integral bug?



Hello,

I was playing with a problem on math.stackexchange.com, and I found this in
Maxima:

> integrate(cos(1/3*exp(-%i*theta)),theta,0,2*%pi);
> 0

WolframAlpha gives me 2*%pi, and it seems obvious that it can't be null
anyway. Any hint to have the computation done right with Maxima ?