No opinion on the change itself but just for reference see #12516 for previous discussion. Also note that currently lightdm test is not part of the tested release (which it probably should if it were to become default dm).

#12690 better version

Another note - lightdm can't be convinced to log to journal which some find valuable (if anything for consistency with other DMs).

I wonder if we might want to remove the slim package as well as I think that keeping unmaintained stuff in our package set just unnecessary extra-work for us...

I don't think we should remove it, as long as it keeps working. I guess some people really prefer it to lightdm and I wouldn't want to take the choice away from them.

The thing is: the reason why I asked is because I'm not a big fan of keeping dead software in a maintained distro.
For instance we deprecated the usage of the EOLed Synaptics Touchpad driver (although it's still working fine, see 9d2777a).

In the end Nix provides a rich composition API, so if people really want to use dead software they can still build an overlay :-)

We shouldn't go at once from having it as default to removing it completely.

Just adding an argument against SDDM in case that was still a candidate. SDDM is not efficient on server hardware: sddm/sddm#323.

I rebased it on the current master and moved the comment to the 18.09 release notes.

Also note that currently lightdm test is not part of the tested release (which it probably should if it were to become default dm).

I also think it should become part of the tested release, otherwise when (not if!) it breaks, it won't be caught.

AFAICT this should be the only blocker for merging, objections?

lightdm is now tested via the gnome3 test (see: #41742).
Is this enough or should we write a seperate test?

Ah, I was going from the commented lightdm test here. I hadn't realized lightdm was transitively tested in the gnome3 tests.

Can somebody explain what was the problem with slim exactly? I've read #12690 but still don't have an idea. What is the problem with "unmaintained"? Emacs and TeX have packages everybody uses that didn't change in the last 20 years. Are there any open bugs that prevent slim from being the default?

Anyway, immediately after rebooting into a system with auto display manager and this patchset applied I get two issues:

• lightdm runs two processes three threads each just to make auto login work (slim had 1 of each), both stay running after .xsession script gets executed. That's a bit too much. I run NixOS with X11 on systems with very low RAM, and this is really bad.

• Normally I exec xmonad at the end of .xsession. With lightdm exiting xmonad brings the system into an unusable state, lightdm just keeps running, showing its background window as if nothing happened, and you have to login in another tty and restart display-manager manually to get anything done.

TBH, I'd just revert this and followups. But if you really want this clearly-non-light-dm thing be the default can we at least get slim for auto at the very least?

I personally dislike e.g. that the UX is rather bad for slim, which doesn't seem suitable for a default. If you don't know slim, you will be hard-pressed to switch sessions, reboot, etc.

can we at least get slim for auto at the very least?

My preferred solution for autologin is pam-autologin, no display manager at all, just use startx as normal user. Unfortunately not yet supported in NixOS.

@vcunat I see, indeed, one does need to read the man. Fine, this might be better as the default interactive login, but I don't want this for auto. I made #46396 which reverts the auto part of this and the workarounds by @xeji that followed.

@xeji

pam-autologin

Hm, right, that would be even better for auto, I'll investigate what needs to be done later. slim is good enough for now, though.

While being a sane default it looks like gnome3 setting gdm as the default display manager can cause problems if the user sets another display manager in their configuration: #47945