| 27 Mar 2026 |
 John Ericson | Mic92: do you have an opinion on this bytes type vs Vec<u8>? | 16:39:32 |
 Mic92 | John Ericson: you refer to this one? https://github.com/tokio-rs/bytes | 16:57:46 |
| John Ericson | yeah | 16:57:58 |
| John Ericson | I don't get the point of it vs Vec<u8> and &[u8] | 16:58:12 |
| John Ericson | I thought this might have come from nix.rs but I am not sure | 16:58:24 |
| John Ericson | (maybe I am just missing something as to why it's better to be clear) | 16:58:42 |
| Mic92 | John Ericson: I think bytes is ref-counted, so you can do cheap clone() | 16:58:49 |
| Mic92 | so useful if you parse protocols | 16:59:09 |
| John Ericson | hmm ok | 16:59:14 |
| John Ericson | interesting | 16:59:20 |
| Mic92 | yes it came from nixrs, I have not used it much before either. | 16:59:44 |
 Mindavi | In reply to @Ericson2314:matrix.org
we should invite anyone we know using hydra If I find some time it'd be cool to play with it. But don't have infra setup for OIDC/OAuth at all yet, so it might be a bit of a time sink to set it all up. But will give feedback if I get anywhere | 17:00:45 |
| John Ericson | sounds good! | 17:01:10 |
| John Ericson | https://github.com/NixOS/hydra/pull/1605/files btw if anyone wants to review that | 17:02:03 |
| John Ericson | it remoes the last C++ :) | 17:02:10 |
| John Ericson | BMG: Is there a valid reason that drvPath is nullable on BuildStep? | 20:51:11 |
| John Ericson | Also it seems like Build / BuildOutput is full of redundant stuff and we should instead track the root BuildStep? | 20:51:42 |
|  Artemis Tosini (NixOS) joined the room. | 20:53:57 |
| John Ericson | (when you see this BMG, please ping Artemis Tosini (NixOS) too) | 21:01:42 |
|  Sapii joined the room. | 22:49:00 |
 BMG | Not off the top of my head but I might have it documented somewhere. It was buildstepoutputs I was looking at | 23:11:43 |
| 28 Mar 2026 |
 Janne | Tbh, the crurent velocity is a bit extreme imo and the speed gives a lot of viebecoding vibes. We still have not deployed the new qr to staging (let alone production) and are adding more and more changes on top of master that will have to be deployed along all fixup commits for the new qr when something goes wrong in production. I am obviously very happy to see C++ removed but I also fear a very high chance breakages - see the Nix/S3 debacle that was Claude-powered | 11:37:51 |
| Janne | We actually have a custom plugin for that that was never as generic as what the PR does … I wonder if we should allow for multiple providers while we're at it. Might be able to remove the Google tracking (or google altogether) from Hydra which has been annoying me for a long time now with h.n.o | 11:38:41 |
| Janne | 
Download image.png | 11:39:36 |
| Janne | These requests in particular | 11:39:39 |
 tomberek | I was looking at doing some Hydra hacking, but not sure what is currently expected to pass tests and what would be new breakages. My guess is that we're in a transition period? | 21:36:37 |
| tomberek | (was looking at: https://github.com/NixOS/hydra/pull/1518) | 21:36:41 |
| 29 Mar 2026 |
| John Ericson | To be clear, that PR should definitely go to staging first, and should not hold up the deployment to prod | 15:59:22 |
| John Ericson | the other changes I have read at code review time quite closely, and I would not be so concerned | 16:00:10 |
| John Ericson | I definitely agree we don't want a buch more dev to pile up and then we end up with something that cannot be deployed | 16:01:01 |
