Alvaro Herrera <alvherre(at)dcc(dot)uchile(dot)cl> writes:
>> Fix bug I introduced in recent rewrite of NUMERIC code: numeric to
>> integer conversions gave the wrong answer for values with stripped
>> trailing zeroes, such as 10000000.
> No regression test?
Don't see the point --- it's a one-off fix. (Wrong loop termination
condition in one place.) I put in regression tests for things that
I think might break again in future, or if I can see a way to test for
a whole class of errors rather than one single routine.
regards, tom lane