James David Smith <james(dot)david(dot)smith(at)gmail(dot)com> writes:
> Yes, I thought the same. Glad to know that I might be along the right
> lines. I've started taking parameters out one-by-one to try and find
> the culprit. Yes, there are a variety of numbers stored as NUMERIC in
> various places. For example:
AFAIR, there is no such thing as an underflow error in the type-NUMERIC
code --- it will happily round to zero instead. You should be looking for
something involving float4 or float8 (aka real/double precision).
regards, tom lane