M2qn1 is a Fortran subroutine that minimizes a smooth function subject to bound constraints. Rooted in Va13a of the Harwell Subroutine Library, it implements a quasi-Newton algorithm, using the BFGS update formula and Wolfe's line-search. To deal with the bound constraints, it implements an active set method, related to the Rosen-Goldfarb technique. Moreover, a facility is granted to make a post-optimal analysis. N2qn1 is the double precision version of M2qn1.