| 27 Dec 2025 |
 K900 | Because in fact systemd good | 16:12:07 |
| K900 | (mostly) | 16:12:13 |
 Spectral (he/him) | I mean if there evet will be something like LixOS then it should probably not be named that to avoid recreating the whole naming headache that nix has | 16:12:50 |
| K900 | It will not be named LixOS no | 16:13:28 |
 Coca | "it's up to you to break generational trauma" | 16:15:25 |
 Acid Bong | at least it's more convenient than Guix being used for both the repo (guixpkgs*), the OS (guixos*) and the tool
(ik the OS is technically called Guix System, but most ppl drop "system") | 16:16:21 |
 piegames | In reply to @k900:0upti.me
There is some work ongoing on the Rust parser, you want to talk to @piegames about that https://git.lix.systems/piegames/lix-parser it has started ✨ | 16:22:27 |
| piegames | Living out of tree while still in the prototype stage | 16:22:48 |
 daslixou | winnow, I see | 16:23:01 |
| piegames | So to set expectations, the timeline for amy of this is "hopefully still in 2026 but no guarantees" | 16:23:17 |
| piegames | * So to set expectations, the timeline for any of this is "hopefully still in 2026 but no guarantees" | 16:23:31 |
| daslixou | Redacted or Malformed Event | 16:23:32 |
| daslixou | time to make a lix account iguess | 16:26:27 |
| daslixou | I have to ask, from where are the whole stateful, bytecode, parserstate and thunk things from? I want to make sense of how they'd be used | 16:36:08 |
 madmccrow | It's more like : I don't like the way to have latest and stable nixpkgs side by side when developing for machines with radically different missions (ie. Gaming desktop, vs backup NAS) .
If you add the fact that the flake layout might require you to have some inputs twice ( like for example if you wanna use Lanzaboote with both stable and latest).
A lockfile by machine would be better by a mile, but I don't know how to do it without splitting my work in multiple repos.
Because I've read that the lix community wasn't satisfied with flakes in it's current state, I would have thought some alternative was being used. | 16:51:34 |
| Acid Bong | then have two Nixpkgs inputs and define hosts from them, there's no problem with that | 16:53:28 |
 QuadRadical (Ping) | but i suppose it would cause you to potentially download an extra unused copy of nixpkgs | 16:54:18 |
| K900 | lanzaboote should be an overlay no? | 16:56:51 |
| piegames | In reply to @daslixou:matrix.org
I have to ask, from where are the whole stateful, bytecode, parserstate and thunk things from? I want to make sense of how they'd be used That's locally defined in the same file, ignore the bytecode scaffolding that's for later | 17:05:35 |
| daslixou | no like, from what other parser is it? | 17:05:57 |
| daslixou | the bytecode thing looks similar to rustpython, but then it's pretty different afterall and I haven't seen such direct mapping from parser to bytecode elsewhere | 17:08:51 |
| madmccrow | Yeah. It is, but for flakes you have to have it in input and use the follows notation. | 17:35:49 |
| K900 | You don't necessarily | 17:36:56 |
| K900 | The nixpkgs flake input should not be used if you're using the NixOS module | 17:37:05 |
| madmccrow | See, to me, it looks like the famous xkcd comic about too many standards. I feel like every nix project uses a slightly different approach, making my OCPD twitch looking at imperfect code. | 17:45:27 |
| K900 | Overlays are generally the Right Way | 17:50:01 |
| K900 | Specifically for this very reason | 17:50:05 |
| piegames | In reply to @daslixou:matrix.org
the bytecode thing looks similar to rustpython, but then it's pretty different afterall and I haven't seen such direct mapping from parser to bytecode elsewhere It's from nowhere because that's my own code. Probably nobody sane would attempt single-pass compailation directly in the parser | 18:02:44 |
| piegames | (The insanity is partially courtesy of horrors) | 18:05:06 |
| madmccrow | Exactly my thoughts. I kinda regret the old /nix/etc way of having a config there, and downloading and building only what's necessary.
I feel like Flakes are very interesting for developing projects and reusable libraries, but they feel way too limiting for managing machines ( secrets, impurities, etc., are still an unsolved issue in my - probably uneducated - opinion)
| 18:10:45 |
