[Nix-dev] Nixbot - a little helper for pull requests

[Michael Raskin]

>You can see it doing its thing here:
>
>    https://github.com/nixbot/nixpkgs-sandbox/pull/9
>
>in this example, the package being modified (openttd, a good game in case
>anyone was wondering) does not list joachifm as a maintainer. This means his
>polite request to merge falls upon deaf ears. fpletz is a maintainer, however,
>which allows him to request a merge. If you'd also like to play with a live
>version of the bot, open pull requests against
>https://github.com/nixbot/nixpkgs-sandbox and comment on them. (This repository
>was created for testing and will likely be removed before long)
>
>While this is a simple implementation currently, more complex things such as
>hydra or travis integration and requiring more than one vote to merge are
>possible in the future if this is a direction people want to take.
>
>So, thoughts and comments on this approach?

A nice tool, but in the short run it will not help on its own.

I think most people who would get nixbot access will currently get full
commit access by promising not to touch things when they are not sure 
they understand the code. And if people do not know that, once in a few
months I run the vanity counter to find high-commit-count people without
commit access and inform them that they have a good chance to get commit
access for pushing updates and similar low-controversy changes.

As for Travis integration, we have so many false Travis failures that 
sometimes people do not notice true Travis failures.

As for Hydra integration… I think having Hydra build PRs would be nice,
but expensive… Is this going to happen?

All this doesn't diminish the niceness of the hack, but getting utility
from this piece of tech will require some effort in other areas, too.

(Yes, one of the functions of this email is to remind people who would 
like commit access to their dozen of small packages that they probably 
can get this access if they promise to be careful with it)