| 6 May 2026 |
 hexa | yeah, that has me a bit worried | 13:17:09 |
 vcunat | I think it looked similar in the past few releases, at least at a very superficial glance. | 13:20:06 |
| vcunat | * I think it looked similar during the past few releases, at least at a very superficial glance. | 13:20:13 |
| 7 May 2026 |
 winston | RE the gnome 50 PR, I worked on it early on in March. now that it's merged, I guess we have a week-ish to settle on how we'll make the gdm nixos module work?
IIRC the problem was just that gdm wasn't running with the correct user, the rest of the PR looked good in a VM | 14:04:27 |
 emily | it was something to do with systemd-userdbd, right? maybe https://matrix.to/#/!SsuQHVzf59iBgk6YIm:0upti.me?via=nixos.org&via=matrix.org&via=tchncs.de can be of assistance (though I agree with https://matrix.to/#/!CTCrFzsBPYmDLmrja4:0upti.me/$2U9w2iyhK1hrYALIqmpThd3tsjENevltL9-NC5SPnjk?via=nixos.org&via=matrix.org&via=tchncs.de that the seeming partial collapse of GNOME maintenance is concerning going forward) | 14:11:46 |
| vcunat | I suppose it's a gamble. The time to merge to master would normally be about two weeks at this point. | 14:12:47 |
| winston | yeah, and for 51 there'll be even more of a dependency on it
https://blogs.gnome.org/adrianvovk/2025/06/10/gnome-systemd-dependencies/ | 14:13:33 |
| winston | there's already an issue for it https://github.com/NixOS/nixpkgs/issues/458058
it wasn't addressed during 49 as well, it was just worked around with fixed user ids | 14:13:54 |
| emily | I know ElvishJerricco has thoughts on userdbd for NixOS | 14:14:24 |
| emily | though AIUI userdbd is meant to be able to pick up records from /etc/passwd still? | 14:14:41 |
| emily | so I don't think it should require full integration? | 14:14:45 |
| winston | thanks for the nixos systemd channel, I'll see if I can get it to work
idk much about the whole stack atm sadly | 14:15:05 |
| emily | I assume it's been tried to just enable the NixOS option for userdbd in the GNOME module and that didn't work? | 14:15:21 |
| winston | I'd be interested in helping with GNOME I just kinda burned out on that PR after the LLM stuff | 14:15:27 |
| emily | "Tested with systemd 257.9 and #205121 but it always failed on various PAM issues." ok :) | 14:15:36 |
| emily | systemd room should be able to help | 14:15:45 |
| emily | (yeah I would be more worried about GNOME maintenance if it did not seem like potentially a one-off incident from how that PR started) | 14:16:08 |
| winston | yeah I did that too, sadly didn't work without further tweaks IIRC, its been a month+ tho | 14:16:12 |
| winston | anyway I'll give the new staging a build, and I'll try helping with getting GNOME over the line | 14:16:57 |
 ElvishJerricco | IIRC the idea is that there's an NSS module that represents userdb records as normal NSS things, and there's functionality in systemd-userdbd to represent NSS things as userdb records | 14:20:43 |
| ElvishJerricco | not terribly surprised if there's PAM confusion, but I know nothing about it. | 14:21:13 |
| ElvishJerricco | my thoughts on userdb are just generally "it sounds like a nice way to kill of /etc/passwd and do declarative JSONs for users instead of imperative file manipulation one day" | 14:21:49 |
| ElvishJerricco | have done zero experimenting | 14:21:57 |
| ElvishJerricco | * my thoughts on userdb are just generally "it sounds like a nice way to kill off /etc/passwd and do declarative JSONs for users instead of imperative file manipulation one day" | 14:23:31 |
| ElvishJerricco | I know GDM wants to use userdb so it can have N GDM users instead of 1 GDM user, but have no idea how that works. Would not be surprised if either it 1) doesn't need systemd-userdbd specifically at all and has its own userdb-JSON-record-serving-daemon (userdb is just a directory of sockets and can be backed by arbitrarily many such services), or 2) creates JSON record files that it expects systemd-userdbd specifically to handle. | 14:26:33 |
| ElvishJerricco | in either case I'm not sure why there would be PAM issues, since AFAIK PAM should just be happy using NSS stuff. Maybe something is starting before nscd.service, and historically fell back on /etc/passwd if that wasn't up? | 14:28:18 |
| ElvishJerricco | (hey look, potentially another reason that nscd makes me want to build volcanoes instead of computers) | 14:29:27 |
| emily | https://blogs.gnome.org/adrianvovk/2025/06/10/gnome-systemd-dependencies/ explains the userdbd dependency | 14:46:47 |
| emily | it needs io.systemd.Multiplexer, getpwent workign for userdbd users, and the userdb API working for /etc/passwd users | 14:47:35 |
| winston | worst case we could revert https://gitlab.gnome.org/GNOME/gdm/-/commit/c3747f1c14b777e1a42b9e5d903214c39e2c0462 to get hardcoded users to work again, at least for 50, but I'd like to avoid kicking the can further down the road if possible | 14:51:15 |
