[Cbc] Possible CbcBranchDynamic.cpp BUG

[Haroldo Santos]

Did more experiments to see when this bug occurs.

To turn off all heuristics and cuts it makes it happens faster.

This is an instance
http://www.decom.ufop.br/haroldo/files/sprint_hidden06.lp.gz
which is quickly solved by CPLEX and has optimal objective function value
130.

CBC crashes at this point:

Cbc0010I After 1700 nodes, 656 on tree, 249 best solution, best possible
99.390873 (139.08 seconds)
cbc: CbcBranchDynamic.cpp:551: virtual int
CbcBranchDynamicDecision::betterBranch(CbcBranchingObject*,
CbcBranchingObject*, double, int, double, int): Assertion `distance >= 0.0'
failed








On Wed, Aug 24, 2011 at 5:43 PM, Haroldo Santos <haroldo.santos at gmail.com>wrote:

> Dear Forrest/CBC Developers,
>
> There is this assertion failing  in :
> CbcBranchDynamic.cpp:551: virtual int
> CbcBranchDynamicDecision::betterBranch(CbcBranchingObject*,
> CbcBranchingObject*, double, int, double, int): Assertion `distance >= 0.0'
> failed.
>
> when I'm solving this MIP:
> http://www.decom.ufop.br/haroldo/files/sprint_hidden04.lp.gz
>
> I did some debug and discover that this fails because at a given point of
> the search the distance is  -0.415667  .
>
> Cbc0010I After 10000 nodes, 4845 on tree, 180 best solution, best possible
> 142.25 (1909.96 seconds)
> Cbc0010I After 10100 nodes, 4849 on tree, 180 best solution, best possible
> 142.25 (1919.14 seconds)
>
> Tested in Linux with CBC 2.7 stable, default compilation options and no
> ThirdParty software included.
>
> The stack trace at this point is:
> #0  0x00007ffff779c5cb in CbcBranchDynamicDecision::betterBranch
> (this=0xb3eee10, thisOne=
>     0x12d60060, changeUp=0, numInfUp=205, changeDown=0, numInfDown=211) at
> CbcBranchDynamic.cpp:555
> #1  0x00007ffff784bf40 in CbcNode::chooseDynamicBranch (this=0x12d9a290,
> model=0x60dcd0,
>     lastNode=0x12aa7510, numberPassesLeft=4) at CbcNode.cpp:3178
> #2  0x00007ffff783521c in CbcModel::chooseBranch (this=0x60dcd0,
> newNode=@0x7fffffff53a0,
>     numberPassesLeft=5, oldNode=0x12aa7510, cuts=...,
> resolved=@0x7fffffff4d18, lastws=0x12e21f40,
>     lowerBefore=0x12ac54b0, upperBefore=0x1396c2e0,
> branches=@0x7fffffff4b08) at CbcModel.cpp:12642
> #3  0x00007ffff783b052 in CbcModel::doOneNode (this=0x60dcd0,
> baseModel=0x60dcd0,
>     node=@0x7fffffff5298, newNode=@0x7fffffff53a0) at CbcModel.cpp:14299
> #4  0x00007ffff78108f5 in CbcModel::branchAndBound (this=0x60dcd0,
> doStatistics=0)
>     at CbcModel.cpp:3797
> #5  0x00007ffff7b7be8b in CbcMain1 (argc=11, argv=0x7fffffffdef8,
> model=...,
>     callBack=0x7ffff7b67569 <dummyCallBack(CbcModel*, int)>) at
> CbcSolver.cpp:5589
> #6  0x00007ffff7b67bb5 in CbcMain1 (argc=11, argv=0x7fffffffdef8,
> model=...) at CbcSolver.cpp:1111
> #7  0x0000000000400f47 in main (argc=11, argv=0x7fffffffdef8) at
> CoinSolve.cpp:336
>
> Cheers,
>
> Haroldo
>
> --
> =============================================================
> Haroldo Gambini Santos
> Computing Department - Universidade Federal de Ouro Preto - UFOP
> email: haroldo [at ] iceb.ufop.br
> home/research page: www.decom.ufop.br/haroldo/
>
> "Computer science is no more about computers than astronomy
> is about telescopes." Edsger Dijkstra
>
>



-- 
=============================================================
Haroldo Gambini Santos
Computing Department - Universidade Federal de Ouro Preto - UFOP
email: haroldo [at ] iceb.ufop.br
home/research page: www.decom.ufop.br/haroldo/

"Computer science is no more about computers than astronomy
is about telescopes." Edsger Dijkstra
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://list.coin-or.org/pipermail/cbc/attachments/20110902/f98015ca/attachment.html