SVD U is not always orthonormal

[chriscoey]

using GenericLinearAlgebra
A = zeros(BigFloat, 2, 2)
F = GenericLinearAlgebra.svd(A, full = false)

LinearAlgebra.SVD{BigFloat, BigFloat, Matrix{BigFloat}}
U factor:
2×2 Matrix{BigFloat}:
 0.0  0.0
 0.0  0.0
singular values:
2-element Vector{BigFloat}:
 0.0
 0.0
Vt factor:
2×2 Matrix{BigFloat}:
 1.0  0.0
 0.0  1.0

U is not orthonormal, though Vt is. It would be more convenient and consistent with e.g. LAPACK if U and Vt are both always orthonormal.

[andreasnoack]

Yeah. It definitely should be. Not sure what is going wrong here. I'll take a look.

[andreasnoack]

This was fixed in 85e8031de66020819bb61146163ed51f9fa3e006