I'm not sure how much it matters, but the netscape 2.0 beta 1 binary - i.e.,
the one that works fine - has the same exact code at the setjmp() point in
it as the 2.0 beta 2 binary does (4.1.3_U1 binary, that is).  I'm not sure
if this means that the action is happening at the longjmp() end, or if there's
something weird that just happens to mess up the register windows at that
particular juncture.  It may take a while to find where the longjmp() is ...

	- Greg