I've moved house and gotten a pmax set up in my new place
(No net connectivity yet, though) and found time to fix
the emacs unexelf.c.  The previous patches stupidly
(my fault, so I can say it) left out relocation of the .mdebug
ECOFF-debug-symbols-in-ELF section. 

Putting that back lets me at least get a stack traceback from emacs
when it dumps core.

Re-building a few emacs source files with cc -g3 shows that the
coredump is happening when emacs tries to create a widget for its menu
bar from x_fns.c, which calls XtCreateWidget() which then calls
_XtCreateWidget, which dies inside another Xt widget-initialization
function. The address gdb claims is the faulting address is somewhere
inside the .dynstr or .hash section of the undumped binary.

That sounds familiar. I wonder if this could be a recurrence of
another old bug with the mips-elf relocation.  Michael, does that ring
any bells to you (or anyone else)?

PS: I'll post diffs (or send a pr to the pkg system) and a stacktrace
just as soon as I figure out how to get it off my machine at home.