[Coin-discuss] problems with OSL on Red Hat [was: AAP_BP example]

[Matthew Galati]

Hi Laci,

I am not sure it is that simple. I put together a small example at:
http://sagan.ie.lehigh.edu/mgalati/test.tar.gz

which reads in an problem from an mps file (primal infeasible) solves it 
using CLP, CPX and OSL. It then calls getDualRays (call it p). It then 
calls my isProof() function on p and -p. You'll notice that p and -p 
from OSL both return false. This corresponds to what I've been seeing in 
my BCP code.

Matt



Laszlo Ladanyi wrote:
> I have fixed OsiOslSolverInterface::getDualRays(). Basically (as Matt noticed)
> the returned vector vas the negative of the real dual rays. It should work
> properly now.
> 
> --Laci
> 
> On Tue, 2 Dec 2003, Stephan Hennig wrote:
> 
> 
>>Matthew Galati schrieb:
>>
>>
>>>[...]
>>>Adding a sanity check "isProof" should show that the return from OsiOsl's
>>>getDualRays returns a ray that is not a certificate. In my check I am using
>>>the following definition:
>>>
>>>     p is a proof iff
>>>     Let z_j = u_j where p' A_j > 0
>>>             = l_j where p' A_j < 0
>>>     p'b - p'Az > 0
>>>
>>>Using the (negative) of the return of OsiClp::getDualRays works fine. 
>>>Using the current return (or its negative) of OsiOsl and OsiCpx does not 
>>>provide a proof. Using the return of CPXdualfarkas also works fine - but 
>>>this is not currently implemented in OsiCpx. I might be implementing 
>>>something wrong - but that is my guess. For anyone that is interested, I 
>>>have some test cases where the isProof(OsiOsl/Cpx::getDualRays) = false.
>>>[...]
>>
>>I think I got it. The wrong solution value on some problem instances and the
>>segfault are two independent issues. Because they appeared to me at once I
>>didn't separate them.
>>
>>The segfault issue can be worked around by the two alternatives Laszlo posted.
>>The second still endures (and you accidentally discovered the same problem
>>with CPLEX yesterday). When I now read the other thread I must admit you
>>already stated that before.
>>
>>Regards,
>>Stephan Hennig
>>
>>
>>
>>_______________________________________________
>>Coin-discuss mailing list
>>Coin-discuss at www-124.ibm.com
>>http://www-124.ibm.com/developerworks/oss/mailman/listinfo/coin-discuss
>>
> 
> 
> _______________________________________________
> Coin-discuss mailing list
> Coin-discuss at www-124.ibm.com
> http://www-124.ibm.com/developerworks/oss/mailman/listinfo/coin-discuss

-- 
Matthew Galati
ISE Lehigh University
IBM Service Parts Solutions
610.758.4042 (Office)
610.882.0779 (Home)
magh at lehigh.edu, magal11 at us.ibm.com
http://sagan.ie.lehigh.edu/mgalati/