Closed dpsanders closed 4 years ago
I believe this comes from a division bug in IntervalArithmetic.jl
when the numerator is exactly 0
:
julia> 0.0/((-10)..10)
[0, 0]
Here the mid
of the interval simply appears to be an exact solution at some point of the computation.
Cf. https://github.com/JuliaIntervals/IntervalArithmetic.jl/issues/264
Has since been fixed:
julia> roots(f5, -10..10)
4-element Array{Root{Interval{Float64}},1}:
Root([0.999999, 1.00001], :unknown)
Root([-1.41422, -1.41421], :unknown)
Root([-1.00001, -0.999999], :unknown)
Root([1.41421, 1.41422], :unknown)
Seems to be a special case when the root is exact: