writing fortran mex file

I have found some difficulties to write fortran mexfile and this small example shows how I have considered the problem of exchanging data between matlab and fortran and "vice versa".
My problem was to use some optimzation fortran codes operating upon function written within matlab and with almost the facilities offered by the optimization toolbox functions, in particular the varargin argument for user special function parameters. I also consider the question of fortran write, because the mexprintf does not work "in real time" R2006, R2007 or R2008.
The example is a simple problem. It consists in finding the zero of a nonlinear function by dichotomy. The submission contains a a matlab test, the matlab code fzdm.m (zero finder) and its fortran mexfile implementation fzd.f