[Ipopt] huge inf_du values

[Damien]

All,

I'm solving an almost-linear base model with a couple of nonlinear 
constraints and a nonlinear objective.  It's not a particularly 
difficult problem, around 60 equality constraints, and a Newton solver 
solves the non-objective version of the problem in two iterations.  
IPOPT has a terrible time with it though.  I'm using limited-memory 
Hessian and it wanders all over the place.  I'm getting inf_du values of 
6.3e58, which is a bit large.  Anyone seen this before in their problems?

Damien