Don't assume Int in _sametype in broadcasting

[dlfivefifty]

These lines are breaking InfiniteArrays.jl in Julia v1.0.5 (but not other v1.x or v1.0.x):

julia/base/broadcast.jl

Line 448 in 3af96bc

_sametype(a::OneTo, b::OneTo) = OneTo{Int}(a)

It seems like a bad idea to assume Int, as they may be BigInt. I can work around it but this should have been considered a breaking change.

[KristofferC]

Wasn't this put into 1.0.3?
#30074

[dlfivefifty]

Ah, I only checked 1.0.2 and 1.0.5

[dlfivefifty]

I realised the issue is with the call to UnitRange, not the use of Int. I need to hack around due to the nature of InfiniteArrays.jl, so from my perspective this issue can be closed.

However, I think the hard coding of Int is very suspicious, and it seems that it should be

_sametype(a::OneTo{T}, b::OneTo{V}) where {T,V} = OneTo{promote_type(T,V)}(a) 

[timholy]

We default to Int for all scalar indexing operations:

julia> to_indices(rand(2), (0x01,))
(1,)

[dlfivefifty]

OK, fair enough. In any case that code has been removed and I've fixed the issue in InfiniteArrays.jl. So I'll close this.