Ensure poetry-plugin-export gets patched or stop using poetry

[bmw]

We have poetry pinned back at

certbot/tools/pinning/current/pyproject.toml

Lines 68 to 71 in 9d8eb6c

	# If poetry>=1.3.0 is selected, multiple versions of virtualenv are selected
	# causing problems when exporting dependencies so let's pin poetry back until
	# https://github.com/python-poetry/poetry-plugin-export/issues/168 is resolved.
	poetry = "<1.3.0"

This is to work around python-poetry/poetry-plugin-export#183.

I think we either need to ensure this gets fixed or stop using poetry (and find another way to pin dependencies across various systems as described in https://github.com/certbot/certbot/blob/9d8eb6ccfd513cf2bfd0c0920c9eb006ad536f4f/tools/pinning/DESIGN.md). If we don't do either, our dependency pinning system will almost certainly break eventually.

[ohemorange]

So, while this is a general issue, looks like we only fail on virtualenv. ...do we actually still need virtualenv for any reason? Given that we don't support python 2 anymore?

[bmw]

Grepping our codebase, it doesn't look like we depend on virtualenv and it's being pulled in as a transitive dependency. After playing with pipdeptree, this seems to be from poetry and tox. At least for those two packages, the need for two versions of virtualenv seems to have been resolved when comparing the latest version of both.

I tried just deleting the line pinning back Poetry and rerunning our pinning script, but Poetry ran for over 10 minutes before I finally killed it prompting me to open #9881 🙁 I didn't debug the problem further than that but I suspect we may need to resolve #9863 first or at the same time as this issue.

If the specific virtualenv problem has indeed been resolved, we could and probably should largely ignore the general problem, but I think it might be good for us to document it in https://github.com/certbot/certbot/blob/7ca1b8f2867b252f8657b49a3fb4c374670a629d/tools/pinning/DESIGN.md and/or in error output from the pinning scripts. I feel bad for the theoretical Certbot dev (who is almost certainly not me) who would otherwise have to independently rediscover the cause of the problem here, but maybe I'm paranoid and that's overkill.