| 8 Apr 2026 |
 John Ericson | (maybe this was better for #infra:nixos.org ) | 17:53:40 |
| John Ericson | BMG: it might be good to collaborate on the feasiblity of some migrations? | 18:25:48 |
| John Ericson | for example "do the root build and root build step, once there exists a ltter which is done, always agree on output info"? | 18:26:19 |
 BMG | I should be able to set aside some time this month | 18:50:12 |
| BMG | Later in the month | 18:50:29 |
| John Ericson | Sounds good! | 18:50:44 |
 hexa | do you mean the version mismatch is resolved? | 23:03:24 |
| 9 Apr 2026 |
 Sapii | Hydra is taking its sweet time testing the nix patch for nixos-unstable ðŸ˜ðŸ˜ | 00:49:42 |
| Sapii | * Hydra is taking its sweet time testing the nix patch for nixos-unstable ðŸ˜ðŸ˜ ( been the entire day ) | 00:49:56 |
| Sapii | A more technical question, is it possible for hydra to pause jobsets and differentiate between commits as to pause a jobset and prioritize building a certain commit? If not, would it be beneficial to add this for future "CVE events"? I am interested in contribuiting to something like this ( as well as learning about hydra internals ) | 01:20:42 |
 Sergei Zimmerman (xokdvium) | In reply to @opandddd:matrix.org
A more technical question, is it possible for hydra to pause jobsets and differentiate between commits as to pause a jobset and prioritize building a certain commit? If not, would it be beneficial to add this for future "CVE events"? I am interested in contribuiting to something like this ( as well as learning about hydra internals ) Hydra already has a way to bump jobs to the front of the queue. When building nix releases (esp security) we do that | 01:23:51 |
| Sapii | Nice, and I am assuming this bump was used for this patch too? | 01:32:17 |
| Sergei Zimmerman (xokdvium) | Nix releases yeah, I don’t think it does much for the nixos package sets since there it’s just a lot of stuff to build | 01:34:01 |
| Sapii | Do all the NixOS specific tests required for nixos-unstable just actaully take that much time? assuming that the cppnix patch is built but its going thru tests right now. | 01:45:48 |
| John Ericson | maybe? I mean there is something to deploy to darwin builders too for staging, so we aren't just testing linux builders | 03:57:36 |
 vcunat | Yes, rebuilding all NixOS tests on Hydra is about 1 full day of time per branch. | 05:29:48 |
| vcunat | I'm not 100% sure why it's so slow. Perhaps the fact that there are many derivations for etc stuff, disk images, etc. | 05:30:23 |
| vcunat | (builders do not get lot of work, it's the central piece that bottle-necks this) | 05:30:54 |
| vcunat | nixos-25.11 finished a couple hours ago. Now it's evaluating nixos-unstable, and I'd hope we can have a bump by tomorrow. | 05:34:47 |
| vcunat | * nixos-25.11 finished a couple hours ago. Now it's evaluating nixos-unstable, and I'd hope we can have the channel update by tomorrow. | 05:35:05 |
| John Ericson | hopefully the new queue runner helps with this! | 20:07:55 |
| vcunat | It's possible. | 20:27:18 |
| vcunat | * It's possible. Or at least I don't know anything disproving that. | 20:27:51 |
| John Ericson | I was told this sort of central bottleneck was the sort of thing it was focused on | 20:28:01 |
| Sapii | Ok thank you! i see 14 out of 201 still pending for unstable so its getting prrety close! | 23:06:17 |
| 10 Apr 2026 |
| vcunat | I don't get what numbers you mean. Either way, the channel should update within an hour now. | 04:46:15 |
| Sapii | Ahh they are the tests https://hydra.nixos.org/job/nixos/unstable/tested#tabs-constituents, sorry for not specifying | 15:19:32 |
| vcunat | They're not relevant for this. Channel bump waits for all tests, not just for this tiny subset. | 15:20:27 |
| vcunat | (we have thousands of test jobs) | 15:20:46 |
| vcunat | * (we have thousands of test jobs; the current nixos-unstable rebuild was about 20k jobs) | 15:21:11 |
