[Ipopt] Does Ipopt guarantee x to be within lower and upper bounds?

[Peng Yu]

> that's no bug, that's a feature.
> There is an option to disable this:
> http://www.coin-or.org/Ipopt/documentation/node43.html#SECTION000114010000000000000

I'm trying to see the effect of this option. I have tried to set
bound_relax_factor to zero for my complex problem, but I still see
evaluations outside the bound. So I need to have a simpler case to
make sure that I set the ipopt parameter correctly.

Are you aware a simple test case that would result in evaluation of
objective/gradient/hessian outside the boundary when the default value
of bound_relax_factor? Thanks.

-- 
Regards,
Peng