On Sun, 2006-01-22 at 13:43 -0500, Tom Lane wrote:
> I object ... this is still trying to enforce tupdesc refcounting on much
> more of the system than I think useful or prudent.
Well, the patch adds management of TupleDesc refcounts solely for the
return value of lookup_rowtype_tupdesc(). If we're going to use
reference counting there at all, I don't see a way do make the
refcounting any less invasive.
-Neil