[Clp] Regression with the current trunk.

[ycollette.nospam at free.fr]

Hello,

I finally found the good set of options for the clp executable.

First, I checkout the Cbc trunk.
Then, I cd into CoinUtils directory and do a 'svn up -r 1698'
Then, I cd into the main trunk directory:
./configure --enable-aboca=2 --prefix=/opt/cbc-devel-ok LDFLAGS=-lpthread
make
make install

Then, I cd into CoinUtils directory and do a 'svn up'
Then, I cd into the main trunk directory:
./configure --enable-aboca=2 --prefix=/opt/cbc-devel-nok LDFLAGS=-lpthread
make clean
make
make install

For my tests, I use the attached lp file.

In a first terminal:
export LD_LIBRARY_PATH=/opt/cbc-devel-nok/lib:$LD_LIBRARY_PATH
/opt/cbc-devel-nok/bin/clp Coin_1.lp -abc on -dualS on -solution Coin_1_nok.sol -solve

In a second terminal:
export LD_LIBRARY_PATH=/opt/cbc-devel-ok/lib:$LD_LIBRARY_PATH
/opt/cbc-devel-ok/bin/clp Coin_1.lp -abc on -dualS on -solution Coin_1_ok.sol -solve

diff Coin_1_ok.sol Coin_1_nok.sol
18c18
<      15 x13                    0                       0
---
>      15 x13                    1                       0

By the way, under fedora 21 64 bits, I needed to add LDFLAGS=-lpthread to allow executables to be built.

Best regards,

YC


----- Mail original -----
De: "Yann COLLETTE" <ycollette.nospam at free.fr>
À: clp at list.coin-or.org
Envoyé: Mercredi 13 Août 2014 19:27:07
Objet: Re: [Clp] Regression with the current trunk.

I will try to see if I am able to extract a problematic lp file from our 
tool.

Best regards,

YC

Le 13/08/2014 18:09, John Forrest a écrit :
> I can't see it being anything other than CoinPresolveDual.cpp changes 
> - dual presolve being tricky.
>
> I could back it out, but it was helping on some difficult problems, so 
> I would prefer to fix it.
>
> Is it easy for you to give me code/problem that breaks?
>
> Regards,
>
> John Forrest
>
>
>
> On 13/08/14 16:19, ycollette.nospam at free.fr wrote:
>> Hello,
>>
>> I was testing the current trunk of clp against our tool.
>> And with the current trunk version, I noticed a regression on some 
>> tests.
>> After trying to understand what was the problematic part I finally 
>> bisected the CoinUtils code until commit 1699.
>>
>> r1699 | forrest | 2014-05-22 19:06:09 +0200 (jeu. 22 mai 2014) | 2 
>> lignes
>>
>>   changes to presolve - hopefully good ones
>>
>> Before this commit, there are no problems in our test base, after, 
>> some tests are failling.
>> I just bisected the coinutils code. I leave the clp code untouched. 
>> So, the problem seems to be really located into coinutils.
>>
>> Best regards,
>>
>> YC
>> _______________________________________________
>> Clp mailing list
>> Clp at list.coin-or.org
>> http://list.coin-or.org/mailman/listinfo/clp
>>
>>
>
> _______________________________________________
> Clp mailing list
> Clp at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/clp

_______________________________________________
Clp mailing list
Clp at list.coin-or.org
http://list.coin-or.org/mailman/listinfo/clp
-------------- next part --------------
A non-text attachment was scrubbed...
Name: Coin_1.lp
Type: application/octet-stream
Size: 1071 bytes
Desc: not available
URL: <http://list.coin-or.org/pipermail/clp/attachments/20140814/9ebc98da/attachment.obj>