diff --git a/base/broadcast.jl b/base/broadcast.jl
index c6651e28489a3..b34a73041708b 100644
--- a/base/broadcast.jl
+++ b/base/broadcast.jl
@@ -1121,19 +1121,20 @@ end
 
 ## scalar-range broadcast operations ##
 # DefaultArrayStyle and \ are not available at the time of range.jl
-broadcasted(::DefaultArrayStyle{1}, ::typeof(+), r::OrdinalRange) = r
-broadcasted(::DefaultArrayStyle{1}, ::typeof(+), r::StepRangeLen) = r
-broadcasted(::DefaultArrayStyle{1}, ::typeof(+), r::LinRange) = r
+broadcasted(::DefaultArrayStyle{1}, ::typeof(+), r::AbstractRange) = r
 
-broadcasted(::DefaultArrayStyle{1}, ::typeof(-), r::OrdinalRange) = range(-first(r), step=-step(r), length=length(r))
+broadcasted(::DefaultArrayStyle{1}, ::typeof(-), r::AbstractRange) = range(-first(r), step=-step(r), length=length(r))
+broadcasted(::DefaultArrayStyle{1}, ::typeof(-), r::OrdinalRange) = range(-first(r), -last(r), step=-step(r))
 broadcasted(::DefaultArrayStyle{1}, ::typeof(-), r::StepRangeLen) = StepRangeLen(-r.ref, -r.step, length(r), r.offset)
 broadcasted(::DefaultArrayStyle{1}, ::typeof(-), r::LinRange) = LinRange(-r.start, -r.stop, length(r))
 
-broadcasted(::DefaultArrayStyle{1}, ::typeof(+), x::Real, r::AbstractUnitRange) = range(x + first(r), length=length(r))
-broadcasted(::DefaultArrayStyle{1}, ::typeof(+), r::AbstractUnitRange, x::Real) = range(first(r) + x, length=length(r))
 # For #18336 we need to prevent promotion of the step type:
 broadcasted(::DefaultArrayStyle{1}, ::typeof(+), r::AbstractRange, x::Number) = range(first(r) + x, step=step(r), length=length(r))
 broadcasted(::DefaultArrayStyle{1}, ::typeof(+), x::Number, r::AbstractRange) = range(x + first(r), step=step(r), length=length(r))
+broadcasted(::DefaultArrayStyle{1}, ::typeof(+), r::OrdinalRange, x::Real) = range(first(r) + x, last(r) + x, step=step(r))
+broadcasted(::DefaultArrayStyle{1}, ::typeof(+), x::Real, r::Real) = range(x + first(r), x + last(r), step=step(r))
+broadcasted(::DefaultArrayStyle{1}, ::typeof(+), r::AbstractUnitRange, x::Real) = range(first(r) + x, last(r) + x)
+broadcasted(::DefaultArrayStyle{1}, ::typeof(+), x::Real, r::AbstractUnitRange) = range(x + first(r), x + last(r))
 broadcasted(::DefaultArrayStyle{1}, ::typeof(+), r::StepRangeLen{T}, x::Number) where T =
     StepRangeLen{typeof(T(r.ref)+x)}(r.ref + x, r.step, length(r), r.offset)
 broadcasted(::DefaultArrayStyle{1}, ::typeof(+), x::Number, r::StepRangeLen{T}) where T =
@@ -1142,9 +1143,11 @@ broadcasted(::DefaultArrayStyle{1}, ::typeof(+), r::LinRange, x::Number) = LinRa
 broadcasted(::DefaultArrayStyle{1}, ::typeof(+), x::Number, r::LinRange) = LinRange(x + r.start, x + r.stop, length(r))
 broadcasted(::DefaultArrayStyle{1}, ::typeof(+), r1::AbstractRange, r2::AbstractRange) = r1 + r2
 
