Inconsistent missing method error

[timholy]

Here's a situation in which @which reports an available method, but actually trying to use the method results in a MethodError. I'm probably doing something wrong, but for the life of me I can't figure out what. This happens on both master and 0.3.

julia> using Images, TestImages, Color

julia> img = testimage("lighthouse")
RGB Image with:
  data: 768x512 Array{RGB{UfixedBase{Uint8,8}},2}
  properties:
    IMcs: sRGB
    spatialorder:  x y
    pixelspacing:  1 1

julia> @which convert(Image{RGB}, img)
convert{Cdest<:Union(AbstractAlphaColorValue{C<:ColorValue{T},T<:Real},ColorValue{T}),Csrc<:Union(AbstractAlphaColorValue{C<:ColorValue{T},T<:Real},ColorValue{T})}(::Type{Image{Cdest<:Union(AbstractAlphaColorValue{C<:ColorValue{T},T<:Real},ColorValue{T}),N,A<:AbstractArray{T,N}}},img::AbstractImageDirect{Csrc<:Union(AbstractAlphaColorValue{C<:ColorValue{T},T<:Real},ColorValue{T}),N}) at /home/tim/.julia/v0.3/Images/src/core.jl:226

julia> convert(Image{RGB}, img)
ERROR: `convert` has no method matching convert(::Type{Image{RGB{T<:Union(FixedPoint,FloatingPoint)},N,A<:AbstractArray{T,N}}}, ::Image{RGB{UfixedBase{Uint8,8}},2,Array{RGB{UfixedBase{Uint8,8}},2}})
 in convert at base.jl:13

# OK, directly test that definition on Images/src/core.jl:226
julia> copy(img, Images._convert(Array{RGB}, data(img)))
RGB Image with:
  data: 768x512 Array{RGB{UfixedBase{Uint8,8}},2}
  properties:
    IMcs: sRGB
    spatialorder:  x y
    pixelspacing:  1 1

# Can we replicate with some function besides convert?
julia> arg1{C<:Images.ColorTypes.ColorType}(::Type{Image{C}}) = "yes"
arg1 (generic function with 1 method)

julia> arg2{C<:Images.ColorTypes.ColorType}(::AbstractImageDirect{C}) = "yes"
arg2 (generic function with 1 method)

julia> arg12{C<:Images.ColorTypes.ColorType, D<:Images.ColorTypes.ColorType}(::Type{Image{C}}, ::AbstractImageDirect{D}) = "yes"
arg12 (generic function with 1 method)

julia> arg1(Image{RGB})
"yes"

julia> arg2(img)
"yes"

julia> arg12(Image{RGB}, img)
"yes"

[simonster]

I'm not sure if it's the same issue, but this looks vaguely similar to #8128.

[timholy]

I bet you're right. I'll close this one and bump that one.