[Cbc] Cbc segfault after many hours of CPU time

[acw at ascent.com]

I have a model which I have been trying to solve thus:

> ./cbc -threads 4 -import trouble.mps -solve -quit

As you can see, the only even slightly exotic thing I'm doing is enabling 
multithreading, which has worked fine for me with less-challenging models.

Cbc works on this for several hours and then segfaults.  I tried it three 
times.  On two occasions, it did about 5.5 M nodes before dying (taking 
about half a CPU-day, but only three or four hours of wallclock time).  On 
the third occasion, it managed more than 27 M nodes and 2.5 CPU-days (but 
only overnight by the wall clock).

On all occasions, the number of nodes on the tree rises rapidly to about 
93 K, then remains approximately constant (within a few hundred) for the 
rest of the ill-fated run.

Has anyone else had similar experiences?

I have not yet tried running the model without multithreading, for obvious 
practical reasons: it would probably take twelve hours if I was lucky, and 
might take sixty.

I'll be glad to share the MPS file (about a quarter of a meg uncompressed) 
with anybody who wants to try this.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.coin-or.org/pipermail/cbc/attachments/20130308/3462fd14/attachment.html>