!9IQChSjwSHXPPWTa:lix.systems

Lix

1194 Members
Lix user channel. Feel free to discuss on-topic issues here and give each other help. For matrix.to links to the rest of the Lix channels, see: https://wiki.lix.systems/books/lix-organisation/page/matrix-rooms329 Servers

Load older messages


SenderMessageTime
29 Jun 2026
@grimmauld:m.grimmauld.deGrimmauld (any/all) the etc overlay things look interesting. I should review those some time soon.... Always a good day when nikstur deletes more bash monstrosities 20:39:08
@rosssmyth:matrix.org@rosssmyth:matrix.orgtrue...21:21:28
@rosssmyth:matrix.org@rosssmyth:matrix.org

This is how I do it
https://github.com/RossSmyth/nixos/blob/main/mkMachine.nix

Works well enough. I also have a bash wrapper about nixos-rebuild that does things like:

  1. nom
  2. -I $(npins get-path nixpkgs)
  3. --attr $(hostname)
  4. --show-trace
21:24:41
@blokyk:matrix.orgzoë (@blokyk)yeah, in practice i also just use a wrapper that builds with nom and activate the new config21:26:16
@blokyk:matrix.orgzoë (@blokyk) but even just having a manual default.nix entry point is so nice for bootstrap, you just do $(nix-build .)/activate and everything works perfectly immediately :) 21:27:43
30 Jun 2026
@buffet:fairydust.spacebuffet, das changed their display name from buffet to buffet, das.10:20:28
@rootile:flausch.socialRutilehttps://github.com/NixOS/nixpkgs/pull/537162 i created a pr in nixpkgs to use our new toml formatting. if anyone has some time for a review, it would be appreciated20:36:29
@rootile:flausch.socialRutile* https://github.com/NixOS/nixpkgs/pull/537162 i created a pr in nixpkgs to use our new toml formatting for remote builders. if anyone has some time for a review, it would be appreciated20:36:40
1 Jul 2026
@easy6:matrix.org@easy6:matrix.org left the room.11:30:40
@blokyk:matrix.orgzoë (@blokyk) *

sorry i was on my phone and should have looked it up before saying that.
i actually meant .overrideScope, which will allow you to have the correct lix derivation used by other derivations in the lixPackageSets.latest set.
for example, you'd do something like:

finalPkgs: prevPkgs: {
  lixPackageSets = prevPkgs.lixPackageSets.extend (finalSet: prevSet: {
    latest = prevSet.latest.overrideScope (finalScope: prevScope: {
      lix = prevScope.lix.override (prevArgs: {
        curl = prevArgs.curl.overrideAttrs { patches = []; };
      });
    });
  });
}
12:23:28
@blokyk:matrix.orgzoë (@blokyk) i'm sure you considered it, so i'd rather ask here than on github: is there a specific reason you chose to make a separate option/module for this, instead of "translating" nix.buildMachines (like some of the stuff in programs/lix.nix)? 13:22:43
@blokyk:matrix.orgzoë (@blokyk) * i'm sure you considered it, so i'd rather ask here than on github: is there a specific reason you chose to make a separate option/module for this, instead of "translating" nix.buildMachines (like some of the stuff in programs/lix.nix does)? 13:23:34
@blokyk:matrix.orgzoë (@blokyk) * i'm sure you considered it, so i'd rather ask here than on github: is there a specific reason you chose to make a separate option/module for this, instead of programmatically "translating" nix.buildMachines (like some of the stuff in programs/lix.nix does, based on nix.package.pname)? 13:27:55
@rootile:flausch.socialRutiledifferent naming of things, and slowly diverging formats and options. (this already is an attrset with named builders instead of just a list) and we will continue to introduce new properties like groups/tags which just aren't possible to handle with the current implementation14:20:25
@blokyk:matrix.orgzoë (@blokyk)okay, that's what i suspected but wanted to know if there were any other considerations. this isn't great ux/discoverability-wise, but oh well14:22:02
@blokyk:matrix.orgzoë (@blokyk)* okay, that's what i suspected but wanted to know if there were any other considerations i missed. this isn't great ux/discoverability-wise, but oh well14:22:12
@rootile:flausch.socialRutilecould we improve discoverability more? (i did add a release note; and we could also advertise it as lix)14:22:43
@rootile:flausch.socialRutileas well as at some point throwing a deprecation warning when using the old format and lix14:23:27
@blokyk:matrix.orgzoë (@blokyk) one idea would be to add a deprecation warning (gated on nix.package.pname to check whether it's lix), but that's maybe a little agressive 14:23:45
@blokyk:matrix.orgzoë (@blokyk) * one idea would be to add a deprecation warning (gated on nix.package.pname to check whether it's lix), but that's maybe a little agressive for now 14:23:49
@rootile:flausch.socialRutilemy words exactly :D14:23:59
@blokyk:matrix.orgzoë (@blokyk)haha sorry14:24:06
@rootile:flausch.socialRutilei'd argue for at least one nixpkgs release cycle before deprecating (i.e. throw warning in 27.05 and then potentially remove support in 27.11 or 28.05)14:25:55
@rosssmyth:matrix.org@rosssmyth:matrix.orgI found a weird pain point last night. For some reason the Niri Cachix started timing out, so while I was developing a program every time I would run "nix build -f." it would just hang waiting for Cachix to reply, which never came. To get around it I just started using "nix build --offline -f.", but the current timeout seems pretty long for waiting for cache replies as it took a very long time (I didn't let it finish, but at least five minutes) to stop waiting. 14:56:17
@raitobezarius:matrix.orgraitobezariusfwiw, i think we should do a lix namespace directly20:56:56
@raitobezarius:matrix.orgraitobezarius https://github.com/NixOS/nixpkgs/pull/445223 20:56:57
@raitobezarius:matrix.orgraitobezariusi have a personal module that contains already a toml formatter for remote builders that i have been using for months now20:57:11
@raitobezarius:matrix.orgraitobezariuswhich extends the lix/ one20:57:14
@raitobezarius:matrix.orgraitobezariuscleaning up the overlay and getting that merged is probably a good idea in addition of getting your changes20:57:27
@raitobezarius:matrix.orgraitobezariusi don't think we should 1:1 replicate the buildMachines API20:57:34

There are no newer messages yet.


Back to Room ListRoom Version: 10