[Cbc] Infeasible for CBC (current stable), Value for CPLEX

[John Forrest]

Fixed in  trunk.

Took some time before I worked out ClpPresolve already had fix. Fixed
OsiPresolve.

If presolve changes some integer bounds it goes round again on original
model to see if it can do better.  If there are duplicate integer
columns this can lead to a bug in extreme cases.

John Forrest 


On Tue, 2011-03-01 at 08:16 +0100, Christoph Cullmann wrote:
> Hi,
> 
> following example ILP is infeasible in CBC standalone solver but gets value in 
> CPLEX.
> 
> Any idea, thanks in advance?
> 
> CBC says:
> 
> cullmann at imp:/local/cullmann/build/lpsolve.clpsolve.O0/lpsolve_clpsolve$ 
> ../usr/bin/cbc /local/cullmann/git/lpsolve/value_18621.lp 
> Welcome to the CBC MILP Solver 
> Version: 2.7 
> Build Date: Mar  1 2011 
> 
> command line - ../usr/bin/cbc /local/cullmann/git/lpsolve/value_18621.lp 
> (default strategy 1)
> Continuous objective value is -18621 - 0.00 seconds
> Cgl0000I Cut generators found to be infeasible!
> Pre-processing says infeasible or unbounded
> 
> Greetings
> Christoph
> 
> _______________________________________________
> Cbc mailing list
> Cbc at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/cbc