| 19 Oct 2025 |
matthewcroughan | because I thought I already had fixed it in my override | 10:27:08 |
matthewcroughan | # https://github.com/NixOS/nixpkgs/pull/451506
python3 = super.python3.override {
packageOverrides = pyfinal: pyprev: {
pytest = pyprev.pytest.overrideAttrs {
dontWrapPythonPrograms = false;
};
};
};
| 10:27:52 |
Puna | * https://github.com/NixOS/nixpkgs/pull/238331
function equality in Nix is unreliable, and the elaborated platform attrsets contain functions | 10:28:49 |
matthewcroughan | But when the VM test gets invoked, somehow this goes out the window | 10:33:01 |
matthewcroughan |  Download image.png | 10:33:10 |
matthewcroughan | Like the above overlay works fine for the nixos system, but not for a vmtest of that same config | 10:35:04 |
matthewcroughan | It's so weird, it's like the test framework is ignoring my overlays | 10:42:18 |
matthewcroughan | # See https://nixos.org/manual/nixpkgs/unstable/#tester-invalidateFetcherByDrvHash
# or doc/build-helpers/testers.chapter.md
nixosTest =
let
/*
The nixos/lib/testing-python.nix module, preapplied with arguments that
make sense for this evaluation of Nixpkgs.
*/
nixosTesting = (
import ../../../nixos/lib/testing-python.nix {
inherit (stdenv.hostPlatform) system;
inherit pkgs;
extraConfigurations = [
(
{ lib, ... }:
{
config.nixpkgs.pkgs = lib.mkDefault pkgsLinux;
}
)
];
}
);
in
test:
let
loadedTest = if builtins.typeOf test == "path" then import test else test;
calledTest = lib.toFunction loadedTest pkgs;
in
nixosTesting.simpleTest calledTest;
| 10:43:50 |
matthewcroughan | 🧠| 10:43:55 |
matthewcroughan | Yeah I don't think there's a way to fix that with an overlay | 10:54:27 |
matthewcroughan | that pytest thing broke a lot | 10:54:39 |
matthewcroughan | I found that -Oz makes a minimal systemd closure 20M -> 18M | 11:22:34 |
matthewcroughan | I now wonder what space saving effect that could have on the rest of the system if it were applied globally | 11:22:52 |
matthewcroughan |  Download image.png | 11:51:05 |
matthewcroughan | weird, why does the systemd initrd depend on two util-linux? | 11:51:14 |
matthewcroughan | Will this work with https://github.com/NixOS/nixpkgs/blob/nixos-unstable/pkgs/os-specific/linux/systemd/default.nix#L496-L498 existing? | 12:07:54 |
Grimmauld (any/all) | i have no idea, i will find out | 12:09:10 |
Grimmauld (any/all) | currently the thing is giving me trouble and i don't know nearly enough meson for this shit | 12:09:26 |
Grimmauld (any/all) | i wanted to do an unstable build, turns out its all kinds of busted | 12:09:39 |
Grimmauld (any/all) | ugh its just broken with LTO | 12:20:05 |
matthewcroughan | What is, systemd? | 12:22:21 |
matthewcroughan | Yeah systemd lto seems not to work | 12:22:28 |
Grimmauld (any/all) | no, pwaccess | 12:22:35 |
matthewcroughan | oh, that too then I guess | 12:22:40 |
Grimmauld (any/all) | its all kinds of broken, but i have a build going | 12:22:56 |
matthewcroughan | I was curious how much space systemd would save with LTO | 12:22:59 |
Grimmauld (any/all) | only took 3 patches... | 12:23:04 |
Grimmauld (any/all) | okay, one patch was unnecessary, but i'll need yet another patch to resolve the cyclic deps against pam | 12:46:21 |
Grimmauld (any/all) | but thats alright, i know pam pretty well. I already chased cyclic deps on their last update... | 12:47:15 |
matthewcroughan | Huh yeah util-linux-minimal is still used even if you set util-linux to the full one in an overlay | 13:11:32 |