| 25 Nov 2025 |
 goldstein | 
Download clipboard.png | 15:31:12 |
 Atemu | I want it to be direct and acually work with the flake.nix "standard" such that I can throw it at some local flake project and eval it without flake tooling BS but I'd probably need to solve many of the same issues as you likely have | 15:31:47 |
 Sofie 🏳️⚧️ (she/her) | well, the issue is that there's really no "standard" | 15:32:47 |
| Atemu | Yeah I know, hence the quotes.. | 15:33:06 |
| Sofie 🏳️⚧️ (she/her) | you'd have to solve that but if we are doing that, why not just fix the whole standard | 15:33:07 |
| goldstein | I think the big question here is whether you want to respect transitive dependency lockfiles
I explicitly do not respect them (the problem I’m trying to solve is a billion of nixpkgs in my system closure), so I need to do a bit of extra work to re-solve dependencies | 15:33:24 |
| goldstein | if you do, then I think just manually dependency-injecting is not that hard | 15:33:41 |
| Atemu | What I mean is the structure of the flake.nix; inputs, outputs and the schema of the outputs | 15:33:38 |
| Sofie 🏳️⚧️ (she/her) | * | 15:33:39 |
| goldstein | if you do not want to respect them, you’ll probably be able to take my solver and just plug in another “backend”, my design is pretty modular | 15:34:49 |
| Atemu | Honestly for any remote inputs, it can do pure eval or whatever it wants; I just want nix projects on my local filesystem to function sensibly | 15:35:25 |
| Atemu | (Like, what Nix would otherwise do with a flake) | 15:35:50 |
| Sofie 🏳️⚧️ (she/her) | Anyways, now that we have a lix committee, times for an proper standard honestly | 15:36:17 |
| goldstein | preach
I did so much reverse engineering for this project | 15:36:35 |
| Sofie 🏳️⚧️ (she/her) | or at least start work on a RFC or something | 15:36:38 |
| Atemu | @goldstein:tty5.dev: Could I DM you an SSH pubkey for access to the WIP code? | 15:37:35 |
| goldstein | oh, one more fun fact: flake.nix is semi-implemented in nixlang, see https://git.lix.systems/lix-project/lix/src/commit/27a94369a9a76c7e41220060b85b984e3dc6da4e/lix/libexpr/flake/call-flake.nix | 15:38:56 |
| goldstein | I can just DM you the file if you want to, it’s not secret or anything, I just need to polish a bit + write some docs | 15:39:19 |
| Atemu | That'd be awesome :)
I don't expect good code quality from a WIP/PoC ;)
| 15:40:28 |
| goldstein | Atemu if matrix doesn’t stop matrixing, you can ping me at root@goldstein.lol emali | 15:45:17 |
| goldstein | * Atemu if matrix doesn’t stop matrixing, you can ping me at root@goldstein.lol email | 15:45:22 |
| Atemu | Thanks lol ^^' | 15:45:54 |
 aloisw | Isn't that what https://git.lix.systems/lix-project/flake-compat is supposed to achieve with copySourceTreeToStore = false? | 17:07:33 |
| aloisw | Of course if the code in the flake itself assumes that it's in the store (which is a valid assumption under flakes) it will either copy anyway or fail. | 17:08:43 |
| Atemu | This was pointed out to me independently 20min ago; will have a look! | 17:11:19 |
 raitobezarius | In reply to @piegames:flausch.social
Do we have a command to easily check (and debug) remote builder connections? If not, can I request that as a feature? Already planned | 17:14:17 |
| raitobezarius | But the new machines format would make this greatly easier | 17:15:03 |
 piegames | I'm trying to run nix-instantiate --parse within a derivation builder (for Flaker purposes), and it fails with error: creating directory '/nix/var': Permission denied | 17:33:30 |
| piegames | this makes no sense to me. Why does nix want to create /nix/var directory? I didn't ask it to! I just want it to parse that silly file | 17:33:57 |
| aloisw | It does a bunch of filesystem accesses on startup, no matter what you end up actually using. | 17:37:22 |
