Components: Update CHANGELOG CI check to support forked repos

[chad1008]

Fixes #49817

What?

Update the Components Package CHANGELOG check to support forked repos

Why?

Without this change the check will always fail because it doesn't know what upstream repo to checkout from

How?

Add a repo declaration!

Testing Instructions

1. Fork Gutenberg
2. Make a change to any file in the components package that isn't a unit test, storybook file, or mobile/native file. Don't modify the CHANGELOG
3. Commit and push your changes
4. Create a PR from the modified branch on your fork
5. Confirm that the CI check runs, and fails because the CHANGELOG has not been updated.
6. Confirm the error output does not reference an error while fetching the repository
7. Update the CHANGELOG with an entry in the Unreleased section
8. Commit and push
9. Confirm the check runs and passes

[t-hamano]

Thanks for the PR, @chad1008!

I would like to test this from the actual forked repository. Check out this PR (component-changelog-ci-forked-repos) in my forked repository, make changes, and submit the PR to trunk. Is this procedure correct?

[t-hamano]

LGTM!

I have confirmed that you have tested with #49905 and it is working as expected 👍

I have another concern. My understanding is that if there is a difference in the packages/components/CHANGELOG.md file between trunk and PR, the file is considered changed and passes the test:

gutenberg/.github/workflows/check-components-changelog.yml

Lines 32 to 35 in fb92cc8

	if git diff --quiet ${{ github.event.pull_request.base.sha }} HEAD -- "$changelog_path"; then
	echo "Please add a CHANGELOG entry to $changelog_path"
	exit 1
	fi

However, the trunk may be ahead of the PR and the changelog may be updated in the trunk.
In this case, I'm concerned that even if the PR doesn't contain the updated CHANGELOG, a difference will still occur and pass the test.

Perhaps it would be better to use an action such as changed-files and check if a particular file has been updated in the PR.

What do you think about this?

[chad1008]

Thanks for reviewing @t-hamano!

trunk may be ahead of the PR and the changelog may be updated in the trunk.
In this case, I'm concerned that even if the PR doesn't contain the updated CHANGELOG, a difference will still occur and pass the test.

Great minds think alike... I noticed the exact same thing yesterday!

The short version of that comment is that the Github context we're using for the diff command is meant to look at the current branch's fork point, but for some reason it's returning the latest commit on the base branch... leading to the false positives you've described when trunk is ahead.

In that PR, I'm also introducing some new checks to make the feedback more specific/meaningful, and those new checks will still catch cases like this, so I think we're okay. The error message would be a little off ("make sure you've linked the right PR" instead of "add an entry") but it will accomplish the same goal of giving the author a heads-up that the CHANGELOG needs attention 😄