Created attachment 752061 [details]
Fix or workaround calling a function via a pointer, that is NULL

Description of problem:
When a Gnome Session (whether fallback or not does not matter) is running in a VNC server, Locking the screen does not work and the window manager becomes somewhat dysfunctional

Version-Release number of selected component (if applicable):
3.4.1-1

How reproducible:
Start a vnc server, connect with vncviewer, try to lock the screen
or let it lock automatically

Steps to Reproduce:
1. prompt# vncserver -extension RANDR
2. prompt# vncviewer :1
   (or whatever session number is output by vncserver)
3. Within the session select in the top menu under the user's name on the right "Lock screen"


Actual results:
Display starts to darken, but then the normal desktop is visible again and the window manager seems to be canceled somehow. In the top bar a message appears "gnome-screensaver crashed ..."


Expected results:
Screen is locked and can be unlocked again entering the password

Additional info:
The gnome-screensaver gets SIGSEGV:
In src/gs-fade.c in 2 lines a function is called via a pointer like this:
fade->priv->screen_priv[i].fade_finish(fade, i);
It seems, when running in VNC, there's no respective capability in the server, so the entire structure fade->priv->screen_priv[i] is zero, so is
fade_finish.
The attached patch adds a check: if fade->priv->screen_priv[i].fade_finish
is NULL, the function is not called.
Frankly i don't know, if this is really a fix or only a workaround. The
maintainer must judge.
However. One more time adding a fancy but superfluous feature introduced a new annoying bug.