Double Reverse mode differentiation results in segfault

[dominic-chang]

I'm trying to calculate the Hessian of a function I have that does not have a Forward rule defined for it, so I tried doing a double Reverse mode differentiation instead of a Forward over reverse mode. Doing so however seems to result in a segfault. Here's a MWE where I think I try to calculate the first row of the Hessian:

using Enzyme

x = [2.0,2.0]
bx = zeros(2)
bx2 = zeros(2)
bbx = [1.0,0.0]

function grad(x, bx)
    Enzyme.autodiff(Reverse, Const(sum), Duplicated(x,bx))
    return nothing
end
Enzyme.autodiff(
    Reverse,
    grad,
    Duplicated(x, bx),
    Duplicated(bx2,bbx)
)