!UUqahLbShAYkkrXmKs:matrix.org

DevOS

34 Members
Seeking help and geeking out together on https://github.com/divnix/devos & https://github.com/divnix/digga10 Servers

Load older messages


SenderMessageTime
12 Jul 2021
@blaggacao:matrix.orgDavid Arnold (blaggacao) *

I'd be greateful for some preliminary feedback on these two PRs that go hand in hand:

  • https://github.com/divnix/devos/pull/338/commits/fb67fc51cc44ce685f99cbdc40d359cc501d427d
  • https://github.com/divnix/digga/pull/84/commits/e2c862ddbec87fe7c2e675ab06c4d50d8800abe0

before I (hopefully) finish those PRs off tonight.

22:33:50
@blaggacao:matrix.orgDavid Arnold (blaggacao) *

I'd be greateful for some preliminary feedback on these two commits that go hand in hand:

  • https://github.com/divnix/devos/pull/338/commits/fb67fc51cc44ce685f99cbdc40d359cc501d427d
  • https://github.com/divnix/digga/pull/84/commits/e2c862ddbec87fe7c2e675ab06c4d50d8800abe0

before I (hopefully) finish those PRs off tonight.

22:34:00
16 Jul 2021
@teutat3s:pub.solar@teutat3s:pub.solar

Does someone else get this, too? (on Devos master branch and with nixpkgs commit 53079ed913181b2f945cf1580746828b57d47edb)

error: value is null while a set was expected

       at /nix/store/9vividw3p8cqwndq7k1sqknnsvinh1ql-source/flake.nix:52:19:

           51|                 moduleDeclarationFile =
           52|                   (builtins.unsafeGetAttrPos "modules" args).file;
             |                   ^
           53|
(use '--show-trace' to show detailed location information)
15:48:18
@kraftnix:matrix.org@kraftnix:matrix.org I got this yesterday, my "fix" was just moving to develop branch on both devos and digga. 15:49:40
@kraftnix:matrix.org@kraftnix:matrix.orgYou could try track down a commit which fixes it amongst the changes in devos/digga, but I was planning on updating devos/digga anyway so I just did that.15:50:51
@teutat3s:pub.solar@teutat3s:pub.solarAlright, thanks for the quick response, I'll wait until the next Devos release 😇15:53:21
@blaggacao:matrix.orgDavid Arnold (blaggacao)THis issue is also reported her: https://github.com/divnix/digga/issues/8715:56:14
@blaggacao:matrix.orgDavid Arnold (blaggacao) * This issue is also reported her: https://github.com/divnix/digga/issues/8715:56:22
@blaggacao:matrix.orgDavid Arnold (blaggacao) /nix/store/9vividw3p8cqwndq7k1sqknnsvinh1ql-source/flake.nix is not a file of digga or devos. WHere does this file come from? 15:57:58
@blaggacao:matrix.orgDavid Arnold (blaggacao) * This issue is also reported here: https://github.com/divnix/digga/issues/8715:58:10
@teutat3s:pub.solar@teutat3s:pub.solar
In reply to @blaggacao:matrix.org
This issue is also reported here: https://github.com/divnix/digga/issues/87
Thanks, subscribed
15:58:18
@teutat3s:pub.solar@teutat3s:pub.solar It looked to me like native Nix flake code 15:58:51
@teutat3s:pub.solar@teutat3s:pub.solar * It looked to me like “native” Nix flake code 15:59:05
@blaggacao:matrix.orgDavid Arnold (blaggacao) Unrelated, but important: Since https://github.com/divnix/devos/pull/342, if you use an unpatched nix version (without https://github.com/NixOS/nix/pull/4641), then you';; need to add this anti corruption layer to your flake.nix 16:06:35
@kraftnix:matrix.org@kraftnix:matrix.org

So I haven't made an iso with flk for while but there seems to be a regression (at least in terms of how I was using it. I have a basic host I use for a super slim iso that pulls in one of my VM users that has some packages + my ssh keys in authorizedKeys. It looks like this:

{ lib, suites, profiles, ... }:
{
  imports = lib.flatten [
    suites.base
    profiles.users.operator
  ];
  services.openssh.enable = true;

  boot.loader.systemd-boot.enable = true;
  boot.loader.efi.canTouchEfiVariables = true;
  networking.interfaces.enp1s0.useDHCP = true;

  fileSystems."/" = { device = "/dev/disk/by-label/nixos"; };
}

Previously this worked fine, and I could ssh into the VM/machine as my operator user, however in the most recent develop branches of devos/digga, the operator user is no longer included when creating an iso with flk iso, I cannot see anything obvious that removes my user, other than perhaps the disabledModules config in isoConfig in digga.

My question would be, is there an easy way to include certain profiles/suites in the iso?

16:24:00
@blaggacao:matrix.orgDavid Arnold (blaggacao) Maybe the quickest way is to not expose users as profiles 16:29:51
@kraftnix:matrix.org@kraftnix:matrix.orgdirect import the operator user as a file path rather than through profiles.users?16:30:36
@blaggacao:matrix.orgDavid Arnold (blaggacao) This indeed seems to be a regression of the way users are plugged into profiles in https://github.com/divnix/devos/commit/d9082066f7bc5bd6a58ab55418db3d6abf4da3d6 16:30:39
@kraftnix:matrix.org@kraftnix:matrix.orgI'll give it a test, I would normally test this kind of stuff myself by running squashimg is a PIA that takes too long :/16:31:24
@kraftnix:matrix.org@kraftnix:matrix.org * I'll give it a test, I would normally test this kind of stuff myself by running quickly but squashimg is a PIA that takes too long :/16:31:36
@blaggacao:matrix.orgDavid Arnold (blaggacao)Wait a sec, I think my previous analysis is wrong. I'll look into it after some stuf here.16:33:29
@kraftnix:matrix.org@kraftnix:matrix.orgI wish mksquashfs was parallelisable or faster :/ i mean through pcie4 nvmes at it helps but otherwise it doesn't run much faster on my laptop vs workstation16:34:39
@blaggacao:matrix.orgDavid Arnold (blaggacao) I think the reason is that all profiles that are incorporated within a suite are disabled by default. So if a user is incorporated into a suite, that would be disabled as well. The reason for disabling is to avoid systemd startups that you don't need on a minimal boostrap image. 16:35:14
@b12f:pub.solarb12fI'm with you that stuff is so slow. We've stopped doing automated ISO builds and instead only do them manually in our CI because it just hogs server resources16:35:23
@blaggacao:matrix.orgDavid Arnold (blaggacao)
In reply to @blaggacao:matrix.org
I think the reason is that all profiles that are incorporated within a suite are disabled by default. So if a user is incorporated into a suite, that would be disabled as well. The reason for disabling is to avoid systemd startups that you don't need on a minimal boostrap image.
So I guess if users.operator is also within suites (anywhere) it will get filtered.
16:37:02
@kraftnix:matrix.org@kraftnix:matrix.org Also the fact that I nearly always have to run flk iso with GC_DONT_GC=1 is mildly annoying 16:37:14
@kraftnix:matrix.org@kraftnix:matrix.org
In reply to @blaggacao:matrix.org
So I guess if users.operator is also within suites (anywhere) it will get filtered.
it does not
16:37:37
@blaggacao:matrix.orgDavid Arnold (blaggacao)Oh yeah, I had that one, too, once.16:37:38
@kraftnix:matrix.org@kraftnix:matrix.orgdirect importing the user works16:37:43
@kraftnix:matrix.org@kraftnix:matrix.org
In reply to @b12f:pub.solar
I'm with you that stuff is so slow. We've stopped doing automated ISO builds and instead only do them manually in our CI because it just hogs server resources
i complain about it, but my min setup takes ~2mins on my workstation vs ~5mins on my laptop. i am glad pcie4 nvme is so much nicer, well worth the money, it literally is almost 3x faster than the nvme's in my laptop
16:40:09

Show newer messages


Back to Room ListRoom Version: 6