-broadcasted(::DefaultArrayStyle{1}, ::typeof(-), r::AbstractUnitRange, x::Number) = range(first(r)-x, length=length(r))
-broadcasted(::DefaultArrayStyle{1}, ::typeof(-), r::AbstractRange, x::Number) = range(first(r)-x, step=step(r), length=length(r))
-broadcasted(::DefaultArrayStyle{1}, ::typeof(-), x::Number, r::AbstractRange) = range(x-first(r), step=-step(r), length=length(r))
+broadcasted(::DefaultArrayStyle{1}, ::typeof(-), r::AbstractRange, x::Number) = range(first(r) - x, step=step(r), length=length(r))
+broadcasted(::DefaultArrayStyle{1}, ::typeof(-), x::Number, r::AbstractRange) = range(x - first(r), step=-step(r), length=length(r))
+broadcasted(::DefaultArrayStyle{1}, ::typeof(-), r::OrdinalRange, x::Real) = range(first(r) - x, last(r) - x, step=step(r))
+broadcasted(::DefaultArrayStyle{1}, ::typeof(-), x::Real, r::OrdinalRange) = range(x - first(r), x - last(r), step=-step(r))
+broadcasted(::DefaultArrayStyle{1}, ::typeof(-), r::AbstractUnitRange, x::Real) = range(first(r) - x, last(r) - x)
 broadcasted(::DefaultArrayStyle{1}, ::typeof(-), r::StepRangeLen{T}, x::Number) where T =
     StepRangeLen{typeof(T(r.ref)-x)}(r.ref - x, r.step, length(r), r.offset)
 broadcasted(::DefaultArrayStyle{1}, ::typeof(-), x::Number, r::StepRangeLen{T}) where T =
diff --git a/base/range.jl b/base/range.jl
index ee3c86c3f103f..b6a8e333cf52e 100644
--- a/base/range.jl
+++ b/base/range.jl
@@ -24,9 +24,9 @@
 _colon(::Ordered, ::Any, start::T, step, stop::T) where {T} = StepRange(start, step, stop)
 # for T<:Union{Float16,Float32,Float64} see twiceprecision.jl
 _colon(::Ordered, ::ArithmeticRounds, start::T, step, stop::T) where {T} =
-    StepRangeLen(start, step, floor(Int, (stop-start)/step)+1)
+    StepRangeLen(start, step, floor(Integer, (stop-start)/step)+1)
 _colon(::Any, ::Any, start::T, step, stop::T) where {T} =
-    StepRangeLen(start, step, floor(Int, (stop-start)/step)+1)
+    StepRangeLen(start, step, floor(Integer, (stop-start)/step)+1)
 
 """
     (:)(start, [step], stop)
@@ -415,10 +415,11 @@ oneto(r) = OneTo(r)
 ## Step ranges parameterized by length
 
 """
-    StepRangeLen{T,R,S}(ref::R, step::S, len, [offset=1]) where {T,R,S}
-    StepRangeLen(       ref::R, step::S, len, [offset=1]) where {  R,S}
+    StepRangeLen(         ref::R, step::S, len, [offset=1]) where {  R,S}
+    StepRangeLen{T,R,S}(  ref::R, step::S, len, [offset=1]) where {T,R,S}
+    StepRangeLen{T,R,S,L}(ref::R, step::S, len, [offset=1]) where {T,R,S,L}
 
-A range `r` where `r[i]` produces values of type `T` (in the second
+A range `r` where `r[i]` produces values of type `T` (in the first
 form, `T` is deduced automatically), parameterized by a `ref`erence
 value, a `step`, and the `len`gth. By default `ref` is the starting
 value `r[1]`, but alternatively you can supply it as the value of
@@ -426,40 +427,45 @@ value `r[1]`, but alternatively you can supply it as the value of
 with `TwicePrecision` this can be used to implement ranges that are
 free of roundoff error.
 """
-struct StepRangeLen{T,R,S} <: AbstractRange{T}
+struct StepRangeLen{T,R,S,L<:Integer} <: AbstractRange{T}
     ref::R       # reference value (might be smallest-magnitude value in the range)
     step::S      # step value
-    len::Int     # length of the range
-    offset::Int  # the index of ref
+    len::L       # length of the range
+    offset::L    # the index of ref
 
-    function StepRangeLen{T,R,S}(ref::R, step::S, len::Integer, offset::Integer = 1) where {T,R,S}
+    function StepRangeLen{T,R,S,L}(ref::R, step::S, len::Integer, offset::Integer = 1) where {T,R,S,L}
         if T <: Integer && !isinteger(ref + step)
             throw(ArgumentError("StepRangeLen{<:Integer} cannot have non-integer step"))
         end
-        len >= 0 || throw(ArgumentError("length cannot be negative, got $len"))
-        1 <= offset <= max(1,len) || throw(ArgumentError("StepRangeLen: offset must be in [1,$len], got $offset"))
-        new(ref, step, len, offset)
+        len = convert(L, len)
+        len >= zero(len) || throw(ArgumentError("length cannot be negative, got $len"))
+        offset = convert(L, offset)
+        L1 = oneunit(typeof(len))
+        L1 <= offset <= max(L1, len) || throw(ArgumentError("StepRangeLen: offset must be in [1,$len], got $offset"))
+        return new(ref, step, len, offset)
     end
 end
 
