> Looks like this is a bug in the new 'invert' which punts to invert_by_lu. Some time ago, I think I said that I didn't feel the love for the adjoint method for matrix inversion. But it does have the advantage of never needing to check if some huge expression vanishes.. --Barton