It seems there is a bug in the latest version of Dynare ++. When computing the 2nd order approximation, the solution for the 1st order expansion changes relative to the one obtained when computing the linear solution. Regardless of this error, the dynare ++ solution (with the --nocentralize option) is not exactly the same as the one from dynare. Although in my example the error is in the second and third decimal, these errors can easily accumulate in simulations.
Pablo A. Guerron
If you mean that the constant term is changing when going from 1st order to 2nd order, this is normal: it is the effect of future volatility on the level of the policy function. In particular, note that Dynare and Dynare++ don’t provide the Taylor expansion of the true solution: they instead provided the exact solution of an approximation of the model, which is quite different.
If you mean something else, it may be a bug: please post your MOD file so that we can replicate the problem and fix it.