Robert Haas <rhaas(at)postgresql(dot)org> writes:
> Avoid coercing a whole-row variable that is already coerced.
This logic seems very strange and more than likely buggy: the
added coerced_var flag feels like the sort of action at a distance
that is likely to have unforeseen side effects.
I'm not entirely sure what the issue is here, but if you're concerned
about not applying two ConvertRowtypeExprs in a row, why not have the
upper one just strip off the lower one? We handle, eg, nested
RelabelTypes that way.
regards, tom lane