| 24 Mar 2026 |
 John Ericson | I think the large point still stands, which is that if we think the way that Helsinki left is was in a known-good state, we should get that deployed and confirm be for me and my people possibly destabilize it :) | 16:05:49 |
| John Ericson | and we would be much more confortable knowing that were not modifying something that's already massively diverged from production too | 16:06:12 |
| John Ericson | (since production would be caught up) | 16:06:33 |
| John Ericson | * I think the large point still stands, which is that if we think the way that Helsinki left is was in a known-good state, we should get that deployed and confirm before me and my people possibly destabilize it :) | 16:06:56 |
| John Ericson | Mic92: if we put the derivations in the data base, what we do do for the old build steps where the derivations are now garbage collected? | 17:18:47 |
| John Ericson | Simon Hauser: FYI we seem to be getting some S3 errors --- probably auth related --- on staging now | 22:53:03 |
| John Ericson | Mic92 and I looked a bit, and so see the stuff in the infra repo | 22:53:27 |
| John Ericson | but it is not yet working | 22:53:30 |
 Mic92 | @simon.hauser:helsinki-systems.de: we just need to dig a little in what S3 SDK environment vars are supported | 22:54:36 |
| Mic92 | https://github.com/NixOS/hydra/pull/1598/changes/2e97e1a32e1383ea46386739551a163479a0ed81 not this one apparently | 22:56:01 |
| 25 Mar 2026 |
| John Ericson | Mic92: it's...working now? | 18:37:32 |
| John Ericson | I went to go poke at it, and it seemed to have done something so that's cool | 18:40:17 |
 BMG | 
Download 743.jpg | 19:25:25 |
| Mic92 | didn't change anything | 21:49:18 |
| 26 Mar 2026 |
 Pratham Patel | Is it possible to get alerts for package breakage in Hydra before those even make it to the respective channels? | 07:34:43 |
| Pratham Patel | I was thinking of looking at the status of a jobset for the last 3 evals and see if the failure is consistent or not. But wanted to ask here if there's a better way nor not. | 07:36:02 |
 vcunat | hydra the SW can send e-mails. Or at least it could in the past. We used to have it turned on for hydra.nixos.org even, but for nixpkgs it didn't turn out practical. | 11:35:40 |
| vcunat | * hydra the SW can send e-mails. Or at least it could in the past, but I don't think it's been removed. We used to have it turned on for hydra.nixos.org even, but for nixpkgs it didn't turn out practical. | 11:35:55 |
| vcunat | Pratham Patel: maybe this thing?
https://zh.fail/failed/by-maintainer/vcunat.html | 11:36:40 |
| vcunat | (I know about no useful notification system for nixpkgs itself so far.) | 11:37:47 |
 Janne | The Main reason was that it notified way too many people | 11:37:49 |
| vcunat | * (I know about no useful notification system for nixpkgs itself so far, but I might've missed some.) | 11:38:05 |
| vcunat | Yes. But perhaps for some specific hydra use (not hydra.nixos.org) it could be usable. I don't know. | 11:38:52 |
| vcunat | Still, the OP asked about channels, so I expect it's about hydra.nixos.org only. | 11:39:34 |
| vcunat | IIRC that system notified committers, not package maintainers. | 11:40:49 |
| vcunat | Maybe it would work well-enough in practice if maintainers were notified instead. | 11:41:11 |
| Janne | I think it notified every committer of every commit in the range between two evals if a package newly failed. For every package | 11:42:03 |
| vcunat | The approach isn't compatible with staging* batching - which we require, as there's not enough build power for more frequent builds. | 11:43:43 |
| John Ericson | https://hydra.nixos.org/jobset/hydra/nix-next#tabs-errors anyone know why this eval is totaly failing? | 15:52:59 |
| John Ericson | * https://hydra.nixos.org/jobset/hydra/nix-next#tabs-errors anyone know why this eval is totally failing? | 15:53:11 |
