Why does CVX have to reject convex problems that violate DCP?

In addition to verifying convex problems, does DCP have other functions in the subsequent solution?

For example:
min y=x1*x2

CVX does what it does. Feel free to write your own tool which does things better.

Anyhow, this problem can be entered “as is” using gp mode. http://cvxr.com/cvx/doc/gp.html