OK, so the method is in there, but the only norms that you can use are L1 and Linf. Any reason why L2 isn’t there? Maybe the error message could be a bit clearer? The way it is written, I am led to believe that the only way to calculate the condition number is to convert to a dense array.
You can get a decent condition estimate for L1 and LInf with an LU facatorization. The L2 condition number needs the largest and smallest singular values, which is more expensive. Maybe ARPack has something you could use for this.