What does fmincon rel max. constraint violation mean ? Why is that compared to TolCon and not "output.constrviolation" ?

Hello,

below is the output of fmincon (2011b, SQP)

Optimization terminated.

                                                          Norm of First-order
 Iter F-count            f(x) Feasibility  Steplength        step  optimality
    0      22  -6.805775e+004  2.649e+002                          2.728e+003
    1      44  -5.056068e+004  1.868e+002  1.000e+000  2.676e+001  8.910e+005
    2      45  -5.056068e+004  1.868e+002  7.000e-001  1.767e-008  8.910e+005

Optimization stopped because the relative changes in all elements of x are
less than options.TolX = 1.000000e-004, but the relative maximum constraint
violation, 7.052354e-001, exceeds options.TolCon = 1.000000e-001.

Optimization Metric                                         Options
max(abs(delta_x./x)) =  1.15e-009                    TolX =  1e-004 (selected)
relative max(constraint violation) =  7.05e-001    TolCon =  1e-001 (selected)

My Question is: At the last iteration the feasibility (constraint violation) is listed as 1.868e+002 which equals to max(f_nonlinc(x)).

At the very end of the optimization run output.constrviolation is also 1.868e+002 BUT the value which is compared to TolCon is the relative max(constraint violation) which is 7.05e-001 ?

What ist the difference between "relative max(constraint violation)" and output.constrviolation, how is the rel. violation computed ?

Why is matlab/fmincon comparing TolCon with the relative violation ? That doesn´t fit to the documentation.

[Vincent]

This is a good question, why has nobody ever answered this? Can someone please have a look at this?