Ping maintainers on breakage

Hydra used to ping maintainers when their packages broke. It doesn’t do that anymore, since there was an incident where it spammed every maintainer.

I would propose to start pinging maintainers once again. There seems to be some kind of consensus that maintainers should have real responsibility and I think fixing breakage is a central part of that responsibility. I have even written a script for myself that scrapes hydra and sends me emails.

For spam mitigation, it may be interesting to look into

  • notifying only the maintainer of the actual broken derivation or the first maintainer(s) up the chain if there is none. This way a stdenv failure won’t spam every maintainer.
  • notifying only once per week (or other timeframe) per package

I’m not sure if I’m up to hacking hydra myself, so this is mostly intended as a discussion without promise of implementation.


Agreed. Especially with new RFC(s) discussing automation and escalated privileges for maintainers I think this is important.

1 Like

This would make my maintainer-ship a more convenient task. Currently I have to keep track of things on a regular basis, and if I don’t keep up things could be broken for a while.

I’d also like to note that maintainers are typically more responsive to notifications of breakage than, for example, updates. Though a possible drawback (maybe past issue?) is being notified repeatedly about transient failures. That would make this lose value.


Yes please. This would be really helpful.

Being notified upon Hydra failure would be very useful. However I would prefer Hydra to open a n issue on GitHub (assigned to me) rather that receiving an email. This is would help tracking the issue and it would also allow to discuss the issue with other developers.


That’s a very good idea. It would probably be best to implement that outside of hydra, is there some kind of API available?


Hydra also has an API:


I also very much miss these notifications.

1 Like

To prevent spam of email/issues- perhaps the title could be of the format- hash-package name (Path inside /nix/store) and if it’s found that such an issue has been closed in the last 10 days then don’t create a new one.

I’d also prefer to receive mails from Hydra.

Here is a later Discourse thread on the same topic. (The feature was not implemented there either, unfortunately.)

1 Like

I would love this as a maintainer (though perhaps I would also like to receive email)

I would also love this as a non-maintainer: this would provide an obvious place to share information on the failure, so that as a ‘casual’ contributor it is easy to read up on what has been already found out about it, use that as a starting point to investigate further and share your your results (even if you didn’t get far enough to be able to produce a PR).