+StepRangeLen{T,R,S}(ref::R, step::S, len::Integer, offset::Integer = 1) where {T,R,S} =
+    StepRangeLen{T,R,S,promote_type(Int,typeof(len))}(ref, step, len, offset)
 StepRangeLen(ref::R, step::S, len::Integer, offset::Integer = 1) where {R,S} =
-    StepRangeLen{typeof(ref+zero(step)),R,S}(ref, step, len, offset)
+    StepRangeLen{typeof(ref+zero(step)),R,S,promote_type(Int,typeof(len))}(ref, step, len, offset)
 StepRangeLen{T}(ref::R, step::S, len::Integer, offset::Integer = 1) where {T,R,S} =
-    StepRangeLen{T,R,S}(ref, step, len, offset)
+    StepRangeLen{T,R,S,promote_type(Int,typeof(len))}(ref, step, len, offset)
 
 ## range with computed step
 
 """
-    LinRange{T}
+    LinRange{T,L}
 
 A range with `len` linearly spaced elements between its `start` and `stop`.
 The size of the spacing is controlled by `len`, which must
-be an `Int`.
+be an `Integer`.
 
 # Examples
 ```jldoctest
 julia> LinRange(1.5, 5.5, 9)
-9-element LinRange{Float64}:
+9-element LinRange{Float64, Int64}:
  1.5,2.0,2.5,3.0,3.5,4.0,4.5,5.0,5.5
 ```
 
