atan2 return value seems wrong

[Keno]

julia> (pivot,opposite)
(FixedSizeArrays.Point{2,Int32}((506,432)),FixedSizeArrays.Point{2,Int32}((282,350)))

julia> atan2(pivot, opposite)
FixedSizeArrays.Point{2,Float64}((1.0623564415840052,0.8898756879564043))

julia> atan2(pivot[1]-opposite[1],pivot[2]-opposite[2])
1.219876329384468

[SimonDanisch]

It applies it vectorized like in base:

julia> p, o = convert(Array, pivot), convert(Array, opposite)
julia> atan2(p, o)
2-element Array{Float64,1}:
 1.06236
 0.889876

Another problem is, that on master I switched back to not let functions change the return type.
So it will do Vec{2, Int32}(atan2(a[1], b[1]), atan2(a[2], b[2])) which will result in an InexactError.
I should change that back at some point, but I need to do this cleanly. Otherwise type inference will cry.

[Keno]

That makes sense for Vecs, but I don't really like that behavior for Points.

[SimonDanisch]

Ah I see your point!
I haven't implemented any Point specific methods yet, so it behaves identical to Vec.
If we start doing things like that, -, +, * need their own definition as well.