| 26 Oct 2023 |
 schmittlauch (he/him) | Or how is it exposed as a name/ attr? | 21:24:16 |
 K900 | No, poetry2nix is the flake input | 21:24:19 |
| K900 | So it'll resolve to the path of the poetry2nix repo | 21:24:27 |
| K900 | And it will import its default.nix | 21:24:54 |
| K900 | Which is defined like this: https://github.com/nix-community/poetry2nix/blob/master/default.nix | 21:24:57 |
| schmittlauch (he/him) | Yes, but when reading through the actual poetry2nix flake.nix file in the repo I do not understand what actually is imported there. | 21:25:15 |
| K900 | default.nix is what is imported | 21:25:55 |
| schmittlauch (he/him) | Oh well, is this a crude workaround or an actual intended way of using this? | 21:26:13 |
| K900 | We might export it as a function in the flake eventually | 21:26:40 |
| K900 | Similar to say nixpkgs.lib.nixosSystem | 21:26:46 |
| schmittlauch (he/him) | I thought the main idea is that you only reference explicit outputs of other flakes | 21:26:50 |
| K900 | But it is a perfectly fine way to use this | 21:26:53 |
| schmittlauch (he/him) | Okay. thx. | 21:27:10 |
| K900 | And it's better than the old way because it avoids yet another nixpkgs instance | 21:27:15 |
| * schmittlauch (he/him) is still a bit weirded out but starts to understand | 21:27:35 |
| schmittlauch (he/him) | In reply to @k900:0upti.me
And it's better than the old way because it avoids yet another nixpkgs instance Ah, that new instance only happens when using it as an overlay? I saw we now hardcode a list of permittedInsecurePackages. | 21:28:19 |
| K900 | That list is very old | 21:28:27 |
| K900 | And the new instance happens when we export the builder functions directly from the flake | 21:28:46 |
| K900 | Because those use the flake's nixpkgsd | 21:28:51 |
| K900 | * Because those use the flake's nixpkgs | 21:28:52 |
| K900 | And not your app's | 21:28:55 |
| schmittlauch (he/him) | I see. Thx. | 21:29:21 |
| schmittlauch (he/him) | Still not qualified enough to figure out the best design choices, but flakes often trigger significant cases of "wat!?" for me ^^ | 21:30:14 |
| schmittlauch (he/him) | In reply to @k900:0upti.me
And the new instance happens when we export the builder functions directly from the flake I see why you're doing it that way, but IMHO needing to do this is still against the basic claimed perks/ goals of flakes, e.g. that they have introspectable, discoverable and (hopefully) enumerable outputs that are then re-used by other flakes. | 22:05:01 |
| K900 | Again, we should probably export poetry2nix.lib.mkPoetry2Nix { pkgs } | 22:06:03 |
| schmittlauch (he/him) | If this has the same footprint regarding number of evaluated nixpkgs instances, then: yes, that'd be great (: | 22:14:41 |
| K900 | It dos | 22:15:13 |
| K900 | * It does | 22:15:15 |
| schmittlauch (he/him) | And as the legacyPackages output was only broken on master a few days ago, it is IMHO tolerable to break the exposed API again. | 22:15:31 |
| K900 | It doesn't break anything | 22:23:18 |
