This seems like a presolver issue. A big hint is the following: > Preprocessing... > 3 rows, 0 columns, 0 non-zeros > 0 integer variables, none of which are binary > Indeed, disabling the mip presolver with --nointopt works as expected: "PROBLEM HAS NO INTEGER FEASIBLE SOLUTION"
Tracing this a bit, the root cause seems to be the rather relaxed tolerance in npp_empty_row() - setting it to 1e-5 (from 1e-3) avoids this issue. Andrew, what do you think? PS. The actual crash is due to the glp_add_cols() call in ios_create_pool(), maybe this should be made conditional as well. Note that with this change the crash is avoided but the solver finds a (wrong) solution to the problem.
_______________________________________________ Bug-glpk mailing list [email protected] https://lists.gnu.org/mailman/listinfo/bug-glpk
