Open tkralphs opened 4 years ago
After spending many hours on this, I traced it down to what seems to be a corruption of CPLEX's own internal problem description. Immediately after calling CPXMipopt
, querying the problem data (in particular, the constraint matrix) returns incorrect data. Neither valgrind nor address sanitizer complain about anything. Setting the parameter SolverMasterAsMip
to 0 allows the problem to solve correctly.
I have no idea how to proceed further. It's very, very strange. I find it hard to believe it is a bug in CPLEX, but that is what it appears.
Building 6aa6708f9a6 with default settings and CPLEX 12.7 on Linux with gcc 4.9.2 results in the following error. Correct solution is found with all other algorithms and solvers.