Backport workflow: Add some feedback comments

[The-Compiler]

It would be good to get some (automated) feedback for the backport workflow. The bot could e.g. add a comment with a link to the backported PR if successful (and also take care of adding the "backported" label maybe?). Perhaps more importantly, if not successful, it should add a comment asking for a manual backport.

Right now, it's kind of "fire and forget", and if we don't carefully double-check, we risk accidentally forgetting about backports which should have been done.

[RonnyPfannschmidt]

the back-port pr should auto-link by referring the origin pr
on failure we should create a issue about bug fixing the backport and/or fixing the backporter

maybe we can use a label as indicator like "backport-failed", and taking the label off and/or switching it to backport needed again could re-trigger a backport

but that may bee a but pushing it on the mechanics

[The-Compiler]

maybe we can use a label as indicator like "backport-failed", and taking the label off and/or switching it to backport needed again could re-trigger a backport

I think transient backporting failures will be very rare. The most common reason a backport will fail is that there are conflicts, and at that point, a human will need to intervene. Still, deleting the branch and applying the label again will already trigger a new backport PR anyways.

[webknjaz]

FYI I have a GitHub App for backporting @ https://github.com/apps/patchback.

Here's the comments I came up with, for example:

• Introduce AppKey for type safety aio-libs/aiohttp#6498 (comment)
• Make pip use --user in GHA CI/CD workflows aio-libs/aiohttp#6588 (comment)

(config: https://github.com/ansible-collections/community.general/blob/main/.github/patchback.yml)

That's in case you'd want to borrow ideas or just use it — it's been online for quite a while now: https://github.com/ansible-collections/community.general/pulls?q=author%3Aapp%2Fpatchback.

P.S. I like the labeling idea. My bot doesn't do it but I should probably implement that :) I think it's a good way for providing a discovery mechanism for PRs with failed backports.