reduce type inference on 1.6

[mzgubic]

On julia 1.5

julia> using Test

julia> x_vecs = ([5, ], [1.0, 2.0, 3.0]);

julia> @inferred reduce(vcat, x_vecs)
4-element Array{Float64,1}:
 5.0
 1.0
 2.0
 3.0

On julia 1.6

julia> @inferred reduce(vcat, x_vecs)
ERROR: return type Vector{Float64} does not match inferred return type AbstractVector{T} where T
Stacktrace:
 [1] error(s::String)
   @ Base ./error.jl:33
 [2] top-level scope
   @ REPL[66]:1

julia> @inferred vcat(x_vecs...)
4-element Vector{Float64}:
 5.0
 1.0
 2.0
 3.0

[mzgubic]

Interestingly,

julia> @inferred reduce(vcat, ([10.0], Bool[], [30.0]))
2-element Vector{Float64}:
 10.0
 30.0

julia> @inferred reduce(vcat, ([10.0], [20.0], Bool[]))
ERROR: return type Vector{Float64} does not match inferred return type AbstractVector{T} where T
Stacktrace:
 [1] error(s::String)
   @ Base ./error.jl:33
 [2] top-level scope
   @ REPL[114]:1

[KristofferC]

I'll try bisect it.

[KristofferC]

Bisection blames b4f2c66, cc @dlfivefifty

[dlfivefifty]

Found a fix, will make a PR:

julia> x_vecs = ([5, ], [1.0, 2.0, 3.0]);

julia> @inferred reduce(vcat, x_vecs);
ERROR: return type Vector{Float64} does not match inferred return type AbstractVector{T} where T
Stacktrace:
 [1] error(s::String)
   @ Base ./error.jl:33
 [2] top-level scope
   @ REPL[2]:1

julia> Base._typed_vcat(::Type{T}, V::Base.AbstractVecOrTuple{AbstractVector}) where T = Base._typed_vcat!(Base._typed_vcat_similar(V, T, sum(map(length,V))), V)

julia> @inferred reduce(vcat, x_vecs);

julia>