RE: lsblibchk update

The line of inquiry I mentioned the other day did prove
fruitful.  The two classes in question are odd because
they don't have an exported vtable in libstdc++.  This
may be an oversight, or it may be intentional.  They do
have a vtable, but as it's not exported we can't reach
it by name.

However, we still have data for these two classes in the
database and generated libchk code (and, sadly, the spec
as well) as if they we complete, and although I don't have
the precise description of what goes wrong, it looks like
libchk ended up following an invalid chain of links
eventually leading to the incorrect results we've been
chasing for a while.

What I'm not sure of is what to do to fix this.
Neither of the apparent options is particularly
appealing.  I'll try to send an update later, and
then we'll need to make some sort of decision.