3rd order approximation of policy function

[vanilla]

Hi,
I have a question about 3rd order approximation.
I compared the 3rd order terms of policy function using dynare 4.1.1(oo_.dr.g_3) and dynare++(dyn_g_3) but it did not match for simple rbc-model (dynare uses "use_dll" option and the MSVC compiler).
So, I wrote a simpler model which has simple analytical expression for third order terms to see the difference of dynare and dynare++ and found that cross terms (e.g x^2*y) coefficients do not appear in dynare( oo_.dr.g_3 ) (up to 2nd order terms both had same result).
Does this mean that dynare 3rd order has some problem?
Do you know how I can get the same coefficients using dynare as dynare++?
I attached mod-file of the simple model for dynare(test.mod) and dynare++(testplus.mod). Also, I attached rbc model mod-file (rbcmodel.mod ) . First part of rbcmodel.mod-file is for dynare and second part is for dynare++.

Thank you very much

Vanilla

[MichelJuillard]

Dear Vanilla,

thank for the post and the careful testing. It helped us to fix a bug in 3rd order approximation in Dynare Matlab. The fix should be available in tomorrow's snapshot of dynare_unstable.

Kind regards,

Michel