Re: Crash of GNU Emacs 21.3.50 on Linux

[Jan D.]

> Symptoms:
>
> Sometimes GNU Emacs from CVS crash after long run - one-two days.
> Usually this happens when i use gnus on large groups, or rebuild
> gnus's spam-stat database.
>
> here is backtrace of GNU Emacs on crash:
...
> #1  0x0811ce81 in fatal_error_signal (sig=5634) at emacs.c:395
> #2  <signal handler called>
> #3  0x4ba1cd87 in XFreeGC () from /usr/X11R6/lib/libX11.so.6
> #4  0x4be443a2 in gdk_x11_font_get_name () from 
> /usr/lib/libgdk-x11-2.0.so.0
> #5  0x476e68dc in g_object_unref () from /usr/lib/libgobject-2.0.so.0
> #6  0x4be2cafb in gdk_screen_get_type () from 
> /usr/lib/libgdk-x11-2.0.so.0
> #7  0x4be4e9d4 in gdk_screen_set_default_colormap () from 
> /usr/lib/libgdk-x11-2.0.so.0
> #8  0x476e6a0f in g_object_run_dispose () from 
> /usr/lib/libgobject-2.0.so.0
> #9  0x4be2cbc7 in gdk_screen_get_type () from 
> /usr/lib/libgdk-x11-2.0.so.0
> #10 0x4be37608 in gdk_x11_display_ungrab () from 
> /usr/lib/libgdk-x11-2.0.so.0
> #11 0x476e6a0f in g_object_run_dispose () from 
> /usr/lib/libgobject-2.0.so.0
> #12 0x4be18224 in gdk_display_close () from 
> /usr/lib/libgdk-x11-2.0.so.0
> #13 0x080fde43 in x_connection_closed (dpy=0x8671620,
>     error_message=0xbfffc770 "X protocol error: BadWindow (invalid 
> Window parameter) on protocol request 18") at xterm.c:7767

GTK has a long standing bug, it can't close displays without crashing.
See http://bugzilla.gnome.org/show_bug.cgi?id=85715.
The real error is the error message, BadWindow.  This is much harder
to track down as the request with the bad window has already happened,
the crash is a result of processing the answer from the X server.  If 
you
can reproduce the bug and run Emacs synchronized (see lisp function
x-synchronize) that might give something.  This is not an option if it 
takes
several hours, synchronized is much slower.

There might be the case that the window that did the selection request
has been deleted.  The stack looks corrupted (event is a pointer, it 
should
not be 0x1), but that can be due to the GTK bug.


> #25 0x08124cf2 in read_avail_input (expected=1) at keyboard.c:6675
> #26 0x08124eb2 in handle_async_input () at keyboard.c:6817
> #27 0x081082eb in x_handle_selection_request (event=0x1) at 
> xselect.c:749
> #28 0x0812527b in swallow_events (do_display=0) at keyboard.c:4216

I guess we could check for BadWindow in the error handler and just 
ignore
it, unless we get several in a row for the same window.  I'll see if 
that
is possible.

        Jan D.