Originally posted by: Enrico77

I have to solve a linear programming problem. I declared a part of the constraints as  the "lazy" ones, this allows me to reduce the execution time. For this reason, I have to change the LP problem into a MILP problem. Otherwise, I cannot use the lazy constraints,

However, I need to get the dual variables. Finally, I would like (i) to use  the lazy constraints and (ii) to get the dual optimal variables.  

I tried to use the CPXsetsolvecallbackfunc function whereby I can access to the root node. In fact, by means of CPXgetcallbacknodelp I can solve the model asking the dual variables, but this is not enough because the model is the pre-processed one.  If I disabled the pre-process, I would lose the effectiveness of the method. How could I do?

Have you tried using CPXuncrushpi() on the presolved model?

Originally posted by: Enrico77

Thanks! I will try your solution. 

Best,

Enrico