Disallow commits to main via pre-commit-hooks

[headtr1ck]

• Towards Protect main from mistaken pushes #8209

[max-sixty]

LGTM. I have a tiny bit of pause that someone just starting out in git is going to get really confused, but that's quite minor...

[max-sixty]

(Also to confirm — this won't fail when run on main outside of a commit, in particular on CI?)

[keewis]

I guess while this can be confusing, it is also partly a blessing: especially people just starting out will open a pull request from main and then run into all sorts of issues. Though in my experience people starting out don't tend to use pre-commit, so this might not be an issue.

[headtr1ck]

(Also to confirm — this won't fail when run on main outside of a commit, in particular on CI?)

I think it will fail... The hook has no way of knowing why it was invoked.
Although, we could change it into a pre-push stage?
I just find it better to fail before committing, so you don't have to clean up your local main history later...

[keewis]

does pre-commit.ci even run on main? Edit: yes, it does. Though as a push event instead of a pr event.

This might still be useful, but I think for now this might be resolved by splitting up the branch protection rule.

[headtr1ck]

does pre-commit.ci even run on main? Edit: yes, it does. Though as a push even instead of a pr event.

Now it does not anymore ;)

[max-sixty]

does pre-commit.ci even run on main? Edit: yes, it does. Though as a push even instead of a pr event.

Now it does not anymore ;)

Presumably this precludes us from using this?

Same if someone run pre-commit run -a on an up-to-date main locally — it's important that passes...

[headtr1ck]

Presumably this precludes us from using this?

why? Now it only runs locally.

Same if someone run pre-commit run -a on an up-to-date main locally — it's important that passes...

Not sure why this has to pass, or why someone would even run that.

But hey, no strong feelings about this!

[max-sixty]

[deleted — previous version was incorrect]

Not sure why this has to pass

I often run some form of pytest && pre-commit run -a, and I would want that to pass on the current best version. I don't use git bisect much but similarly I think the principle of "tests always pass on main" is good.

But very happy to defer to whatever others think. It's also easy to change — I would encourage more change and fewer vetos!

[max-sixty]

Unless there are supporters, I would vote to close :'(