Re: BUG #5748: Invalid oidvector data during binary recv

BTW ... it strikes me there's another inconsistency between oidvectorin
and oidvectorrecv, namely that the former enforces a maximum of
FUNC_MAX_ARGS elements whereas the latter has no such limit.  Should we
do something about that, and if so what?  It wouldn't be too hard to get
rid of the former's maximum, but I'm worried about whether any code is
dependent on an assumption that oidvectors can't get toasted.  The
type's marked untoastable in pg_type, so maybe enforcing the same limit
in the latter is a safer idea.

            regards, tom lane