[Coin-discuss] OsiCpxSolverInterface incompatible with CGL

[Ted Ralphs]

Jonathan,

Actually, Tobi has modified the CPLEX interface and the bug has been 
"mostly" fixed. By commenting out the CPLEX calls in setInteger() and 
setContinuous(), so that variable statuses only get set in the local 
cache, you can get things to work. I still get a "No solution exists" 
error from CPLEX with some of the CGL calls, but it doesn't seem to 
cause any problems---I am still tracking down the cause of that error. 
At any rate, I am using the full suite of cut generators with OSI and 
CPLEX and have done some pretty extensive testing with MIPLIB3 and 
haven't had any problems so far (except that the Simple Rounding CGL 
seems to be buggy).

By the way, the above fix breaks the branchAndBound call since the 
status of the integer variables doesn't get passed on to CPLEX anymore 
when calling setInteger. However, this should be easy to fix by passing 
the contents of the cache to CPLEX just before calling CPXmipopt.

After I've played around with it a bit more and fixed the branchAndBound 
call, I'll post the patched version. Let me know if you want what I 
currently have.

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