Thank you for your fast reply.

But unfortunately this doesn't help me to solve my "real" problem.

The aforementioned procedures f and g are just simplyfied ones in order to
point out the problem with fsolve and procedures. Sadly I simplified the
procedure f too much so that your suggested method is able to solve this
specific problem.

Actually the procedure f is of the same type as the procedure g. So  the
procedure cannot return  a symolic expression in (x,y).

Is there a possibility, that maple solves a set of equations in the form

f_i(x_1,x_2,..x_n)=0  ; i=1,..,n

just by handling the functions f_i as "black boxes" and make use of, say
e.g. a Newton-Raphson method ?

Thanks,
iznogoud