Contributing Scrutiny to Nixpkgs | Hacker News Reader

Contributing Scrutiny to Nixpkgs

jnsgr.uk

68 pointsjnsgruk16 comments

nixpkgs is a funny situation where it is incredibly easy to just run with patches and additions locally. It is often easier to make a service config just for you than something reusable and broadly useful.

Thankfully the upstream contribution process is also pretty easy. So it seems like a lot does make it upstream.

It's easy to open a PR, but not so easy to get someone to actually review it.

There's currently 165 open PRs by first-time contributors adding a new package, some of which have been just sitting there without review comments for years. https://github.com/NixOS/nixpkgs/pulls?q=is%3Apr+label%3A%22...

At least they're meticulously labeled so it's easy to find them.

Yeah if your PR isn't getting the attention it needs, you probably just need to go to the discourse forum and post a request in the appropriate thread. There's one for PRs that need a review and one for PRs that have been reviewed but are pending a final check and merge by maintainers.

https://discourse.nixos.org/t/prs-ready-for-review/3032

https://discourse.nixos.org/t/prs-already-reviewed/2617

IMHO there could be clearer guidance regarding this but as it is it's not terribly difficult to find and it works well enough.

I suspect the reason that posting on Discourse works is because this advice is not universally followed, so the workload is reduced relative to reviewing all open PRs. In other words, it's a way to distribute limited grease to only the squeakiest wheels, not to grease everything better.

But maybe I'm wrong and simply posting one of the 165 PRs per day to Discourse would be enough to clear that particular backlog in less than a year.

There's also https://matrix.to/#/#review-requests:nixos.org

While trying to get reviews, I posted on a forum a few times (I think it was this one). Got a little traction, got great advice and updated the patches. Sadly I hit the forum’s new user posting limits and was unable to get the reviews to give the final approval. So I’ve had some patches accepted and others that have been sitting there for months/years.

It has meant that after a burst of packaging, I’ve stopped submitting. I’d love to be able to get reviews even if it takes months. Any suggestions are welcome.

Try to come on Matrix; you'll be able to get much better traction there.

Some community members are also working on a merge-bot[1] with the goal of allowing package maintainers to self-service their PRs. This will allow nixpkgs to scale much better. For now, it's limited to package upgrades submitted by the ryantm-updater bot.

[1]: https://github.com/NixOS/nixpkgs-merge-bot

Thanks! I’ll give Matrix a go. It is fun packaging software up and getting it out there.

That is not too bad, considering there are about 90000 packages in nixpks now.

I’ve ended up with a small but excellent network of folks I reach out to on Mastodon who usually help me out with reviews.

Most of the maintainers I’ve interacted with are open to a ping so long as you’re not hounding them.

I’ve also found posting the PR on Mastodon with the #NixOS tag has been helpful for attracting attention.

In a perfect world some of the development could happen in nix-community repos which are usually a bit more topic specific. The packages and modules would "move up" to nixpkgs on tag releases. This is how I initially thought the repos work anyway, but now I see that most of them just live as overlays and never quite sync up with the main repo.

On this topic - I think nixpkgs could really use some tooling to delegate merge permissions (mediated by a bot, probably) to more folks for less critical areas.

Nixpkgs needs more maintainers with commit permissions, but for a project as important as nixpkgs its hard to vet enough people.

This is exceptional content. I'll be linking this frequently when I help users on Discourse and encourage them to become contributors.