This is related to the thread about tan(pi/2) unevaluated in the
result of some integral.
What should maxima do with tan(%pi/2)? Currently, it does nothing.
No error, no infinity. What should it do?
Compare this with csc(0), which throws an error about 0 not being in
the domain.
When that is answered, I think we can fix the original integral
"correctly".
Ray