[7266] in testers
Re: Focus getting stuck
daemon@ATHENA.MIT.EDU (Greg Hudson)
Thu Jul 14 17:04:10 2005
Date: Thu, 14 Jul 2005 17:03:03 -0400
Message-Id: <200507142103.j6EL33pM019881@egyptian-gods.mit.edu>
From: Greg Hudson <ghudson@MIT.EDU>
To: Greg Hudson <ghudson@MIT.EDU>
CC: testers@MIT.EDU
In-reply-to: <1117745727.3570.158.camel@egyptian-gods.mit.edu>
Laura finally ran into this problem at a time when I could try to
debug it remotely. Unfortunately, I could not produce a gdb binary
for Solaris which could read metacity's symbol table (the gnu locker
versions both dumped core before attaching the process; a binary built
from the most recently released gdb source code could read the symbol
table for the libraries metacity was linked against, but not the table
for metacity itself, so I couldn't print out any global variable
values or anything). Here's what I did learn, from a more in-depth
conversation with Laura:
1. The observed behavior is as if some client had the mouse pointer
grabbed, but not the keyboard or the server. Keystrokes would go
to the application which currently had keyboard focus, but mouse
movement or clicks would have no effect. Alt-tab and alt-esc did
not work for changing the keyboard focus, but I think that's due
to http://bugzilla.gnome.org/show_bug.cgi?id=151500. Also,
xscreensaver complained to the console:
xscreensaver: couldn't grab pointer! (AlreadyGrabbed)
2. In this case, the cursor shape suggested maybe zwgc had the
pointer grabbed, but mouse clicks sent to zwgc windows had no
apparent effect. According to Laura, in the past the cursor has
been the zwgc cursor, the eterm cursor, and the emacs cursor.
3. According to Laura, killing off the program which matches the
cursor shape resolves the problem.
My two best theories are: (a) some rogue application is grabbing the
mouse pointer without changing the cursor shape, and (b) metacity is
getting confused into thinking something has the pointer grabbed when
nothing does. Theory (a) is not really consistent with (3), and
theory (b) is not really consistent with mouse clicks having no effect
anywhere (as far as I know, anyway), and I also couldn't find any
bugzilla reports to substantiate it.
So far, it seems like this problem only happens to Laura.
