Re: [Help-glpk] GLPSPX2

[Andrew Makhorin]

> I think in the code :
> 
>  case LPX_E_INSTAB:
>             if (spx->msg_lev >= 2)
>                print(prefix "numerical instability (primal simplex, pha"
>                   "se II)");
>             goto feas;
> 
> of the routine GLPSPX2 the last statement causes a loop (the routine doesn't
> return). I replaced this statement by ' goto done ' and then it worked.

This case happens when the current basic solution becomes primal
infeasible due to excessive round-off errors, and the solver tries to
restore its primal feasibility switching to the phase I. That's why
it goes to 'feas'.

Andrew Makhorin