| 14 Sep 2023 |
 K900 ⚡️ | And very bad at delimiting what is actually opinion and what is baseline required | 20:22:34 |
| K900 ⚡️ | And it's not really because nixos-hardware is bad necessarily | 20:23:50 |
| K900 ⚡️ | It's just that the line is in fact extremely blurry | 20:23:56 |
 emily | yeah, I researched and dug around for a bit longer and I am aware now.
I just didn't send a follow up message here.
I am planing on writing a jobset for hydra now (soon™) | 20:28:19 |
| K900 ⚡️ | I would also, FWIW, be very cautious about adding another kernel variant to nixpkgs | 20:29:34 |
| K900 ⚡️ | I'd argue we already have entirely too many | 20:29:39 |
| K900 ⚡️ | And it's already causing us pain in places | 20:30:15 |
| emily | (in case it wasn't clear -- I am no longer planing upstreaming that kernel into nixpkgs) | 20:33:00 |
| K900 ⚡️ | I kinda really wish we had a "universe" style source that's not generally expected to 100% work, but is still kinda blessed and maintained to best effort | 20:34:07 |
| emily | (on a related note: am I am allowed to merge kernel bumps via bors in nixos-hardware as a nixpkgs committer? :)) | 20:39:13 |
| K900 ⚡️ | I think you need the commit bit on nixos-hardware for that | 20:39:41 |
| emily | pretty sure committers have write access to nixos-hardware
https://github.com/orgs/NixOS/teams/nixpkgs-committers/repositories
or is this a bors thing? | 20:40:51 |
 woobilicious | In reply to @k900:0upti.me
And it's not really because nixos-hardware is bad necessarily Just to clarify, when I said it should be in nixpkgs, what I mean is that my hardware should work "as expected", without having to add a poorly documented extra repo (which was suggested to me on my first install). I'm not too familiar with the repo itself, I just know my system isn't currently using pstate which seems like a sane default to have for Zen, but yeah I'm against opinionated settings. | 23:41:03 |
| 15 Sep 2023 |
| K900 ⚡️ | nixos-hardware is a repository of opinionated-ish hardware specific workarounds | 04:24:17 |
| K900 ⚡️ | And amd_pstate is the default starting with kernel 6.5, which is when upstream decided it is appropriate to enable by default. | 04:25:11 |
| woobilicious | I guess I miss understood the point of it then. | 04:28:40 |
| K900 ⚡️ | And the problem with nixos-hardware is mostly that the line between "workaround" and "solution" is pretty thin | 04:32:37 |
| K900 ⚡️ | Like, it enables custom kernels for Surface or Apple devices | 04:32:50 |
| K900 ⚡️ | Those are entire other kernel trees, often multiple patches behind | 04:33:06 |
| K900 ⚡️ | Are we OK with doing that to people by default? | 04:33:14 |
| K900 ⚡️ | I don't know | 04:33:16 |
| K900 ⚡️ | If the computer doesn't work without it, then yeah, probably | 04:33:50 |
| K900 ⚡️ | But then where is the line | 04:34:01 |
| woobilicious | In reply to @k900:0upti.me
And amd_pstate is the default starting with kernel 6.5, which is when upstream decided it is appropriate to enable by default. Mines still disabled and I'm on 6.5 🤷 | 04:44:24 |
| K900 ⚡️ | It also requires firmware support | 04:44:37 |
| K900 ⚡️ | So it's very possible your hardware just doesn't have it | 04:44:50 |
| K900 ⚡️ | Especially if it's older | 04:44:57 |
| K900 ⚡️ | I think it should be available for all Zen3 and above at least | 04:45:20 |
| woobilicious | Yeah I'm on Zen 3 | 04:45:53 |
| woobilicious | dmesg says: amd_pstate: driver load is disabled, boot with specific mode to enable this | 04:46:43 |
