Pass through of options to kkt() from within solvebilevel()

[Zeb Tate]

I was attempting to turn off the computation of dual bounds during the solution of a bilevel problem, and I noticed that the following call from within solvebilevel does not pass through the options structure:

solvebilevel.m, line 56:
    K = kkt(InnerConstraints,InnerObjective,z);

Is there a reason that this call to kkt() is uses the default options (rather than have the call be kkt(InnerConstraints,InnerObjective,z,options))?

[Johan Löfberg]

Not intentional. I wouldn't recommend turning off the dual bound propagation though, a KKT approach must have good dual bounds. (most likely the derivation of dual bounds fails though anyway, so it is really recommended to set up the problem manually using the kkt operator and then explicitly add bounds on the dual variables for which no bounds were derived)

[Johan Löfberg]

Fixed in the next version.