[Coin-discuss] Strange termination

[Matthias Koeppe]

Kasper Bonne Rasmussen <s992293 at imm.dtu.dk> writes:

> On Fri, 17 Jun 2005, Jean-Sebastien Roy wrote:
>> Mr Rasmussen, could you try with the presolve disabled and see if the 
>> status returned is correct ?
>> 
>> If it does, does the following patch fixes the status in all cases ?
>
> It compiles okay (with CPX_MIP_INForUNBD changed to CPXMIP_INForUNBD), but
> it does not solve the problem. The status is still CPXMIP_INForUNBD
> (119) after the second solve.

In my code (not related to COIN-OR), I use the following technique to
handle CPXMIP_INForUNBD.

    {
	int save_preind;
	int save_relaxpreind;
	(void) CPXgetintparam(env, CPX_PARAM_PREIND, &save_preind);
	(void) CPXgetintparam(env, CPX_PARAM_RELAXPREIND, &save_relaxpreind);
	/* "infeasible or unbounded" -- but we need to know which!
	   So turn off the presolver for this problem and try again. */
	(void) CPXsetintparam(env, CPX_PARAM_PREIND, CPX_OFF);
	(void) CPXsetintparam(env, CPX_PARAM_RELAXPREIND, CPX_OFF);
	/* In CPLEX 9, this is a way to discard the presolve information: */
	{
	    int objsen = CPXgetobjsen(env, lp);
	    CPXchgobjsen(env, lp, -objsen);
	    CPXchgobjsen(env, lp, objsen);
	}
	cpxstat = CPXmipopt(env, lp);
	(void) CPXsetintparam(env, CPX_PARAM_PREIND, save_preind);
	(void) CPXsetintparam(env, CPX_PARAM_RELAXPREIND, save_relaxpreind);
    }

Regards
-- 
Matthias Koeppe -- http://www.math.uni-magdeburg.de/~mkoeppe