Make setuptools-rust play nice with cffi

[alex]

The primary benefit of this is that it'll work when you have a single project using both setuptools-rust as well as cffi.

[alex]

(Apologies that this is a WIP mess! I'm having some trouble testing locally, so I'm relying on travis!)

[alex]

👋

So the reason for this is that currently, if you use both cffi and setuptools-rust in the same setup() invocation, you have a sad time. See pyca/cryptography#5357 as an example that hits this.

This switches from the monkeypatching approach to using the same approach that cffi uses, which should play nice with other things that use the same strategy. It's a bit gross, but seems to work!

❤️ to @dstufft for explaining this to me.

We can't use maturin for the same reason: we have a large existing setup.py and we want to do a gradual migration.

[alex]

If there's anything I can be doing to help get this reviewed and landed, please don't hesitate to ask me to do more work :-)

[davidhewitt]

As a pyo3 maintainer I'm not super familiar with the internals of setuptools-rust, though am happy to help you get this through.

It looks like all functionality is preserved with this PR, though what is not clear to me is whether this is a breaking change to public API of setuptools-rust. Can you elaborate on this?

Also, as this PR is intended to add compatibility with cffi, it would be great to add an example / test which uses both together in the same project and proves it works in CI.

[alex]

a) I don't think this changes anything, at least as used by normal setup.py. If people were relying on importing these classes for some reason, I guess it potentially breaks them?
b) Yes, good call, I'll write an example/test.

[alex]

Question: Would it be ok to add some tiny CFFI usage to the html-py-ever module, or woiuld you prefer it be an entirely separate example?

[alex]

Took me a few try to get .travis.yml happy, but we now have tests. Huzzah!

[davidhewitt]

This is brilliant; thanks very much!

I'm writing a PR this afternoon to update the release process for setuptools-rust to use github actions, once I've got that sorted I'll put out a new setuptools_rust version.

[alex]

🎉 thanks!

[davidhewitt]

(Just an apology to note that I had too many things I was making progress with this afternoon and didn't get around to the release process PR. It's on my list of things to make time for in the week ahead. Sorry for the delay.)

[alex]

Totally understand! If I can be helpful, let me know.

…

On Sun, Aug 2, 2020 at 6:06 PM David Hewitt ***@***.***> wrote: (Just an apology to note that I had too many things I was making progress with this afternoon and didn't get around to the release process PR. It's on my list of things to make time for in the week ahead. Sorry for the delay.) — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#68 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAAAGBCEYYS22B7CF6QAPY3R6XPNJANCNFSM4PJATVKA> .

-- All that is necessary for evil to succeed is for good people to do nothing.

[davidhewitt]

0.11.0 is now released and should be good for you to use!

[alex]

Woooo!

…

On Thu, Aug 6, 2020 at 8:24 AM David Hewitt ***@***.***> wrote: 0.11.0 is now released and should be good for you to use! — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#68 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAAAGBETTYLGNKWAAXUCLILR7KOHXANCNFSM4PJATVKA> .

-- All that is necessary for evil to succeed is for good people to do nothing.