gdm_session_worker_process_pam_message() contains this code:

                         *response_text = strndup (user_answer,
PAM_MAX_RESP_SIZE - 1);
                        (*response_text)[PAM_MAX_RESP_SIZE - 1] = '\0';

If the string pointed to by user_answer is shorter than PAM_MAX_RESP_SIZE - 1
(which will generally be the case), the second line clobbers unrelated memory.
On this powerpc laptop, that causes gdm-session-worker to crash while verifying
the password, leaving me unable to log into any user session.

strndup() already ensures that the resulting string is 0-terminated anyway, so
this commit just removes the second line.

it could be you want one of the other methods to be strictly required.
This commit just adds the config key for checking that.

Just a small leak fix.

slave_name_id is supposed to track the bus name watch of the slave, but
it was never getting set.

This commit sets it, and makes sure the watch is removed in unmanage.

https://bugzilla.gnome.org/show_bug.cgi?id=698370

Right now we do:

g_signal_connect (dbus_server, "...", callback, self);

in a few places.  These signal connections should go away when the dbus
server is unreffed in dispose, but to be on the safe side we should
use g_signal_connect_object anyway.  (See for instance the leak
fix in the previous commit).

Without this fix, GDM refuses to start up saying that pam_systemd.so
isn't initialised.

https://bugzilla.gnome.org/show_bug.cgi?id=696577