[Ipopt] Crash in Matlab Interface for larger problems

[Tony Kelman]

Patrick,

Aha! Jonathan Currie just got to the bottom of the memcpy issue I mentioned, 
and it may be what you're running into. It's similar to a previously seen 
and fixed bug (https://projects.coin-or.org/Ipopt/ticket/187) with sparse 
gradients, and we can fix it in Ipopt trunk without too much trouble. In the 
meantime as a workaround, check and make sure that your initial condition 
vector X0 is full(), and that your jacobianstructure and/or hessianstructure 
callbacks return double() sparse matrices, not logical data types.

-Tony