[Scip] scip reports optimal solution along with violated constraints

[Tobias Achterberg]

This is strange. I guess that something wrong must have happened in your model building.

Is there anything strange in the SCIP output? For example, what does presolving say about 
the size of the model? What numbers appear in the node log, in particular the "rows" and 
"cols" column? Is the model solved directly at the root node?

Which SCIP version are you using?


Tobias


Martin Bergner wrote:
> Hello all,
>
> I am again building a scip problem in C++ and try to solve it. The
> problem gets created normally with SCIPcreateProp,
> SCIPcreate/addCons/Var. I include the default plugins and I only use
> linear constraints. When I try so solve the problem via SCIPsolve, the
> status (SCIPgetStatus) is SCIP_STATUS_OPTIMAL. However, the following
> output appears:
>
> [linear]<c2>: +1<x9>[B] +1<x10>[B] +1<x11>[B] +1<x12>[B] +1<x13>[B]
> +1<x14>[B] +1<x15>[B] == 2;
> violation: left hand side is violated by 2
> best solution is not feasible in original problem
>
> The solution as well as the objective function is not correct.
>
> Now to the best part: I can exported my problems via
> SCIPprintOrigProblem(.., "lp", false)
> and solving this from the command line gives me valid (say: nonzero)
> solutions.
>
> I create the constraints with:
>
> SCIPcreateConsLinear(scip,&cons, name, 0, NULL, NULL, lhs, rhs,
> // initial, separated, enforce, checked, propagated, local, modifiable,
>     true,    true,      true,    true,    true,       false, false,
> // dynamic, removable, sticking
>     false,   false,     false));
>
> Do you have any idea or hints where the problem might be?
>
> Regards,
> Martin
>
> P.S.: Some additional background information: I do a matrix
> decomposition and create some disjoint problems in a relaxator which
> should solve these problems in the relax method. So the problems (each a
> new scip instance) are created in some init function of the relaxator
> and then solved in the relaxator relax method. The sum of the objective
> values should provide a lower bound for the given (at the moment only
> the root) node.
>
>
> _______________________________________________
> Scip mailing list
> Scip at zib.de
> http://listserv.zib.de/mailman/listinfo/scip