@@ -483,26 +489,35 @@ julia> collect(LinRange(-0.1, 0.3, 5))
   0.3
 ```
 """
-struct LinRange{T} <: AbstractRange{T}
+struct LinRange{T,L<:Integer} <: AbstractRange{T}
     start::T
     stop::T
-    len::Int
-    lendiv::Int
+    len::L
+    lendiv::L
 
-    function LinRange{T}(start,stop,len) where T
+    function LinRange{T,L}(start::T, stop::T, len::L) where {T,L<:Integer}
         len >= 0 || throw(ArgumentError("range($start, stop=$stop, length=$len): negative length"))
-        if len == 1
+        onelen = oneunit(typeof(len))
+        if len == onelen
             start == stop || throw(ArgumentError("range($start, stop=$stop, length=$len): endpoints differ"))
-            return new(start, stop, 1, 1)
+            return new(start, stop, len, len)
         end
-        lendiv = max(len-1, 1)
+        lendiv = max(len - onelen, onelen)
         if T <: Integer && !iszero(mod(stop-start, lendiv))
             throw(ArgumentError("LinRange{<:Integer} cannot have non-integer step"))
         end
-        new(start,stop,len,lendiv)
+        return new(start, stop, len, lendiv)
     end
 end
 
+function LinRange{T,L}(start, stop, len::Integer) where {T,L}
+    LinRange{T,L}(convert(T, start), convert(T, stop), convert(L, len))
+end
+
+function LinRange{T}(start, stop, len::Integer) where T
+    LinRange{T,promote_type(Int,typeof(len))}(start, stop, len)
+end
+
 function LinRange(start, stop, len::Integer)
     T = typeof((stop-start)/len)
     LinRange{T}(start, stop, len)
@@ -621,6 +636,7 @@ step(r::StepRangeLen) = r.step
 step(r::StepRangeLen{T}) where {T<:AbstractFloat} = T(r.step)
 step(r::LinRange) = (last(r)-first(r))/r.lendiv
 
+# high-precision step
 step_hp(r::StepRangeLen) = r.step
 step_hp(r::AbstractRange) = step(r)
 
@@ -648,7 +664,7 @@ function checked_length(r::OrdinalRange{T}) where T
         diff = checked_sub(stop, start)
     end
     a = Integer(div(diff, s))
-    return checked_add(a, one(a))
+    return checked_add(a, oneunit(a))
 end
 
 function checked_length(r::AbstractUnitRange{T}) where T
@@ -657,7 +673,7 @@ function checked_length(r::AbstractUnitRange{T}) where T
         return Integer(first(r) - first(r))
     end
     a = Integer(checked_add(checked_sub(last(r), first(r))))
-    return checked_add(a, one(a))
+    return checked_add(a, oneunit(a))
 end
 
 function length(r::OrdinalRange{T}) where T
@@ -675,14 +691,14 @@ function length(r::OrdinalRange{T}) where T
         diff = stop - start
     end
     a = Integer(div(diff, s))
-    return a + one(a)
+    return a + oneunit(a)
 end
 
 
 function length(r::AbstractUnitRange{T}) where T
     @_inline_meta
     a = Integer(last(r) - first(r)) # even when isempty, by construction (with overflow)
-    return a + one(a)
+    return a + oneunit(a)
 end
 
 length(r::OneTo) = Integer(r.stop - zero(r.stop))
@@ -710,7 +726,7 @@ let bigints = Union{Int, UInt, Int64, UInt64, Int128, UInt128}
         else
             a = div(unsigned(diff), s) % typeof(diff)
         end
-        return Integer(a) + one(a)
+        return Integer(a) + oneunit(a)
     end
     function checked_length(r::OrdinalRange{T}) where T<:bigints
         s = step(r)
@@ -729,7 +745,7 @@ let bigints = Union{Int, UInt, Int64, UInt64, Int128, UInt128}
         else
             a = div(checked_sub(start, stop), -s)
         end
-        return checked_add(a, one(a))
+        return checked_add(a, oneunit(a))
     end
 end
 
@@ -803,10 +819,11 @@ copy(r::AbstractRange) = r
 
 ## iteration
 
-function iterate(r::Union{LinRange,StepRangeLen}, i::Int=1)
+function iterate(r::Union{StepRangeLen,LinRange}, i::Integer=zero(length(r)))
     @_inline_meta
+    i += oneunit(i)
     length(r) < i && return nothing
-    unsafe_getindex(r, i), i + 1
+    unsafe_getindex(r, i), i
 end
 
