The SDP code for creating dual problems had two problems (see #104): slicing was being done in a way not resistant to reordering of the variables (which was happening spontaneously), and coefficients in the objective function associated to known variables were being double-counted.
The SDP code for creating dual problems had two problems (see #104): slicing was being done in a way not resistant to reordering of the variables (which was happening spontaneously), and coefficients in the objective function associated to known variables were being double-counted.
The code passes all tests.