Setting presolve = False makes the problem feasible
I have a problem which HiGHS tells me is infeasible, but when I disable presolve, it becomes feasible
Model file:
I am reasonably sure the problem is feasible. So the question is: why does presolve = "on" make the problem appear infeasible?
I'm using highspy 1.10.0 and cvxpy 1.6.0
This is a bug in the MIP presolve @fwesselm (tagging you so I can also assign it to you).
This looks like a tolerance-related issue (in the forcing row reduction). I am investigating.
I cannot reproduce the infeasibility as of today with the latest branch. However, HiGHS with presolve finds a slightly different optimal objective value.