At the moment, cashocs shows the norm of the unprojected gradient and uses this as convergence criterion when mesh quality constraints are used. Better would be to use the norm of the projected shape gradient.
As the mesh quality constraints are a semi-discrete, semi-continuous ansatz, either the 2-norm (discrete) of the shape gradient, or the elasticity norm of the shape gradient (continuous) could be used. I tend to favor the second approach at the moment.
At the moment, cashocs shows the norm of the unprojected gradient and uses this as convergence criterion when mesh quality constraints are used. Better would be to use the norm of the projected shape gradient.
As the mesh quality constraints are a semi-discrete, semi-continuous ansatz, either the 2-norm (discrete) of the shape gradient, or the elasticity norm of the shape gradient (continuous) could be used. I tend to favor the second approach at the moment.