Fix missing ENOTSUP on 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

1. Only add errno.ENOTSUP to FTP failure codes if that's present.
2. Removed a few unused imports.

Closes #338

[coveralls]

Coverage increased (+0.0002%) to 99.001% when pulling 554ce16 on dargueta:pypy-enotsup into 84719be on PyFilesystem:master.

[coveralls]

Coverage remained the same at 99.001% when pulling a7c3a96 on dargueta:pypy-enotsup into 84719be on PyFilesystem:master.

[lurch]

I wonder if it might be better to fix this with try ... catch rather than adding arbitrary error numbers? 🤷‍♂️

[lurch]

P.S. Is this at all relevant? giampaolo/pysendfile#28

[dargueta]

I wonder if it might be better to fix this with try ... catch rather than adding arbitrary error numbers?

And what would we do on the catch?

Never mind, I see. We could omit that error code from _sendfile_error_codes if it's not present. Probably a better idea.

P.S. Is this at all relevant? giampaolo/pysendfile#28

I don't think so?

[dargueta]

Crashes in PyPy appear to be unrelated, due to an SFTP connection issue. Stack trace here.

[willmcgugan]

The pypy build is failing on the ftpfs tests. I've seen this before and never figured it out. It's unrelated to your changes.

I'm happy to merge this and treat the pypy test fails as another task. What do you think?

[dargueta]

I'm happy to merge this and treat the pypy test fails as another task.

Conceptually it makes sense, but anyone creating a new PR is probably gonna be really mad that their PR is broken because of something I did. Up to you.

[willmcgugan]

How about remove the pypy builds from travis for now, and I'll add an issue to add them later?

[willmcgugan]

Added an issue. #342

If you wouldn't mind removing the pypy builds I'll merge this...

[dargueta]

This is gonna have a merge conflict with the pytest PR #337 which modifies the Travis file and will remove the PyPy tests. Let's get that in first and then this will be good to go.

[willmcgugan]

@dargueta Would you mind fixing that merge conflict and we should be good to go.

[lurch]

Does this change have any impact on non-pypy usages of pyfilesystem?

[willmcgugan]

I think this was purely because the lock is redundant, as there is a lock on L714.

[dargueta]

I think this was purely because the lock is redundant, as there is a lock on L714.

Yep. I thought this was the cause of a "cannot release lock not held" (or whatever that message was) that only shows up on PyPy. Apparently this isn't, because that warning still shows up.

[lurch]

Will this PR also close #342 ?

[dargueta]

Ah, good point. Yes.

[dargueta]

Okay never mind, I give up. Let's not do 342 in this PR.

[dargueta]

@willmcgugan I think we're good now.

[willmcgugan]

Thanks @dargueta !