 iterate(r::OrdinalRange) = isempty(r) ? nothing : (first(r), first(r))
@@ -897,7 +914,7 @@ function getindex(r::AbstractUnitRange, s::AbstractUnitRange{T}) where {T<:Integ
     @boundscheck checkbounds(r, s)
 
     if T === Bool
-        range(first(s) ? first(r) : last(r), length = Int(last(s)))
+        range(first(s) ? first(r) : last(r), length = Integer(last(s)))
     else
         f = first(r)
         st = oftype(f, f + first(s)-firstindex(r))
@@ -916,7 +933,7 @@ function getindex(r::AbstractUnitRange, s::StepRange{T}) where {T<:Integer}
     @boundscheck checkbounds(r, s)
 
     if T === Bool
-        range(first(s) ? first(r) : last(r), step=oneunit(eltype(r)), length = Int(last(s)))
+        range(first(s) ? first(r) : last(r), step=oneunit(eltype(r)), length = Integer(last(s)))
     else
         st = oftype(first(r), first(r) + s.start-firstindex(r))
         return range(st, step=step(s), length=length(s))
@@ -949,24 +966,29 @@ function getindex(r::StepRangeLen{T}, s::OrdinalRange{S}) where {T, S<:Integer}
     @_inline_meta
     @boundscheck checkbounds(r, s)
 
+    len = length(s)
+    sstep = step_hp(s)
+    rstep = step_hp(r)
+    L = typeof(len)
     if S === Bool
-        if length(s) == 0
-            return StepRangeLen{T}(first(r), step(r), 0, 1)
-        elseif length(s) == 1
+        rstep *= one(sstep)
+        if len == 0
+            return StepRangeLen{T}(first(r), rstep, zero(L), oneunit(L))
+        elseif len == 1
             if first(s)
-                return StepRangeLen{T}(first(r), step(r), 1, 1)
+                return StepRangeLen{T}(first(r), rstep, oneunit(L), oneunit(L))
             else
-                return StepRangeLen{T}(first(r), step(r), 0, 1)
+                return StepRangeLen{T}(first(r), rstep, zero(L), oneunit(L))
             end
-        else # length(s) == 2
-            return StepRangeLen{T}(last(r), step(r), 1, 1)
+        else # len == 2
+            return StepRangeLen{T}(last(r), rstep, oneunit(L), oneunit(L))
         end
     else
         # Find closest approach to offset by s
         ind = LinearIndices(s)
-        offset = max(min(1 + round(Int, (r.offset - first(s))/step(s)), last(ind)), first(ind))
-        ref = _getindex_hiprec(r, first(s) + (offset-1)*step(s))
-        return StepRangeLen{T}(ref, r.step*step(s), length(s), offset)
+        offset = L(max(min(1 + round(L, (r.offset - first(s))/sstep), last(ind)), first(ind)))
+        ref = _getindex_hiprec(r, first(s) + (offset-1)*sstep)
+        return StepRangeLen{T}(ref, rstep*sstep, len, offset)
     end
 end
 
@@ -974,22 +996,24 @@ function getindex(r::LinRange{T}, s::OrdinalRange{S}) where {T, S<:Integer}
     @_inline_meta
     @boundscheck checkbounds(r, s)
 
+    len = length(s)
+    L = typeof(len)
     if S === Bool
-        if length(s) == 0
-            return LinRange(first(r), first(r), 0)
-        elseif length(s) == 1
+        if len == 0
+            return LinRange{T}(first(r), first(r), len)
+        elseif len == 1
             if first(s)
-                return LinRange(first(r), first(r), 1)
+                return LinRange{T}(first(r), first(r), len)
             else
-                return LinRange(first(r), first(r), 0)
+                return LinRange{T}(first(r), first(r), zero(L))
             end
         else # length(s) == 2
-            return LinRange(last(r), last(r), 1)
+            return LinRange{T}(last(r), last(r), oneunit(L))
         end
     else
         vfirst = unsafe_getindex(r, first(s))
         vlast  = unsafe_getindex(r, last(s))
-        return LinRange{T}(vfirst, vlast, length(s))
+        return LinRange{T}(vfirst, vlast, len)
     end
 end
 
@@ -1158,8 +1182,8 @@ issubset(r::AbstractUnitRange{<:Integer}, s::AbstractUnitRange{<:Integer}) =
