Wrong results for forward-mode `exp!` half of the time about chainrules.jl HOT 5 OPEN

julia> using ChainRules, ChainRulesTestUtils, LinearAlgebra, Random, Test

julia> Random.seed!(42);

julia> @testset "exp!" begin
           Xs = (randn(4, 4) for _ in 1:20_000)
           @testset for X in Xs
               test_frule(LinearAlgebra.exp!, X)
           end
       end;
Test Summary: |  Pass  Total   Time
exp!          | 80000  80000  13.2s

ERROR: MethodError: no method matching iterate(::Nothing)

Closest candidates are:
  iterate(::Union{LinRange, StepRangeLen})
   @ Base range.jl:880
  iterate(::Union{LinRange, StepRangeLen}, ::Integer)
   @ Base range.jl:880
  iterate(::T) where T<:Union{Base.KeySet{<:Any, <:Dict}, Base.ValueIterator{<:Dict}}
   @ Base dict.jl:698
  ...

Stacktrace:
 [1] indexed_iterate(I::Nothing, i::Int64)
   @ Base ./tuple.jl:91
 [2] exp!(x1::Matrix{ForwardDiff.Dual{ForwardDiff.Tag{typeof(test_exp), Float64}, Float64, 8}})
   @ Main ~/.julia/packages/ForwardDiffChainRules/2Xt9G/src/ForwardDiffChainRules.jl:81
 [3] test_exp(x::Vector{ForwardDiff.Dual{ForwardDiff.Tag{typeof(test_exp), Float64}, Float64, 8}})
   @ Main ./REPL[4]:3
 [4] chunk_mode_gradient(f::typeof(test_exp), x::Vector{Float64}, cfg::ForwardDiff.GradientConfig{ForwardDiff.Tag{typeof(test_exp), Float64}, Float64, 8, Vector{ForwardDiff.Dual{ForwardDiff.Tag{typeof(test_exp), Float64}, Float64, 8}}})
   @ ForwardDiff ~/.julia/packages/ForwardDiff/vXysl/src/gradient.jl:123
 [5] gradient(f::Function, x::Vector{Float64}, cfg::ForwardDiff.GradientConfig{ForwardDiff.Tag{typeof(test_exp), Float64}, Float64, 8, Vector{ForwardDiff.Dual{ForwardDiff.Tag{typeof(test_exp), Float64}, Float64, 8}}}, ::Val{true})
   @ ForwardDiff ~/.julia/packages/ForwardDiff/vXysl/src/gradient.jl:21
 [6] gradient(f::Function, x::Vector{Float64}, cfg::ForwardDiff.GradientConfig{ForwardDiff.Tag{typeof(test_exp), Float64}, Float64, 8, Vector{ForwardDiff.Dual{ForwardDiff.Tag{typeof(test_exp), Float64}, Float64, 8}}})
   @ ForwardDiff ~/.julia/packages/ForwardDiff/vXysl/src/gradient.jl:17
 [7] gradient(f::Function, x::Vector{Float64})
   @ ForwardDiff ~/.julia/packages/ForwardDiff/vXysl/src/gradient.jl:17
 [8] top-level scope
   @ ./REPL[5]:4

using LinearAlgebra, ForwardDiff, ForwardDiffChainRules, FiniteDifferences
@ForwardDiff_frule LinearAlgebra.exp!(x1::AbstractMatrix{<:ForwardDiff.Dual})
function test_exp(x)
    X = copy(reshape(x, 4, 4))
    X2 = LinearAlgebra.exp!(X)
    sum(X2)
end

for i = 1:20
    X = randn(4,4)
    X = X'X
    x = vec(X)
    g1 = ForwardDiff.gradient(test_exp, x)
    g2 = FiniteDifferences.grad(central_fdm(5, 1), test_exp, x)[1]
    @show norm(g1-g2)
end