I apologise for the messy commit log. You may want to use squash merge 😅.

No worries :) Feel free to force-push to fix it! After a first code review it can be good to avoid force push though, to make incremental review easier.

This seems like a good idea, but we need to make an exception for auto-merged dependabot PRs. Example: #2761. Either based on branch (starts with dependabot/...) or based on that the author is dependabot. We handle this by drafting release notes right before a release: https://github.com/sharkdp/bat/blob/master/doc/release-checklist.md#changelogmd-updates. But it's nice if the CHANGELOG is prepared as much as possible anyway.

Rebased on master.

git rebase -i master and squashed some messy commits.

Ah shit. I think I messed up. branches-ignore: refers to the target branch, not the source branch.

This is hard to verify before it's merged I suppose? So I'm tempted to merge it after you fix the comment so we can verify and iterate on it.

Oh wait, right, it runs already (after you fix the bug)

This is hard to verify before it's merged I suppose? So I'm tempted to merge it after you fix the comment so we can verify and iterate on it.

I made sure this workflow errors properly before I added the changelog for this PR. So if it passes now (as it should), I think it should be mostly fine.

@cyqsimon Hi! Any chance you could fix a bug in the script? Right now it uses the last commit in a PR to determine author. That will give the wrong author if a maintainer adds a commit to the PR. We likely need to use gh pr view 2779 --json author | jq -r .author.login or something similar. Thanks in advance!

Example of where the check fail when it shouldn't: #2779

Arrrrgh. There's always an edge case that slips by. I'll take a look.