| 18 Dec 2024 |
 guiniol | so, cache is setup, and I tried to build lix for it, but... it fetched it from cache.nixos.org. So why isn't the nixos system doing the same... | 23:12:53 |
| guiniol | Is there a way to see which architecture/sub architecture/march nix build is targeting? I imagine my old machine is slightly different and that's why it's not fetching from the cache | 23:36:34 |
| guiniol | builtins.currentSystem is the same on both machines | 23:39:14 |
| guiniol | what can I check to find out the differences between the systems that force it to rebuild? | 23:39:32 |
 Grimmauld | Maybe you can nix-diff it? | 23:40:57 |
| Grimmauld | Getting the baseline should be the easy part with a nix copy, but I suppose getting the compare rev would be kinda Tricky if your compile never finished | 23:42:39 |
| guiniol | lix should finish at some point since it's not the first time it has to compile it (and I do keep interrupting it to try stuff around the cache) | 23:45:30 |
| guiniol | I would like to know how it checks the cache and decides to build instead | 23:45:53 |
| guiniol | since lix is available in cache.nixos.org | 23:47:01 |
| guiniol | is there a way to debug that | 23:47:15 |
| Grimmauld | In reply to @guiniol:matrix.org
since lix is available in cache.nixos.org Only if you have a cached branch and no overlays messing with the caches | 23:47:47 |
| Grimmauld | I suppose you could also go clone nixpkgs, git switch <branch> and nix build .#lix and see what happens if you just build lix, independent of surrounding config. You'd not be dealing with wonky overlays and stuff. If that still doesn't cache, something else is seriously wrong | 23:50:34 |
| Grimmauld | ((if you are doing this already, then uhhhh.... But your earlier statement "why isn't the nixos system doing the same..." makes me guess you are trying to build a nixos config. Which obviously has more things that can interfere)) | 23:52:25 |
| 19 Dec 2024 |
| guiniol | so, one host is not nixos and that's the one I am using as a cache and that fetched from cache.nixos.org | 00:01:10 |
| guiniol | it is using the unstable branch | 00:01:17 |
| guiniol | the other system is a nixos system with flakes | 00:01:25 |
| guiniol | and I just followed the instructions in "how to install lix" | 00:01:43 |
| guiniol | I've got a lix-module in the input, pointing to 2.91.1-2 and following nixpkgs | 00:02:34 |
| guiniol | maybe there's a new version and that's why the old one is not in the cache? But it also built the previous version | 00:02:52 |
| guiniol | looking at the branches on nixpkgs, there is nixos-unstable and nixpkgs-unstable, so maybe that's the diff? Since the non-nixos machine is probably using the nixpkgs-unstable while the nixos machine is using nixos-unstable I would guess | 00:05:27 |
| guiniol | should lix be one of the packages in systemPackages? | 00:06:56 |
| guiniol | and why is it even rebuilding lix if there is no new version?? | 00:18:00 |
 Adam Chance | Inputs could’ve changed | 00:28:23 |
| Grimmauld | In reply to @guiniol:matrix.org
I've got a lix-module in the input, pointing to 2.91.1-2 and following nixpkgs Oh yeah the overlay points to git versions of lix last I checked.
Just set nix.package = pkgs lix; and drop all the overlays/modules and you'll get caches | 00:29:25 |
| guiniol | yes, I had just noticed the beginning of that section saying It will result in building Lix from source, which takes a few minutes on every update, which is a perfect time to get up, get some water, and stretch for a bit. | 00:30:51 |
| guiniol | but | 00:33:02 |
| guiniol | if I could get the cache machine to build the same thing, it could fetch it from there, yes? | 00:33:18 |
| guiniol | I don't see how I could do that as I don't understand how the package is even defined. Something to do with the follows maybe? | 00:55:41 |
| guiniol | alright, just used nixos-rebuild build. Let's see if that gets into the cache | 01:00:32 |
|  @bbigras:matrix.org left the room. | 01:36:02 |
