Ulrich Drepper <drepper(a)cygnus.com&gt; writes: Well the stable egcs 1.0.3 and gcc 2.8.1. releases are responsible for about half the crashes in our bug tracking archive, so I have got into the habit of asking. However this almost for sure a real XEmacs bug. Well. I think that unless you also set x_debug_events to 1 using a debugger it won't help much. I am pretty sure (based on Karl's backtrace), that the crash can be made to go away if we simply chop the XIC xic = FRAME_X_XIC (x_any_window_to_frame (get_device_from_display (event->display), event->window)); line in event-Xt.c into pieces and do nothing if x_any_window_to_frame returns NULL. However as Kyle says, it would be much nicer to actually not get into that situation in the first place. Thus, observing that the window still has a parent that we didn't delete: Can you try out this patch? In particular can you trigger the assert? (which replaces the crash). The patch is of the "untested but obvious type". Index: event-Xt.c =================================================================== RCS file: /usr/CVSroot/XEmacs/xemacs/src/event-Xt.c,v retrieving revision 1.41.2.11 diff -u -u -r1.41.2.11 event-Xt.c --- event-Xt.c 1999/07/30 09:14:01 1.41.2.11 +++ event-Xt.c 1999/08/17 00:20:50 ＠＠ -780,9 +780,16 ＠＠ int bufsiz = sizeof (buffer); Status status; #ifdef XIM_XLIB - XIC xic = FRAME_X_XIC (x_any_window_to_frame + XIC xic; + struct frame *f; + + /* Sometimes we get a keysym for a window that we have deleted, + but its parent is still known too us. */ + f = x_any_window_or_parent_to_frame (get_device_from_display (event->display), - event->window)); + event->window); + assert(f != NULL); + XIC xic = FRAME_X_XIC (f); #endif /* XIM_XLIB */ #endif /* HAVE_XIM */