| 26 Nov 2025 |
 leona | and it can be even longer when the staging-next cycle is difficult (and that is likely because it will have multiple complicated things) | 16:59:58 |
| leona | * and it can be even longer when the staging-next cycle is difficult (and that is likely because it will (probably) have multiple complicated things) | 17:00:09 |
 Marcus | I guess I'm confused about how this works. I assumed that when 25.11 was forked off from master, master would start getting the changes from staging that won't be present in the release again. | 17:03:53 |
 K900 | Yes, but staging is done on one branch at a time | 17:04:30 |
| K900 | So right now it's building staging-25.05 | 17:04:34 |
| K900 | Next will be staging-25.11 | 17:04:37 |
| K900 | And then staging-next (for master) | 17:04:41 |
| leona | no, staging is for large (and sometimes even world rebuilding) packages, it's just that staging is earlier unrestricted for breaking changes | 17:05:02 |
| leona | at some point there will be a merge staging (via staging-next) -> master, but this depends on hydra building everything | 17:05:25 |
| leona | and takes time and people fixing things | 17:05:39 |
| leona | ref https://github.com/NixOS/nixpkgs/blob/master/CONTRIBUTING.md#staging | 17:06:06 |
| Marcus | I'm aware of what it's for. 🙂 Just wasn't aware that it's being run one branch at a time. But I guess I don't have to worry about perl 5.42 reaching unstable anytime soon then. | 17:08:49 |
| leona | hydra is too underdimensioned to run multiple, and right now it takes about 1.5-2 weeks with no problems for one staging-next | 17:09:32 |
| K900 | Hydra also does not have capacity to run multiple stagings at once in its current state | 17:11:11 |
| Marcus | How does this correlate with the perl and python jobsets? They often take half a week or so, and are run separately from the staging process. I guess they then steal capacity from ongoing staging process? | 17:13:49 |
| Marcus | (and other similar ones) | 17:14:15 |
| K900 | I don't think there's a perl jobset | 17:14:42 |
| K900 | python-updates does steal some capacity | 17:14:47 |
| K900 | But it's configured to have a lower priority | 17:15:01 |
| Marcus | I'm pretty sure there is 😄 | 17:15:01 |
| Marcus | https://hydra.nixos.org/jobset/nixpkgs/perl-updates | 17:15:26 |
| K900 | That is currently disabled and configured to use 600 shares | 17:16:31 |
| leona | python-updates do currently primarily run on a personal hydra | 17:16:31 |
| K900 | Which is very few shares | 17:16:33 |
| K900 | Though Hydra scheduling is exceptionally fucked | 17:16:42 |
| K900 | And the number of shares is kind of arbitrary and irrelevant | 17:16:54 |
 whispers (it/fae) | i have been curious, how do y'all actually monitor the status of a staging cycle? is it through the eval results at e.g. https://hydra.nixos.org/eval/1820494 or…? | 17:18:45 |
| K900 | Basically that yes | 17:18:59 |
| Marcus | I imagine not a lot of folks want to touch the hydra code base to improve it, given it's catalyst 🙂 | 17:19:57 |
 vcunat | The PRs have some further useful links:
https://github.com/NixOS/nixpkgs/pull/464512 | 17:36:05 |
