Please take a look on the example indicating a false behaviour of compressed_matrix of ublas. In this example, I tried to fill a compressed_matrix by given row pointers and column indices. The matrix has dimension of 186x20 and contains 141 entries.
After fill I tried to print out the values of index1_data and index2_data and observe different results with the inputs. The last component of index1_data is zero which is not correct since compressed format require it to be nnz. Furthermore, the index2_data vector is also longer than the initial column pointers.
I don't think my algorithm to fill the matrix is wrong. In this case, which should be the reason for such behaviour?