[Coin-ipopt] NLPAPI memory problem?

Andreas Waechter andreasw at watson.ibm.com
Fri May 6 15:00:27 EDT 2005

Hi Matt,

Looking at your example, I found a bug in the memory allocation in the
Ipopt interface for the NLPAPI (memory of length 0 was allocated - which
seems Ok on Linux, but not AIX).  I fixed this and the code is in the CVS
repository now.

But that does not seem to be the problem you describe (since you would
have seen that mistake already when solving just one NLP).  We have
corrected a few other bugs in the NLPAPI and submitted the corrections
just yesterday, so you probably didn't have the latest code.  Please
update your code of the NLPAPI (if you don't use CVS you have to wait
until tomorrow to get the latest changes in the tarball).  If you still
have a problem please let us know (and tell us what operating system and
compiler you are using).



On Fri, 6 May 2005, Matthew Guthaus wrote:

> This should probably go to the NLPAPI mailing list, but I do not see
> one.
> I have attached a small example that uses an equality constraint. If I
> repeat the optimization in a loop while (properly, according to the
> examples) freeing the memory, it appears that something gets corrupted.
> Removing the equality constraint allows it to complete fine.
> Am I making a mistake? Is this a problem in the NLPAPI?
> Thanks,
> Matt

More information about the Coin-ipopt mailing list