You would define such an integral by choosing a parameterization for the path. Then it reduces to a regular integral in terms of the parameter. I don't know what "zeta" has to do with it but all the limits are hidden in the definition of the regular integral.
I'm reviewing Meirovitch's "Methods of Analytical Dynamics," and I don't understand the commutation of the derivative from r to dr:
$$
\mathbf{F} \cdot d\mathbf{r} = m \ddot{\mathbf{r}} \cdot d\mathbf{r} = m\mathbf{\dot{r}} \cdot d\mathbf{\dot{r}}
$$