From: Howard Brazee on
On Thu, 24 Jul 2008 20:47:52 GMT, "William M. Klein"
<wmklein(a)nospam.netcom.com> wrote:

>It is with noting that NUMVAL is NOT required to produce "exact" results, i.e.
>it MAY (in a Standard conforming implementation) return a floating-point value
>which MAY be different than what is expected. However, I would consider what
>you report to be a "bug". Have you contacted your vendor?

First test with a very simple program the way a couple of us have in
response to the problem. Then add complexity until you get to the
spot that repeats the problem (or you discover that it's a programmer
bug - did you double check USAGE?).