integrate returns undefined



John Lapeyre wrote:
> Its not a question of CAS design. There is no need to go
> into the details of how Maxima computes the answer (although
> that may be interesting). Maxima is giving the only correct
> answer. The user is saying:
>  
>   For positive integers m and n what is the definite integral
>    integral_sign cos(m*x)*cos(n*x) dx  from 0 to 2 pi   ?
> 
> The answer to that question is 
>       0   for m != n
>       and the integral is undefined for m=n.

Why is it undefined for m=n?  It seems well-defined to me.

Ray