possible Xlib problem?

[Ken Raeburn]

So, I've been having these core files left around sometimes by
Emacs.... I've just reported some of the specifics to the pretest-bug
list earlier tonight (which stuff should I send there and which
here?).

The main thing seems to be, when I'm running Gnus and it's put up its
somewhat colorful "*Group*" buffer with text in at least three colors,
and then the sole X connection goes away (timeout, or "~." in ssh),
Emacs crashes.  Near as I can figure, it's trying to free up some
XIM-related data structures when it runs across a function pointer
that's NULL.  The function pointer is being set to NULL by XFreeColors
because that routine somehow thinks that the location of the function
pointer is part of a request buffer it can store things into to ship
them off to the X server.  (We'll ignore for the moment the logic in
trying to assemble messages to send to a server you already know you
can't talk to any more.)  XFreeColors is being called from the Xt
library.  So it may be that Xt, multiple colors, and XIM are all
relevant.

Oh, and I'm using XFree86 3.3.5 as shipped with Red Hat 6.1.

Granted, it's making the core file after I'm no longer able to use
Emacs anyways, since it was running only on the one display, but if it
can't clean up a single display, how can we expect the
multiple-display support to work?

Do other people see this problem too, of leaving core files when the X
connection is dropped on a colorful buffer?

Ken