elogind plan

[Mark Hindley]

On Sun, Dec 02, 2018 at 11:42:26PM +0100, Adam Borowski wrote:
> So... what's the plan with elogind?
> 
> We need some sort of transition, _plus_ testing.  And Buster's freeze is
> afoot.

Yes, absolutely!

I prepared a package of upstream 239.2 last week which I posted to the list and
Ian is reviewing. Upstream 239.3 is imminent.

> There's of course that policykit-1 issue; do we have a plan?

We need the package of 239.2-1+debian1 uploaded before we can ask the
policykit-1 devs to include libelogind in their builds

> Besides that, with libpam-elogind-compat installed xfce works for me, and
> since today so does mate.

Good news.
> But, that compat package is not going to hit buster (nor even sid), so we'd
> need to fix dependencies.  I think smcv's suggestion from January would work
> -- we can replace depends on "libpam-systemd" with "default-logind | logind",
> with elogind declaring Provides: logind (and on some derivatives, also
> default-logind).  That would handle packages needing this part of the
> functionality.  Does this sound ok to you?  If so, please say so -- I still
> have a set of patches implementing this, thus filing bugs would be far less
> work for me.  And if you have a different but similar solution, it's likely a
> sed job on the patches would be enough.

I don't see any problems with that approach, but we will need to get the virtual
package names formally agreed.

> There's also the matter of upgrades: there's a need to be able to switch
> from systemd to sysvinit+elogind, which you can't currently do (systemd's
> prerm refuses to remove while it's pid 1).

The impression I had received was that it might be difficult to persuade
systemd's maintainers to remove that restriction. Do you know if anyone has
actually tried asking them?

Mark