Bump dependencies from commit 93933b

[A5rocks]

I just realized that our pre-commit autoupdates are kinda unnecessary. Should we disable those (if we can) and then maybe make this biweekly?

[codecov[bot]]

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

Project coverage is 99.63%. Comparing base (93933b5) to head (74401a4).

Additional details and impacted files

```diff @@ Coverage Diff @@ ## master #3008 +/- ## ========================================== - Coverage 99.63% 99.63% -0.01% ========================================== Files 120 120 Lines 17865 17855 -10 Branches 3213 3212 -1 ========================================== - Hits 17800 17790 -10 Misses 46 46 Partials 19 19 ``` [see 6 files with indirect coverage changes](https://app.codecov.io/gh/python-trio/trio/pull/3008/indirect-changes?src=pr&el=tree-more&utm_medium=referral&utm_source=github&utm_content=comment&utm_campaign=pr+comments&utm_term=python-trio)

[CoolCat467]

I just realized that our pre-commit autoupdates are kinda unnecessary. Should we disable those (if we can) and then maybe make this biweekly?

Apparently can't disable pre-commit autoupdates (not seeing anything on their website), but we could change autoupdate_schedule to monthly or quarterly instead of weekly. Maybe while we are at it we could trial enabling autofix_prs?

[A5rocks]

I still don't really think autofix_prs would work well tbh. Also monthly probably sounds fine, I'm super surprised we can't disable autoupdates though. Maybe it's a setting in the CI? Nevermind, checked the site and I cannot.

(though my opinion on autofix_prs is very weakly held)

[A5rocks]

We could probably use https://pre-commit.ci/lite which might be slightly slower but a) uses github actions and b) simply cannot autoupdate. (it supposedly can still autofix though so that's a thing) What do you think?

[CoolCat467]

I think they want things to be auto-updated because if you think about it from their perspective, if you have tons of projects using similar tools, suggesting for everyone to update to new versions means they can purge old copies and save disk space. And really most of the main reasons for using pre-commit.ci in the first place is keeping everything up to date and the PR autofixing.

I don't see much of a point to using the lite version, as what it seems to be intended to replicate is the autofix part. Auto updating is as simple as running pre-commit autoupdate, which is a part of autodeps already.

I personally like pre-commit's auto update PRs because it means all the hooks stay more up to date. If we had to get rid of it for some reason though, autodeps running more often would accomplish the same thing.

[A5rocks]

I personally like pre-commit's auto update PRs because it means all the hooks stay more up to date. If we had to get rid of it for some reason though, autodeps running more often would accomplish the same thing.

oh yeah, that's what I meant. I'd personally like to consolidate autodeps and the pre-commit.ci prs (by removing the pre-commit.ci prs and making autodeps biweekly, or maybe weekly? probably biweekly).

[CoolCat467]

There are pros and cons to merging the two in my eyes.

Pros:

• Not nearly as easily possible for there to be differences between tools' hook versions and versions used in CI testing

Cons:

• pre-commit.ci is free for public repos and isn't eating into our Github Actions minutes unlike merging them would be
• Can't do PR autofixing anymore if we disable pre-commit.ci integration (only way to truly disable autoupdates apparently)

[A5rocks]

yep the PR autofixing was why I was pointing out the lite version. the way I see it, the versions being different is annoying and neither of the cons is something we care about as much.