 ## linear operations on ranges ##
 
 -(r::OrdinalRange) = range(-first(r), step=-step(r), length=length(r))
--(r::StepRangeLen{T,R,S}) where {T,R,S} =
-    StepRangeLen{T,R,S}(-r.ref, -r.step, length(r), r.offset)
+-(r::StepRangeLen{T,R,S,L}) where {T,R,S,L} =
+    StepRangeLen{T,R,S,L}(-r.ref, -r.step, r.len, r.offset)
 function -(r::LinRange)
     start = -r.start
     LinRange{typeof(start)}(start, -r.stop, length(r))
@@ -1173,12 +1197,12 @@ el_same(::Type{T}, a::Type{<:AbstractArray{S,n}}, b::Type{<:AbstractArray{T,n}})
 el_same(::Type, a, b) = promote_typejoin(a, b)
 
 promote_rule(a::Type{UnitRange{T1}}, b::Type{UnitRange{T2}}) where {T1,T2} =
-    el_same(promote_type(T1,T2), a, b)
+    el_same(promote_type(T1, T2), a, b)
 UnitRange{T}(r::UnitRange{T}) where {T<:Real} = r
 UnitRange{T}(r::UnitRange) where {T<:Real} = UnitRange{T}(r.start, r.stop)
 
 promote_rule(a::Type{OneTo{T1}}, b::Type{OneTo{T2}}) where {T1,T2} =
-    el_same(promote_type(T1,T2), a, b)
+    el_same(promote_type(T1, T2), a, b)
 OneTo{T}(r::OneTo{T}) where {T<:Integer} = r
 OneTo{T}(r::OneTo) where {T<:Integer} = OneTo{T}(r.stop)
 
@@ -1196,11 +1220,11 @@ OrdinalRange{T1, T2}(r::AbstractUnitRange{T1}) where {T1, T2<:Integer} = r
 OrdinalRange{T1, T2}(r::UnitRange) where {T1, T2<:Integer} = UnitRange{T1}(r)
 OrdinalRange{T1, T2}(r::OneTo) where {T1, T2<:Integer} = OneTo{T1}(r)
 
-promote_rule(::Type{StepRange{T1a,T1b}}, ::Type{StepRange{T2a,T2b}}) where {T1a,T1b,T2a,T2b} =
-    el_same(promote_type(T1a,T2a),
-            # el_same only operates on array element type, so just promote second type parameter
-            StepRange{T1a, promote_type(T1b,T2b)},
-            StepRange{T2a, promote_type(T1b,T2b)})
+function promote_rule(::Type{StepRange{T1a,T1b}}, ::Type{StepRange{T2a,T2b}}) where {T1a,T1b,T2a,T2b}
+    Tb = promote_type(T1b, T2b)
+    # el_same only operates on array element type, so just promote second type parameter
+    el_same(promote_type(T1a, T2a), StepRange{T1a,Tb}, StepRange{T2a,Tb})
+end
 StepRange{T1,T2}(r::StepRange{T1,T2}) where {T1,T2} = r
 
 promote_rule(a::Type{StepRange{T1a,T1b}}, ::Type{UR}) where {T1a,T1b,UR<:AbstractUnitRange} =
@@ -1211,35 +1235,38 @@ StepRange(r::AbstractUnitRange{T}) where {T} =
     StepRange{T,T}(first(r), step(r), last(r))
 (StepRange{T1,T2} where T1)(r::AbstractRange) where {T2} = StepRange{eltype(r),T2}(r)
 
-promote_rule(::Type{StepRangeLen{T1,R1,S1}},::Type{StepRangeLen{T2,R2,S2}}) where {T1,T2,R1,R2,S1,S2} =
-    el_same(promote_type(T1,T2),
-            StepRangeLen{T1,promote_type(R1,R2),promote_type(S1,S2)},
-            StepRangeLen{T2,promote_type(R1,R2),promote_type(S1,S2)})
-StepRangeLen{T,R,S}(r::StepRangeLen{T,R,S}) where {T,R,S} = r
-StepRangeLen{T,R,S}(r::StepRangeLen) where {T,R,S} =
-    StepRangeLen{T,R,S}(convert(R, r.ref), convert(S, r.step), length(r), r.offset)
+function promote_rule(::Type{StepRangeLen{T1,R1,S1,L1}},::Type{StepRangeLen{T2,R2,S2,L2}}) where {T1,T2,R1,R2,S1,S2,L1,L2}
+    R, S, L = promote_type(R1, R2), promote_type(S1, S2), promote_type(L1, L2)
+    el_same(promote_type(T1, T2), StepRangeLen{T1,R,S,L}, StepRangeLen{T2,R,S,L})
+end
+StepRangeLen{T,R,S,L}(r::StepRangeLen{T,R,S,L}) where {T,R,S,L} = r
+StepRangeLen{T,R,S,L}(r::StepRangeLen) where {T,R,S,L} =
+    StepRangeLen{T,R,S,L}(convert(R, r.ref), convert(S, r.step), convert(L, r.len), convert(L, r.offset))
 StepRangeLen{T}(r::StepRangeLen) where {T} =
-    StepRangeLen(convert(T, r.ref), convert(T, r.step), length(r), r.offset)
+    StepRangeLen(convert(T, r.ref), convert(T, r.step), r.len, r.offset)
 
-promote_rule(a::Type{StepRangeLen{T,R,S}}, ::Type{OR}) where {T,R,S,OR<:AbstractRange} =
-    promote_rule(a, StepRangeLen{eltype(OR), eltype(OR), eltype(OR)})
-StepRangeLen{T,R,S}(r::AbstractRange) where {T,R,S} =
