`attrs` version

[mmiller-max]

Hi there,

Is there any reason for this upperbound?

https://github.com/allegroai/clearml-agent/blob/e3e6a1dda81bee2dd20a64d09746568e415f1823/requirements.txt#L1

Looking at the release log I don't think clearml-agent will be affected by any of the breaking changes, but I'm not an expert!

If it can be upgraded to attrs>=18.0,<21.5.0 I'll make a PR.

[jkhenning]

Well, there were a few breaking changes that caused us to limit the version (IIRC it was related to the cmp argument). I have no objection to such a PR, but it will require in-depth testing to make sure nothing was broken, and I would appreciate a summary of all (potentially) breaking changes in attrs versions (>20.3.0) for features used in our code 🙂

[Honzys]

Well, there were a few breaking changes that caused us to limit the version (IIRC it was related to the cmp argument). I have no objection to such a PR, but it will require in-depth testing to make sure nothing was broken, and I would appreciate a summary of all (potentially) breaking changes in attrs versions (>20.3.0) for features used in our code slightly_smiling_face

Hello @jkhenning , Can I kindly ask you what is the status on this issue?

We would also really appreciate freeing up the upper version limit for attrs.

AFAIK it should not break any other functionality except the cmp property, which could be just fixed using the .eq and .order. You can see this here - https://github.com/python-attrs/attrs/pull/939/files

List of breaking changes can be seen here - https://www.attrs.org/en/stable/changelog.html#id36 Version 21.1.0 have been yanked from the PyPI, so there's only 4 new versions that we need to thoroughly check.

And if I am not mistaken, there's no other changes to the attr module, they've only made changes to the attrs module.

So what do you think? Could we create a PR to release the upper limit?

[jkhenning]

Hi @Honzys,

Sure, we would welcome a PR 🙂 👍