| 7 Sep 2023 |
 DavHau | In reply to @infinisil:matrix.org
Maybe we should have pname match the attribute name too though Yes please. Establishing that pattern is also important for downstream projects. Associating the log of a failed build with the correct package file can often be a nightmare, especially for team members with little nix experience. | 20:55:43 |
| DavHau | Its also important if we want to better integrate package updates. We can attach an update script to a package via passthru, but that script needs to know the location of the package definition. Hardcoding the location breaks portability. A mechanism to compute the package location from the attribute path is needed. | 23:10:24 |
| DavHau | * Its also important if we want to better integrate package updates. We can attach an update script to a package via passthru, but that script needs to know the location of the package definition. Hardcoding the location breaks portability. A mechanism to compute the package location from the package name is benefitial. | 23:12:44 |
| 8 Sep 2023 |
 @piegames:matrix.org | I think currently update scripts are executed in the package's folder, and finding the Nixpkgs root can be done with a simple git command | 06:04:25 |
| @piegames:matrix.org | Update scripts should already be invariant to their location | 06:04:49 |
|  Alesya joined the room. | 09:59:01 |
 Artturin | When python...callPackage is is gotten rid of everything should be switched to pythonXXPackages instead of python.pkgs.
https://github.com/NixOS/nixpkgs/issues/211340 | 10:44:48 |
 figsoda | In reply to @infinisil:matrix.org
Ah neat! Note that the folder structure should be treated as internal to Nixpkgs, there is no API guarantee that it will not change over time. Considering that, it would be great if we could figure out a proper API that Nixpkgs could expose for this. Maybe it should be as simple as pkgs.preferredDirectoryForPackage "foo" = ./pkgs/by-name/fo/foo just implemented the feature in nix-init, I think one thing that would help is if some constants and shard_for_package are exposed in a crate | 18:34:46 |
 tomberek | infinisil: Amazing talk today. Very well done. Bravo! | 22:54:03 |
 infinisil | Thanks! ☺️ | 22:55:26 |
| infinisil | In reply to @figsoda:matrix.org
just implemented the feature in nix-init, I think one thing that would help is if some constants and shard_for_package are exposed in a crate It is internal, but we could maybe just mark it as such. Can you open an issue or a PR to discuss this? | 22:59:40 |
