This bug may be related to #908: 

https://bugzilla.icculus.org/show_bug.cgi?id=908


Symptoms: 
---------

No program responds to keyboard input, neither the directly connected kbd, the vnc kbd, nor the auxilliary kbd. 

Steps to replicate:
-------------------

Intermittent Bug. (sigh)

It seems to be associated with periods of inactivity, but that could just be related to how the bug is discovered. 

We'll be putting debug statements into keyboard_grab, and callers, to see if we can get more information on what part of OB grabbed the kbd last.

>>>>Action item: if anyone knows what additional info to collect, let us know!

Frequency of occurence:
-----------------------

Of 400 stations using OB, this bug is reported 6-8 times a day. 

Additional Info:
----------------

* Eliminating all other processes does not restore the keyboard, suggesting that it is OB that has grabbed.

* OB does not respond to any of its mapped keystrokes.

* The mouse continues to function in terms of repositioning windows and so on, oddly, the windows do NOT appear as focused.

* Opening an xterm (via ssh) and clicking on the client space of the window terminates our ability to reposition that window with the mouse.



* Sending SIGUSR2 (reconfigure) does not restore the kbd.

* Sending SIGUSR1 (reload) DOES restore the kbd.

*xsession-errors reports this, ad infinitem, indicating that something has grabbed and frozen the kbd:

xscreensaver: 00:02:51: unable to grab keyboard or mouse!  Blanking aborted.
xscreensaver: 00:03:55: couldn't grab keyboard!  (GrabFrozen)
xscreensaver: 00:03:59: couldn't grab keyboard!  (GrabFrozen)
xscreensaver: 00:04:03: couldn't grab pointer!  (AlreadyGrabbed)


Workarounds being considered:
------------------------------

Well, we'd like to authentically identify the source of the bug and kill it proper, but failing that, we're considering the yucky hack of externally detecting the condition and modify a sighandler to release all its grabs.