| 16 Jun 2026 |
 Capypara | * | 17:24:56 |
| Capypara | or maybe make one big PR for staging... might be easier to deal with | 17:25:44 |
 winston | there's some automation you can stumble on, e.g.
nix-shell ./maintainers/scripts/update.nix --arg predicate '(path: pkg: pkg.updateScript.name or null == "gnome-update-script")'
which is mentioned in that update.nix script | 17:25:50 |
| winston | I completely missed 50.2 tho as I was busy with other stuff the last 2 weeks, but yea, I'd be interested in helping out 🫡 | 17:26:50 |
| winston | oh yea and regarding GDM, I think the best way to deal with the gdm-greeter UID warning would be to
1) restore 25.11 UIDs so the warning for people upgrading to 26.05 goes away
2) use systemd-userdbd for nixos-unstable so we can get rid of managing the UIDs altogether | 17:31:28 |
| winston | but there's blockers before we can do 2) as I understand, because we'd need to change how nixbld users are created, too | 17:32:04 |
| Capypara | there's also a new issue on this: https://github.com/NixOS/nixpkgs/issues/529995 which might be related, it reminds me of that issue we had as a blocker for a while in the initial 50 PR | 17:33:05 |
| Capypara | I also really need to get my main PC updated to 26.05. I ran into gdm issues myself a few weeks ago when I tried, but haven't really tried since, I just barely had any time, but really want to help here. Still on 25.11 here. I want to get the backports ready and then probably will try switching to 26.05-staging + backports to get an overview.
My unstable testing VM has been running 50 just fine, even the upgrade went smooth, but I also don't daily drive that one. | 17:36:51 |
| winston | appreciate more people testing and helping with gdm, I've not really been able to learn much more about the issues people are reporting :( | 17:38:39 |
| Capypara | I also hope I'll find some time to look more into it, what we have right now seems very hacky for gdm. but from the brief research I did gdm also seems to have a lot of hardcoded and not well documented preconditions that need to be met for it to really work correctly | 17:40:33 |
 tom | yeah, let's backport.
This was also done for other minor versions in the past. I'm not in front of a computer right now, so feel free to add the label if that does the trick, otherwise I'll do a manual backport as soon as I get to it. :) | 17:43:35 |
| Capypara | * https://github.com/NixOS/nixpkgs/pull/517777
https://github.com/NixOS/nixpkgs/pull/521789
(edit: nevermind, those were before branch-off, misremebered)
https://github.com/NixOS/nixpkgs/pull/530103
| 18:18:51 |
| Capypara | I think btw that bobby is not active on Matrix (or Discourse from what I can see) and Jan was last active in February here | 18:23:31 |
| tom | Just checked out the 50.2 backport, didn't encounter any issues as far as my testing went (see my comments on the backport PR). The remaining open PRs for 50.2 are (all backports):
- https://github.com/NixOS/nixpkgs/pull/532400
- https://github.com/NixOS/nixpkgs/pull/530596
| 22:26:50 |
| tom | * Just checked out the 50.2 backport, didn't encounter any issues as far as my testing went (see my comments on the backport PR). The remaining open PRs for 50.2 are all backports now:
- https://github.com/NixOS/nixpkgs/pull/532400
- https://github.com/NixOS/nixpkgs/pull/530596
| 22:29:29 |
| 18 Jun 2026 |
|  whispers [& it/fae] changed their display name from whispers [& it/fae] to meow meow. | 18:46:27 |
| whispers [& it/fae] changed their display name from meow meow to whispers [& it/fae]. | 19:12:03 |
| 22 Jun 2026 |
|  Yves (she/they) changed their profile picture. | 08:48:16 |
| 23 Jun 2026 |
| Capypara | Does somebody know how we can get those PRs merged without having contact to the GNOME maintainers?
I'm currently checked out on this branch and can't really update my system anymore until it's merged. 😅 | 15:44:08 |
| tom | Maybe try pinging bobby on GitHub if they're not active here | 16:59:33 |
| 25 Jun 2026 |
|  l0b0 joined the room. | 20:21:20 |
| l0b0 | When trying to run a custom NixOS test, gnome-shell is warning about "Missing required core component Settings, expect trouble…". Does anyone know how to rectify that? | 20:21:55 |
| l0b0 | services.gnome.gnome-settings-daemon.enable = true; doesn't help. | 20:24:50 |
| 28 Jun 2026 |
 ElvishJerricco | winston: ok, so yea my hunch was correct. You don't actually need systemd-userdbd running at all. It's just like io.systemd.DynamicUser; GDM just exposes its dynamic users with the userdb protocol, the rest of the system can pick it up with NSS, and systemd-userdbd doesn't play any role at all. | 04:21:50 |
| ElvishJerricco | in the process, I figured out what's going on with unix_chkpwd and with the sleep 1 thing. | 04:23:04 |
| ElvishJerricco | the sleep 1 thing is dumb and the unix_chkpwd one is actually fairly concerning | 04:23:24 |
| ElvishJerricco | the sleep 1 thing is just our resolvconf module explicitly restarting nscd all the time, which means nscd happens to be down when GDM tries to get information about its greeter user (which is a little silly, because it could just get that information from itself, but it's using generic code paths that use NSS and then NSS calls back into GDM via userdb / nss_systemd to get the gdm-greeter user info) | 04:26:18 |
| ElvishJerricco | if nscd is down, no nss_systemd, so no access to userdb accounts via NSS, so GDM fails to see info about gdm-greeter | 04:26:58 |
| ElvishJerricco | The unix_chkpwd thing is a much worse version of that. TIL apparently getent shadow just never uses nscd, which means NixOS fundamentally cannot do NSS module shadow entries without LD_LIBRARY_PATH. Insane. That's why unix_chkpwd needs LD_LIBRARY_PATH; it's the same thing where it needs the NSS module to see the info it wants (and the reason it wants a shadow entry for gdm-greeter is basically just because that's ancillary information picked up along the code paths used in the pam stack), but it won't use nscd, this time because that's just not something that uses nscd, and so it doesn't see nss_systemd's shadow entries from userdb sockets | 04:30:47 |
| ElvishJerricco | We can fix the resolvconf thing easily enough by just disabling that logic when nsncd is used, because that logic is there to deal with the caching that nsncd does not do. | 04:33:05 |
