Originally posted by: JesusCerquides

Some days ago, I reported a bug in the presolver.

https://www.ibm.com/developerworks/community/forums/html/topic?id=af040717-eb79-4e78-98eb-62927676828b&ps=25

As DanielJunglas suggested, I have set CPX_PARAM_AGGIND to 0, from there on, to avoid that problem.

Now, it seems that I hit some other strange behavior. Looks like the kind of problems I am solving have some specific characteristic that CPLEX does not like. :) 

In the attached problem, CPLEX fails to find the optimal solution when CPX_PARAM_AGGIND is set to 0. However it does find it when CPX_PARAM_AGGIND is set to -1. In both cases the solutions are reported as optimal. I do also attach both solutions.

My platform is Linux 64 and CPLEX is version 12.6.0.0 

Could this be due to the MIP gap tolerance?

Jesús

Originally posted by: JesusCerquides

It does. Setting both absolute and relative mip gap tolerances to 0 solves the problem. Sorry for bothering.

Jesús.