In Bug 2953369
<http://sourceforge.net/tracker/index.php?func=detail&aid=2953369&group_id=4933&atid=104933>;,
maxima evaluates the integral incorrectly to zero.
The problem is due to the fact that maxima cannot determine if the roots
of b*yy^2-2*a*yy+b are in the unit circle are not. In fact maxima
thinks they are all outside the unit circle. Because of that, maxima
thinks the integral is zero because maxima is evaluating the integral
via contour integration around the unit circle.
As a work around, I was thinking of changing the test to use asksign.
At least then the user can supply the correct sign. However, asksign
(and $asksign) don't ask. What controls that?
Ray