[BtbN]

Description

This is the other part to the previous PR #231.
If the backport fails, but then something triggers a re-run of the backport action, like additional lables being added or removed, it could get incredibly spammy, since naturally the backport will fail every time.

This adds a small commented out tag to each failure comment, and then scans the existing comments for that tag before even trying to do the same backport again.
For a retry the comment has to be deleted or the tag edited out of it.

How Has This Been Tested?

This has also been actively in use on the FFmpeg Forgejo instance, and done a couple dozen of successfull backport PRs.

Checklist

• Tests added if applicable.
• Documentation updated if applicable.

[github-actions]

Coverage report

St.❔	Category	Percentage	Covered / Total
🟢	Statements	88.05% (-1.67% 🔻)	641/728
🟢	Branches	83.02% (-3.28% 🔻)	308/371
🟢	Functions	85.31% (-2.57% 🔻)	151/177
🟢	Lines	88.16% (-1.35% 🔻)	618/701

Show files with reduced coverage 🔻

St.❔	File	Statements	Branches	Functions	Lines
🟢	runner/runner.ts	88.82% (+0.41% 🔼)	92.45% (+0.62% 🔼)	82.35% (-0.98% 🔻)	88.74% (+0.34% 🔼)
🟢	... / github-client.ts	83.49% (-6.76% 🔻)	72.58% (-15.51% 🔻)	68% (-5.68% 🔻)	84.62% (-5.26% 🔻)
🟢	... / gitlab-client.ts	80.2% (-5.07% 🔻)	65.52% (-10.48% 🔻)	81.48% (-6.52% 🔻)	80.21% (-4.41% 🔻)

Test suite run success

230 tests passing in 19 suites.

Report generated by 🧪jest coverage report action from 4826977

[lampajr]

Hey again @BtbN 👋🏻

Overall looks good and, even though the last comment might not be the one posted by the failure, I think that's a good tradeoff.

Could you please add some small section about this behavior in the README? This might be subtle if not properly documented.

I merged the other PR, do you mind rebasing and resolving conflicts?

[BtbN]

It's not the last comment, but the last "page", as per the pagination of the API, of comments.
So if it does end up retrying, it won't be right away where it'd be super spammy.
Though thinking about it, I think I overlooked that the last page there might only be a singular comment, given the page-rollover happens every N, usually 100, comments, hm.
Maybe looking at the last two pages would be a bit better?

Could also add a line like "Delete this comment to allow retries.".
And of course document it in the README.

Rebased as-is for now.

[lampajr]

Oh I missed it is the last page, I believe that's a fair tradeoff for now - I don't think there will be many cases where this won't work. Either way we can improve it later.

Could also add a line like "Delete this comment to allow retries.".
And of course document it in the README.

Yeah these would be great, or something like "Subsequent backports are blocked. Delete this comment to unblock retries"

[BtbN]

I've extended it a bit so it's the last two pages. Since on GitHub, the last page can be just one singular comment.
Readme and message are updated accordingly.

[lampajr]

LGTM

[lampajr]

Thanks a lot @BtbN 🙏