Marvin is an experimental bot with the goal of improving the nixpkgs PR workflow. Because its experimental, it only acts on PRs that have explicitly opted-in to it and its behavior may change at any time.
Marvin is supposed to make the life of PR authors and reviewers (with or without commit permission) easier. See USAGE.md and the bot's profile for details on the usage. This document focuses on the high-level aspects.
You can opt-in to use marvin on your PRs. You can only do that on your own PRs. Once your PR is opted-in, as indicated by the "marvin" GitHub label, marvin will
- Semi-automatically keep the status of your pull request. Does it currently need a new reviewer (label
needs_review)? Is it in the middle of the review cycle (labelawaiting_reviewerorawaiting_changes)? Has the reviewer become unresponsive? - Based on this status, it will nag the reviewer in case they forgot to respond to your questions or changes. This happens after three days of inactivity in the
awaiting_reviewerstatus. This keeps your PR from being forgotten. - If the reviewer remains unresponsive, it will put the PR back in the review queue. This way the PR will be discovered by a new reviewer eventually.
- If nobody reviews your PR, it will assign a reviewer to it. There are limited capacities for assigning reviews, so this might take some time. Reviewers are assigned to
needs_reviewerPRs on an oldest-pr-first basis. Your PR will never "drown" in the queue!
If your PR is managed by marvin and you are getting impatient, consider spending some time as a reviewer yourself. That way you can give back to the community and reduce the size of the queue. Smaller queue means more reviewers for your PR. You do not need any permissions or prerequisites for reviewing, just read on.
If you want to help out with the PR backlog, marvin makes it easy to find a PR that does not have a reviewer yet. Once you have picked a PR to review, marvin empowers you to "merge by proxy".
Do this on a honest best-effort basis. If you had the permission, would you push the merge button? If the secondary reviewer has additional feedback, that is perfectly fine. Consider it a learning experience for the next review. You do not need to have review experience to start reviewing PRs.
If you have the permission to merge PRs yourself, you can still benefit from the discoverability features from marvin. You can focus on PRs that have already been positively reviewed or simply pick PRs from the queue of PRs that need a reviewer.
You can help out.
Whoever you are, if you have some minutes to spare you can review a PR. See the sections above for more information.
Marvin can also assign you to PRs to review. This is a nice way to make a sustainable and substantial contribution. Would you review a PR a week if an email invited you to do it? Sign up for it! See this issue for more information. Keep in mind that the experience is still a bit rough around the edges, especially the rate limiting is a bit unintuitive for now.
It is currently difficult to get started with contributions since marvin is hard to test. If you want to get your hands dirty anyway, check out CONTRIBUTING.md.
Marvin can be considered trial run of rfc 30. It is experimental on an opt-in basis because the behavior is not finalized and there has not been a community decision yet. It will likely remain experimental for a long time since I (@timokau) work on it in my free time (and who has enough of that?). It is not at all certain that it will ever become official. It will require some more experimentation and probably at least a second maintainer.
Once I (or whoever maintains marvin then) think marvin is ready for stabilization, it will go through an RFC process.
Further readings: