| 12 Oct 2025 |
 emily | do the nice thing on staging, add the bodges to make it not a mass rebuild when backporting to -next. | 17:09:50 |
 matthewcroughan | ah okay it's because it's conditional on musl it avoids the rebuild labels | 17:10:04 |
| emily | that way the most you can neglect to do is getting the fix early, rather than neglecting to clean up. | 17:10:12 |
 Yureka (she/her) | that's because hydra doesn't build pkgsMusl.systemd | 17:10:12 |
| Yureka (she/her) | or anything depending on it | 17:10:16 |
| matthewcroughan | What's needed to make hydra keep track of this? :P | 17:10:25 |
| matthewcroughan | I wanted to make a flake to keep track of it, I'm in the process of doing that atm | 17:10:34 |
| matthewcroughan | hercules is nice but hydra is made for nixpkgs more than hercules is | 17:10:49 |
| emily | you can run your own Hydra, building tons of stuff on dynamic Musl isn't a good use of our limited build resources | 17:11:14 |
| matthewcroughan | Yeah that makes sense, but I'm happy to put my resources to use on musl | 17:11:38 |
| matthewcroughan | My https://github.com/MatthewCroughan/nixos-musl/tree/refactor flake is keeping track of what is necessary in these profiles | 17:12:34 |
| matthewcroughan | idea is that musl.nix contains only want is currently necessary to build a nixpkgs.lib.nixosSystem {} on musl, and nothing more | 17:13:00 |
 Ihar Hrachyshka | on a relevant point: how do folks feel like about a packaging having a passthru.tests.static = pkgsStatic.<itself> to signal it is desired / expected to build? or is it redundant? | 17:13:05 |
| matthewcroughan | * idea is that musl.nix contains only what is currently necessary to build a nixpkgs.lib.nixosSystem {} on musl, and nothing more | 17:13:06 |
| emily | can be useful, OTOH is in tension with moving variant package sets to variants.nix which we'd generally like to do. | 17:13:44 |
| emily | it depends a bit on just how "expected" it is I think | 17:14:09 |
| emily | like "would people be okay with an update being held back just because it regresses pkgsStatic for that package" | 17:14:27 |
| emily | if it's part of a channel blocker then certainly | 17:14:33 |
| matthewcroughan | I don't want things to be held back, but it would be nice if as much as possible was cached, and that status could be checked | 17:14:51 |
| Ihar Hrachyshka | aren't passthru.tests already kinda advisory? (I don't think we gate on them and I don't see a lot of nixpkgs-reviews ran with -p passthru.tests ) | 17:15:25 |
| matthewcroughan | K900: Can I merge https://github.com/NixOS/nixpkgs/pull/451338 when ready? It's what I intended now that it's musl conditional | 17:18:40 |
| emily | caching is unlikely to happen | 17:18:57 |
| emily | I think if passthru.tests.X fails you're meant to be worried, at least. | 17:19:12 |
| matthewcroughan | I mean I've been doing an okay job warming up my own caches by doing as many builds as possible, at all times | 17:19:23 |
| emily | as in, you would usually add something to passthru.tests if it failing would be a reason for you to at least consider blocking the PR. | 17:19:29 |
| matthewcroughan | nix build .#checks.aarch64-linux.{base,musl,musl-llvm,gnu-musl} --keep-going for example | 17:19:37 |
| matthewcroughan | in my flake | 17:19:39 |
| matthewcroughan | so I'll let those build on a bunch of builders 24/7 as I make changes to files.. and it's better than if I hadn't attempted the builds at all | 17:19:54 |
| matthewcroughan | since at least then the base toolchains and stuff are cached | 17:20:43 |
| matthewcroughan | I spent a few hours to build https://github.com/NixOS/nixpkgs/pull/451274 | 19:58:11 |
