Bug#1060465: gnome-control-center: crashes loading System panel if gnome-remote-desktop is missing

[Josh Triplett]

On Thu, 11 Jan 2024 18:00:58 -0500 =?UTF-8?Q?Jeremy_B=C3=ADcha?= <jeremy.bicha at canonical.com> wrote:
> On Thu, Jan 11, 2024 at 5:27 PM Arnaud Ferraris <aferraris at debian.org> wrote:
> > Package: gnome-control-center
> > Version: 1:46~alpha-1
> …
> >
> > On a system without gnome-remote-desktop, trying to access the "System" panel
> > (containing the "Users", "Date & Time" panels among other important things)
> > results in a crash with the following message:
> >
> >   GLib-GIO[25147]: ERROR: Settings schema 'org.gnome.desktop.remote-
> > desktop.rdp' is not installed
> >
> > I believe either g-c-c should handle the lack of gnome-remote-desktop more
> > gracefully, or the latter should be promoted to Depends instead of Recommends.
> 
> Ok, I'll promote gnome-remote-desktop to Depends again.
> 
> Josh, I remember you complained about this dependency in
> https://bugs.debian.org/1014879
> 
> GNOME Remote Desktop is going to be more tightly entwined in GNOME 46
> to support the new integrated "headless" mode allowing remote access
> even when the user is not already logged into the host computer. It
> doesn't feel feasible to me to make this dependency optional because
> of the integration in gnome-session, gdm3, gnome-settings-daemon, and
> as a separate page in gnome-control-center.

I appreciate that, and I understand.

I do continue to feel that having gnome-remote-desktop installed and
*enabled* by default feels risky in a similar way to having an installed
sshd on a system that shouldn't be remotely accessed. But, of course, we
cannot have settings pages crashing, either.

For the "headless" mode, would it be possible to confirm that it isn't
enabled by default, that nothing is running if it *isn't* enabled, and
that it requires administrator permissions to enable? (e.g. enabling the
requisite setting should require a policykit prompt.)

For the per-user version, would it be possible to similarly confirm that
it isn't enabled by default, and thus there's *no* remote socket
connection without it first being explicitly enabled, not even a
socket-activateable connection?