance setuptools can simply ignore 2_to_3 setting, rather than ignore when false ?

UPDATE: see below - I withdraw the proposal. It was bad idea

[plumdog]

@potiuk See #2775 (comment) and #2769 for rationale behind not ignoring the flag - the installed package almost certainly won't be importable.

[potiuk]

Yeah. Absolutely. Actually we had an interesting case because the Flask-openid actually was working with just removing 2_to_3 - because it was really small, the changes were harmless (like removing u` from unicod strings) and the only non-compatible python 3 code was under "if python < 3" anyway. Thankfully we got response from the maintainers and we became maintainers ourselves and we released a new version of package (after manually converting it with 2_to_3 and making it python3 - only). We also released a .whl package which remove the need of using setuptools at all.

I agree it's not possible and tha approach you come up with with "false" is good.