meson: Fix systemd_dep assertions to match assertion message

See merge request GNOME/gdm!247

udev rules: Add check for nvidia S0ix suspend

See merge request GNOME/gdm!248

Dan Nicholson authored Mar 13, 2024 and Ray Strode committed Mar 23, 2024

During the PAM conversation, `pam_gnome_keyring` will try to unlock the
login keyring. If the keyring created by gnome-initial-setup hasn't
changed ownership, then `gnome-initial-setup-copy-worker` can't move it
in place and `gnome-keyring-daemon` will try to create a new login
keyring. This either means the keyring created in g-i-s is lost or g-k-d
gets confused about a keyring that shows up after initializing but
before it's received the unlock/create request.

Similarly, the systemd user session is started by `pam_systemd` and will
trigger any number of services that might consume the config files g-i-s
has created.

To ensure that files are ready before any consumers have started, chown
them right after the user credentials have been established but before
the PAM session has been started.

Helps: https://gitlab.gnome.org/GNOME/gnome-keyring/-/issues/137

manager: Move g-i-s chown to when PAM conversation started

See merge request GNOME/gdm!246

Marco Trevisan authored Mar 26, 2024 and Ray Strode committed Apr 04, 2024

This was already handled but avoids a static analyzer warning.

pam-extension: Avoid gnu-variable-sized-type-not-at-end warning

See merge request GNOME/gdm!250

data: Re-enable wayland on hybrid graphics laptops with NVIDIA driver

See merge request GNOME/gdm!237

commit 4291dde6 was a little
too overzealous in its trimming, creating bugs for non-nvidia hardware.

This commit brings back an important part of the gpu logic.

data: Revert part of wayland hybrid graphics change

See merge request GNOME/gdm!254

The remote display factory has a counter of how many remote displays
are. With that counter when there's at least one, block suspend, when
there's none unblock suspend.

This is necessary because other sessions might want to suspend without
being aware that there might be remote sessions being used. On suspend
the RDP client would disconnect and that makes the remote session die.
In the future the sessions would not die on disconnecting, but still is
undesirable the RDP client disconnecting.

remote-display-factory: Block suspend when any remote display exists

See merge request GNOME/gdm!258

Joan Torres authored Mar 22, 2024 and Ray Strode committed Jun 07, 2024

Now two headless sessions are compatible.
On reauthentication, the login display will get the property
of its reauth_caller_pid.
Now when a session is reauthenticated, on "Reauthenticated" signal it
is given the pid_of_caller property. This way it can be known from which
login display it has been reauthenticated, when multiple login displays
attempted to reauth the same user.
On succesfull reauthentication, get the login display using that
property. If that display is remote, set the "remote-id" of the login
display to the user display, this makes the g-r-d daemon-system, handover
an RDP client to that reauthenticated session.

GitLab doesn't seem to be able to mark a manually merged MR as merged
in the UI.

This commit adds an empty commit in the hopes that will wake it up.

Support persistent headless sessions

See merge request GNOME/gdm!249

- Support for gnome-remote-desktop persistent sessions
- Misc changes
- Translation updates