| 8 Mar 2024 |
 @jade_:matrix.org | the last category is extremely unfit for purpose imo: there is no way of searching all of them cleanly | 21:06:05 |
 Qyriad | It definitely is, yeah. We try to think of it as describing the roles that things fill, at least as the primary focus. "makeScoope is used in these places to create a package set with these properties" is easier to think about than trying to reverse engineer all the pre- and post- assumptions of a poorly documented behavior | 21:06:07 |
 infinisil | Agreed, I'd be in favor of making the contributor docs use some better rendering | 21:06:28 |
| infinisil | I don't think it should be in the Nixpkgs manual though, this is where it was before https://github.com/NixOS/nixpkgs/pull/245243 | 21:06:54 |
| @jade_:matrix.org | (also, the nixpkgs issue tracker is unusable for these kinds of meta issues ime because they get buried; is there some trick y'all use for actually putting these bugs somewhere they aren't lost?) | 21:07:39 |
| infinisil | jade_: Labels are somewhat effective. I'll soon create an "architecture" label, but I also created the "significant" label with decent success | 21:08:59 |
| infinisil | Honestly at some point I'd like Nixpkgs to be split into two repos, where one is the builders repo and the other is the package collection | 21:10:34 |
| @jade_:matrix.org | these things wind up really squishy tho | 21:10:49 |
| infinisil | Yeah.. needs a lot of thought. But it would make issue tracking much better | 21:11:11 |
| @jade_:matrix.org | there's a lot of surface area where it is not obviously one or the other. like, haskell or such.
I really think the github issue tracker and our tooling is the issues problem | 21:11:26 |
| infinisil | Or maybe we just need better issue/PR filtering links/support | 21:11:28 |
| @jade_:matrix.org | yeah | 21:11:31 |
| @jade_:matrix.org | I think we need a better issue search tool | 21:11:39 |
| infinisil | We could maybe add automatic labels based on whether builders, packages, or both are changed | 21:12:25 |
| @jade_:matrix.org | that would help for pr filtering yes | 21:12:39 |
| infinisil | It's hard to untangle this via files, but pkgs/by-name will help here | 21:12:48 |
| @jade_:matrix.org | but for "does it touch this attribute", there is no way to find prs lol | 21:12:50 |
| infinisil | Oh yeah, I wish github had a "search PRs by touched files" feature! | 21:13:18 |
| @jade_:matrix.org | In reply to @infinisil:matrix.org
It's hard to untangle this via files, but pkgs/by-name will help here i have a galaxy brained idea which is to associate the file of a package from the meta.location with the attr name, then collect a db table of pr files touched so i can search prs by attr path | 21:13:45 |
| @jade_:matrix.org | https://pad.lassul.us/opendata-projects# wrote some of this in here | 21:14:11 |
| infinisil | jade_: Nice! | 21:15:40 |
| infinisil | Generally I think more post-Nixpkgs processing would be really nice | 21:15:54 |
| @jade_:matrix.org | yeah | 21:15:59 |
| infinisil | programs.sqlite is an example, and it's really valuable | 21:16:04 |
| infinisil | Here's a smallish project: Replace programs.sqlite with nix-index | 21:16:29 |
| infinisil | Would deprecate https://github.com/Mic92/nix-index-database and give everybody a nix-index database by default | 21:16:50 |
| infinisil | Well, as long as you fetch channels and not git branches | 21:17:00 |
| @jade_:matrix.org | hm. channels have been killed, but we could make a hydra job for the nix-index of the current channel version.... or something? | 21:17:11 |
| @jade_:matrix.org | well | 21:17:14 |
| @jade_:matrix.org | flakes have killed channels | 21:17:18 |
