Cholesky bug? #67
Description
I've got a specific case where a cholesky decomposition fails (gets an NaN in the result) using sparse-linear-algebra but succeeds in hmatrix. I believe hmatrix is correct.
I've replicated the example and results here (apologies for the terrible formatting. Brittany!)
https://gist.github.com/adamConnerSax/a77900d61c13c2bd982f6916bba30282
That matrix comes from a linear mixed model example in R (the "Rails" example).