[Coin-lpsolver] checkClear() in createRim() and NDEBUG compilation flag
Kish Shen
ks15 at icparc.ic.ac.uk
Mon Aug 15 16:40:42 EDT 2005
Hi John,
John Forrest <jjforre at us.ibm.com> wrote:
>If it fails it is a serious bug. The checkClear is expensive so should not be
>in for production work. Try if a clear() before gets rid of error and if so
I am also running into this assert failure in several problems, when I use Clp
in conjunction with Cbc:
Cbc0015I Node 40 Obj 931.231 Unsat 77 depth 26
Coin0507I Presolve determined that the problem was infeasible with tolerance of 1e-08
eclipse.exe: ClpSimplex.cpp:3367: bool ClpSimplex::createRim(int, bool, int): Assertion `rowArray_[iRow]->capacity()==length' failed.
Signal 6
I have been trying various MIP problems with Osi/Cbc/Clp combination, and
out of 20 problem instances, 4 problems aborted with assert failures, and
at least 2 of these aborted at the above capacity()==length. I don't know
if the problems are due to Cbc or Clp, but the assert failures are in Cbc.
We are using Cbc/Clp as a MP solver for our own ECLiPSe system, and the
problem is loaded into a CoinPackedMatrix in a column-wise fashion. I have
managed to log the calls to Osi from our system, and I have been able to
compile these logged calls as a C++ program, and running this program
reproduce the problem above. If you are willing to look at this problem (I
am afraid I don't know enough about Simplex or Cbc/Clp to look at this
problem myself :-(), I can send you this program.
[the logged program is quite large, as it records all the calls to setting
the matrix coefficients, but it gzips down to about 42K.]
Yours sincerely,
Kish Shen
IC-Parc
Imperial College London
More information about the Clp
mailing list