[Coin-discuss] Strange termination

[Ted Ralphs]

Jean-Sebastien Roy wrote:
> Matthew Saltzman wrote:
> 
>> The docs are not clear what the circumstances are that lead to this 
>> return code.  There is a CPXMIP_UNBOUNDED (unbdd ray found) and a 
>> CPXMIP_INFEAS (proven infeasible).  So I will guess that 
>> CPXMIP_INForUNBD must arise in some ambiguous case.  We'd need to get 
>> CPLEX to weigh in on whether the two cases are distinguishable with 
>> some additional calls.
> 
> 
> In the non-mip case, CPX_STAT_INForUNBD may be returned when the 
> presolve is present, since CPLEX's presolve cannot distinguish the two 
> cases. Disabling the presolve and resolving lead to the proper return 
> code (see OsiCpxSolverInterface.cpp for the relevant code).
> 
> Maybe it's the same in the mip case ?

That's what I was going to guess...in the LP case, we explicitly check 
for this return status and turn off the presolve before resolving the 
problem in this case. This might not be prudent in the MIP case. Take a 
look at the code in initialSolve(). Something similar could be done in 
the MIP case.

Ted
-- 
Dr. Ted Ralphs
Assistant Professor
Industrial and Systems Engineering
Lehigh University
(610)758-4784
tkralphs at lehigh.edu
www.lehigh.edu/~tkr2