Simple but crucial bug fix to gdb

[Jim Blandy]

Fernando Nasser <fnasser@cygnus.com> writes:
> I cannot approve the patch because I am not the maintainer of the
> stabs reader, but (for the reasons you've explained above) this
> seems to fall clearly into the obvious fix rule.

This is *not* an obvious fix.

The stabs reader has seen a symbol definition, before it has any idea
what compilation unit that symbol belongs to.  Are you *sure* it's
okay to just ignore that symbol definition?  Won't the psymtab's start
and end addresses get set wrong?