Originally posted by: JesusCerquides

Hello, 

I have found a problem whilst solving the MIP problem attached. If I use the default opt command (which performs presolving) I get a non-feasible solution reported as optimal (with a value of 6.14). However, if I disable presolving by doing "set preprocessing presolve no", then I get the correct optimal solution with value 4.55. 

Thus, it looks like the presolver is performing some inadequate simplifications. Am I getting something wrong. 

Jesús.

P.S. The problem is correctly solved by gurobi and glpk. 

Are you sure you did attach the right problem? Solving the attached model here I consistently get 49.75 as optimal solution, no matter whether I disable presolve or not.

What version of CPLEX are you using?

Originally posted by: JesusCerquides

Sorry Daniel, 

This is because I am just stupid and attached the wrong file. Please try again with this one. 

Jesús.

Originally posted by: UserCplex

Yes. Seems like a bug. :--(

Got dissimilar answers with and without setting presolve.

Using Windows, CPLEX 12.6.0.0

I could reproduce the faulty behavior here. So it indeed looks like a bug. I have not yet found the reason for the problem but a less drastic workaround seems to be to only disable the aggregrator (set CPX_PARAM_AGGIND to 0) instead of all of presolve.

We have found the problem. Indeed, the best workaround is to set CPX_PARAM_AGGIND to 0. The bug will be fixed in the next release.

Originally posted by: JesusCerquides

I am solving with CPLEX 12.6.0.0 on Ubuntu 14.04 64 bits.