Re: [Bug-glpk] glpsol error

[Andrew Makhorin]

> I have been working with GLPK and Gurobi for some time now.
> Today when trying to solve a LP problem I got a conflict. Gurobi found
> a solution and GLPK reported that the problem had no primal feasible
> solution.
> I have also checked with cplex to confirm that there is in fact a
> solution to the problem, so I have come to the conclusion that there
> might be an error in GLPK.
> 
> 
> I've attached a zip file containing the model, it is in the MPS file
> format.
> I have been using the GLPK stand-alone solver, glpsol, both with and
> without GMP, neither of which found a solution.
> 

Thank your for your report.

I tried to solve your instance with glpsol 4.47 using different options,
and most likely its lp relaxation indeed has no primal feasible solution
at least within tolerances used in the glpk simplex solver by default.
The lp relaxation of your instance is highly degenerate (about half of
the equality constraints are linearly dependent within the working
precision), so it may happen that another solver using weaker tolerances
is able to find a solution that could be considered as feasible.
Unfortunately, the current version of glpk doesn't have a procedure to
prove infeasibility, so glpsol cannot provide more detailed information.


Andrew Makhorin