Typically most tools in MATLAB, or any language that works in floating point arithmetic, must allow a small amount of slop, a tolerance constraint, wherein the constraint may actually be exceeded by a small amount that is less than the tolerance. Much of the time, that excess is tiny, on the order of floating point trash.
A problem arises when your equations are ill-conditioned. Then the slop becomes magnified, and difficult to maintain the desired limits.
However, does your use of the word hard bounds mean that you wish to do more than that? Do you want to see strict inequalities, thus of the form C(X)>0, as opposed to C(X)>=0?
As I said above, floating point trash can get into all numerical computations done with floating point arithmetic. So asking for s strict inequality is quite difficult to do. Even asking for a solution that strictly obeys a nonlinear inequality of the form C(X)>=0, with no tolerance allowed for exceeding the bound is difficult.
