SYMPHONY Reports No solution Found

[svigerske]

Issue created by migration from Trac.

Original creator: anh509

Original creation time: 2015-02-23 04:30:37

Assignee: @tkralphs

Version: trunk (SVN)

SYMPHONY trunk reports "no solution found" when a node limit is enforced. See the attached MPS and parameters file. This is with CPLEX or Clp as the LP solver. The compile script for both cases is as follows.

with Clp as the LP solver built with
../configure --enable-debug --enable-sensitivity-analysis --disable-openmp

or with CPLEX with the lp solver with ../configure --enable-debug --enable-sensitivity-analysis --disable-openmp --with-cplex-lib=-L/usr/local/CPLEX_Studio_125/cplex/lib/x86-64_sles10_4.1/static_pic/ -lcplex -lpt
hread -lm --with-cplex-incdir=/usr/local/CPLEX_Studio_125/cplex/include/ilcplex --with-lp-solver=cplex --no-create --no-recursion

Please check: ./symphony -F recourse18_5.MPS -f recparams

anh509@polyp1:/SYM-trunk/buildSeq-Clp/bin
OR
anh509@polyp1:/SYM-trunk/buildSeq-Cpx/bin

./symphony -F ~/GenBen-MILPD/recourse18_5.MPS -f ~/GenBen-MILPD/recparams

[svigerske]

Attachment recourse18_5.MPS by anh509 created at 2015-02-23 04:31:07

[svigerske]

Attachment recparams by anh509 created at 2015-02-23 04:31:35

[svigerske]

Comment by @tkralphs created at 2015-03-01 23:53:00

I can't replicate this so far. To begin with, the problem solves in less than 10 nodes, so the parameter file attached doesn't result in the node limit being reached. There is a problem in that the code crashes at the very end when freeing the tree if sensitivity analysis is enabled. This seems to be an issue with an invalid pointer being left around. when I reduce the node limit or try it with other instances, it does, however, seem to report finding a solution, even when there is a node limit for all the cases I have tried. Of course, if no solution has actually been found by the time the node limit is reach, then it report this. I'm assuming you checked that a solution had been found and was just not reported.

I updated your checkout to the most current trunk and re-built it to see if that would fix the problem, so it's possible that this made a difference, though I can't see why it would. Can you check this with the current trunk in your code and see if you're seeing the same thing? I will try to track down the source of the crash at the end when sensitivity analysis is enabled.