Hello,

I’m using the following Lagrangian definition in my solver: \mathcal{L}(x, y, z) = f(x) - y^T c(x) - z, while the ASL Hessian is defined as \nabla^2 \mathcal{L}(x, y) = \nabla^2 f(x) + \sum_{j=1}^m y_j \nabla^2 c_j(x).

Therefore I call ASL’s `lagscale`

with scaling coefficient -1, but this also flips the signs of the dual solution when I call `write_sol`

. I would like to avoid that and write my dual solution as is.

A call to `lagscale`

with scaling coefficient 1 just before calling `write_sol`

has no effect (it terminates immediately). What are my options here?

Thank you,

Charlie