When I was trying to use the LDLT decomposition inside Eigen2, I found that it only works when the matrix is positive definite.

However, what I know is that the Cholesky decomposition(LLT here) requires the matrix to be positive definite, but the LDLT should only require the matrix to be symmetric and square.

Am I wrong with that? Thanks!

This is definitely fixed in the development version ("default branch", to become 2.1). I don't remember whether we backported that to the stable branch (2.0).