I wrote:
> Andrew Gierth <andrew@tao11.riddles.org.uk> writes:
>> I can see two possible fixes: one to correct the assumptions in the
>> macros, the other to check for NaN before calling init_var_from_num in
>> numeric_send (all the other functions seem to do this check explicitly).
>> Which would be preferable?
> I'm inclined to think special-casing NaN in numeric_send is the thing to
> do, since it is that way everywhere else. If we could push down all the
> special casing into init_var_from_num then that would probably be better,
> but in most cases that looks unattractive.
After taking a closer look I realized that you were right with your first
alternative: the field access macros are just plain wrong, and we should
fix 'em. Done now.
regards, tom lane