[Ipopt] Pb with get_variables_linearity

[Stefan Vigerske]

Hi,

I have actually never checked it with my code, when it gets executed.
Hmm, and now where I have a closer look I see that there are additional
virtual TNLP functions called get_number_of_nonlinear_variables() and
get_list_of_nonlinear_variables() (hmm, I never recognized them :-(). I
thought yesterday their output would be computed from the output of
get_variables_linearity, but looks like one has to implement both (why??).

So I have to change my previous mail. You may better check your
implementation of get_number_of_nonlinear_variables() and
get_list_of_nonlinear_variables(). The output of that one seem to be the
input for ExpansionMatrixSpace where the valgrind complain occurs. See
also TNLPAdapter::GetQuasiNewtonApproximationSpaces in TNLPAdapter.cpp:1941.

Stefan

Am 20.05.2010 10:34, schrieb Yann COLLETTE:
> Hello,
> 
> You where right Stephan, my problem effectively comes from the 
> get_variables_linearity.
> I have defined my problem in a TNLP class.
> I a redefined the method get_variables_linearity.
> But when I  start the resolution, the method get_variables_linearity is 
> never called.
> Do I need to set a special option so that the method is called ?
> 
> 
> YC
> 
> _______________________________________________
> Ipopt mailing list
> Ipopt at list.coin-or.org
> http://list.coin-or.org/mailman/listinfo/ipopt
>