| 8 Jul 2022 |
 matthewcroughan - nix.how | Alyssa Ross: based on what you were saying about nixos-hardware, does this mean we should be building and caching nixos-hardware? | 17:10:24 |
| matthewcroughan - nix.how | that's the only way I can fathom getting rid of vendor kernels from nixpkgs | 17:10:35 |
 Alyssa Ross | Yeah I think that's a great idea | 17:10:47 |
| matthewcroughan - nix.how | cache.nixos.org would then be able to build and redistribute all vendor kernels from nixos-hardware | 17:11:02 |
| matthewcroughan - nix.how | I think what you're saying about vendor kernels might be premature though | 17:11:24 |
| Alyssa Ross | There are a lot of improvements somebody who wanted to put the effort in could make to nixos-hardware, and that's definitely a big one | 17:11:24 |
| matthewcroughan - nix.how | the reason is because flakes aren't stable yet, and this functionality all relies upon flakes | 17:11:33 |
| Alyssa Ross | no it doesn't | 17:11:38 |
| Alyssa Ross | I've been using nixos-hardware since before flakes existed | 17:11:51 |
| matthewcroughan - nix.how | in practice, nixos-hardware is a flake without legacyPackages | 17:11:51 |
| matthewcroughan - nix.how | it's definitely dependent on flakes as functionality | 17:11:57 |
| matthewcroughan - nix.how | I'm not saying this cannot be expressed without flakes, nor should it, but something to keep in mind | 17:12:08 |
| Alyssa Ross | I don't understand at all the point you're trying to make | 17:12:24 |
| matthewcroughan - nix.how | how people will ultimately use this stuff will be as a flake, since nixos-hardware doesn't have any flake-compat | 17:12:57 |
| Alyssa Ross | you can use nixos-hardware just fine using any of the mechanisms you'd use to evaluate nix code | 17:13:01 |
| Alyssa Ross | I personally use a git subtree | 17:13:06 |
| Alyssa Ross | you could also use niv | 17:13:09 |
| Alyssa Ross | or a channel | 17:13:15 |
| Alyssa Ross | literally any of the existing mechanisms | 17:13:21 |
| matthewcroughan - nix.how | should there be a packages output to the nixos-hardware flake? | 17:15:06 |
| matthewcroughan - nix.how | it could contain all of the kernel packages for a given hardware name | 17:15:39 |
| Alyssa Ross | it should definitely expose its packages without having to go through a NixOS evaluation, yeah, that's another improvement I'd like to see | 17:15:46 |
| matthewcroughan - nix.how | packages.linuxPackages_visionfive | 17:15:50 |
| Alyssa Ross | (and it should do it in a non-flakes way as well) | 17:15:56 |
| matthewcroughan - nix.how | baggage | 17:16:03 |
| matthewcroughan - nix.how | to do things in a non-flake way is baggage, but do you see the conflict? | 17:16:13 |
| Alyssa Ross | again, i have no idea what you mean | 17:16:34 |
| matthewcroughan - nix.how | to avoid using flakes for this, is going to make the resulting code much more complex and harder to read | 17:16:49 |
| matthewcroughan - nix.how | less simple | 17:16:51 |
| matthewcroughan - nix.how | but we have to do that if we want to get out of nixpkgs whilst flakes aren't stable | 17:17:00 |
