fzn-gecode, mod, wrong solutions

[Mats Carlsson]

The following model is unsat, but fzn-gecode finds several wrong solutions:

var -5 .. 5: A;
var -5 .. 5: B;
constraint A mod B=B;
solve :: int_search([A,B], input_order, indomain_median) satisfy;