| 27 Mar 2026 |
 hexa (signing key rotation when) | @emilazy:matrix.org @k900:0upti.me @vcunat:matrix.org let's make a call on this | 10:44:33 |
 vcunat | It certainly looks good at a glance and in principle. | 10:47:34 |
| vcunat | But I haven't thought about details, e.g. whether it's OK to have this nesting (aggregate of aggregates). | 10:48:40 |
| vcunat | Overall the situation around what's channel blockers and what isn't - I find it messy, not very maintained. | 10:49:44 |
| vcunat | I suppose we can merge optimistically. | 10:57:00 |
 K900 | I think it's fine unless there's a reason Hydra will explode | 11:02:37 |
 emily | oh I had a suggestion around this in general | 11:02:38 |
| K900 | But it would be nice to maybe one day kill off the branch separation | 11:02:52 |
| emily | from before this PR actually | 11:03:04 |
| emily | can't we just merge the jobsets. like keep the channel advancement conditions decoupled but kill nixpkgs:unstable | 11:03:43 |
| emily | that way you can much more easily pick commits that are both nixpkgs-unstable and nixos-unstable | 11:04:05 |
| K900 | Conceptually yes, but it would require significant channel-scriptage | 11:04:14 |
| emily | and Hydra saves queueing and building a bunch of redundant Linux jobs from nearby commits | 11:04:22 |
| K900 | And no one wants to touch that | 11:04:27 |
| emily | it would just mean looking at the same job set for both, no? the actual tested job wouldn't change | 11:04:58 |
| K900 | It would also mean looking at the completion state of the other jobs | 11:05:22 |
| K900 | Unless we want to keep the basic "everything" check | 11:05:36 |
| emily | this would also fix the nixos-YY.MM vs. nixpkgs-YY.MM-darwin discrepancy with stable that I hate | 11:05:38 |
| K900 | Which would slow Darwin down quite a bit probably | 11:05:43 |
| emily | that seems fine to me | 11:06:01 |
| emily | like you worry about nixpkgs-unstable blocking on NixOS test builds? | 11:06:28 |
| K900 | Yeah | 11:06:40 |
| emily | I think nixos-unstable updates often enough tbh (and the freq of evals could be bumped to compensate for nixpkgs:unstable jobs going away) | 11:07:04 |
| K900 | Possibly | 11:07:13 |
| K900 | I think another problem we will have is getting people off those channels | 11:07:31 |
| emily | again I don't propose any change to how channels advance | 11:08:00 |
| emily | I genuinely just mean merging the jobsets themselves | 11:08:15 |
| K900 | Hm | 11:08:52 |
| K900 | We can probably do that fairly easily yeah | 11:08:59 |
| emily | this came up because I wanted to have one Nixpkgs pin that passed both nixos-unstable and nixpkgs-unstable gates which is hard to do manually right now because they're almost never building the same commit | 11:09:03 |
