Start testing PyPy

[dargueta]

Type of changes

• Bug fix
• New feature
• Documentation / docstrings
• Tests
• Other

Checklist

• I've run the latest black with default args on new code.
• I've updated CHANGELOG.md and CONTRIBUTORS.md where appropriate.
• I've added tests for new code.
• I accept that @willmcgugan may be pedantic in the code review.

Description

There's a problem with the SFTP server daemon in our tests that causes PyPy to fail on them (see #342). We can still test PyPy compatibility with other parts of the codebase without failing the build by using Travis' allow_failures directive. This will still run tests on PyPy, but won't fail the build if the PyPy builds don't pass. We can remove the allow_failures part once #342 is fixed.

[dargueta]

@willmcgugan do you know anything about the MyPy failures?

[willmcgugan]

No, they appeared recently. It's on my list to take a look at.

I'm traveling atm. Back next week.

[althonos]

seems like the recent version of mypy wants X.__new__ to return a instance of X, but that's incompatible with the factory pattern we are using in ZipFS and TarFS.

[dargueta]

seems like the recent version of mypy

Should we pin it to an earlier version for now?

[althonos]

Dissecting the version show that the new build fails with mypy == 0.740 but succeeds with mypy == 0.720. Incriminated PR is likely python/mypy#7656.

[willmcgugan]

I think I'd prefer to ignore those errors rather than pin to older Mypy.

The factory pattern is a little unorthodox, but I wouldn't say it's a bug.

[althonos]

@willmcgugan : same, this can be ignored IMO.

[lurch]

I know nothing about MyPy, but is it worth reporting a bug / feature-request asking them to support the "unorthodox factory pattern" being used in PyFilesystem? Or is a "local ignore" the only sensible way around this?

[willmcgugan]

I reckon 99% of the time that would be a bug. So Mypy is right to flag it, but we have a legit reason to ignore those errors.

I'll review that code though. Maybe there is a way of making it less unorthodox that Mypy approves of.

[dargueta]

I think I'd prefer to ignore those errors rather than pin to older Mypy.

I forgot we can do that 😄

[willmcgugan]

Thanks @dargueta