-    StepRangeLen{T,R,S}(R(first(r)), S(step(r)), length(r))
+promote_rule(a::Type{StepRangeLen{T,R,S,L}}, ::Type{OR}) where {T,R,S,L,OR<:AbstractRange} =
+    promote_rule(a, StepRangeLen{eltype(OR), eltype(OR), eltype(OR), Int})
+StepRangeLen{T,R,S,L}(r::AbstractRange) where {T,R,S,L} =
+    StepRangeLen{T,R,S,L}(R(first(r)), S(step(r)), length(r))
 StepRangeLen{T}(r::AbstractRange) where {T} =
     StepRangeLen(T(first(r)), T(step(r)), length(r))
 StepRangeLen(r::AbstractRange) = StepRangeLen{eltype(r)}(r)
 
-promote_rule(a::Type{LinRange{T1}}, b::Type{LinRange{T2}}) where {T1,T2} =
-    el_same(promote_type(T1,T2), a, b)
-LinRange{T}(r::LinRange{T}) where {T} = r
-LinRange{T}(r::AbstractRange) where {T} = LinRange{T}(first(r), last(r), length(r))
+function promote_rule(a::Type{LinRange{T1,L1}}, b::Type{LinRange{T2,L2}}) where {T1,T2,L1,L2}
+    L = promote_type(L1, L2)
+    el_same(promote_type(T1, T2), LinRange{T1,L}, LinRange{T2,L})
+end
+LinRange{T,L}(r::LinRange{T,L}) where {T,L} = r
+LinRange{T,L}(r::AbstractRange) where {T,L} = LinRange{T,L}(first(r), last(r), length(r))
+LinRange{T}(r::AbstractRange) where {T} = LinRange{T,typeof(length(r))}(first(r), last(r), length(r))
 LinRange(r::AbstractRange{T}) where {T} = LinRange{T}(r)
 
-promote_rule(a::Type{LinRange{T}}, ::Type{OR}) where {T,OR<:OrdinalRange} =
-    promote_rule(a, LinRange{eltype(OR)})
+promote_rule(a::Type{LinRange{T,L}}, ::Type{OR}) where {T,L,OR<:OrdinalRange} =
+    promote_rule(a, LinRange{eltype(OR),L})
 
-promote_rule(::Type{LinRange{L}}, b::Type{StepRangeLen{T,R,S}}) where {L,T,R,S} =
-    promote_rule(StepRangeLen{L,L,L}, b)
+promote_rule(::Type{LinRange{A,L}}, b::Type{StepRangeLen{T2,R2,S2,L2}}) where {A,L,T2,R2,S2,L2} =
+    promote_rule(StepRangeLen{A,A,A,L}, b)
 
 ## concatenation ##
 
@@ -1266,9 +1293,9 @@ function _reverse(r::StepRangeLen, ::Colon)
     # invalid. As `reverse(r)` is also empty, any offset would work so we keep
     # `r.offset`
     offset = isempty(r) ? r.offset : length(r)-r.offset+1
-    StepRangeLen(r.ref, -r.step, length(r), offset)
+    return typeof(r)(r.ref, -r.step, length(r), offset)
 end
-_reverse(r::LinRange{T}, ::Colon) where {T} = LinRange{T}(r.stop, r.start, length(r))
+_reverse(r::LinRange{T}, ::Colon) where {T} = typeof(r)(r.stop, r.start, length(r))
 
 ## sorting ##
 
diff --git a/base/twiceprecision.jl b/base/twiceprecision.jl
index 55cdc59371674..a7ff116b56b36 100644
--- a/base/twiceprecision.jl
+++ b/base/twiceprecision.jl
@@ -194,6 +194,10 @@ function TwicePrecision{T}(x) where {T}
     TwicePrecision{T}(xT, T(Δx))
 end
 
+function TwicePrecision{T}(x::TwicePrecision) where {T}
+    TwicePrecision{T}(x.hi, x.lo)
+end
+
 TwicePrecision{T}(i::Integer) where {T<:AbstractFloat} =
     TwicePrecision{T}(canonicalize2(splitprec(T, i)...)...)
 
@@ -207,7 +211,7 @@ end
 
 function TwicePrecision{T}(nd::Tuple{Any,Any}) where {T}
     n, d = nd
-    TwicePrecision{T}(n) / d
+    TwicePrecision{T}(TwicePrecision{T}(n) / d)
 end
 
 function TwicePrecision{T}(nd::Tuple{I,I}, nb::Integer) where {T,I}
@@ -329,13 +333,13 @@ function steprangelen_hp(::Type{Float64}, ref::Tuple{Integer,Integer},
                          step::Tuple{Integer,Integer}, nb::Integer,
                          len::Integer, offset::Integer)
     StepRangeLen(TwicePrecision{Float64}(ref),
-                 TwicePrecision{Float64}(step, nb), Int(len), offset)
+                 TwicePrecision{Float64}(step, nb), len, offset)
 end
 
 function steprangelen_hp(::Type{T}, ref::Tuple{Integer,Integer},
                          step::Tuple{Integer,Integer}, nb::Integer,
                          len::Integer, offset::Integer) where {T<:IEEEFloat}
-    StepRangeLen{T}(ref[1]/ref[2], step[1]/step[2], Int(len), offset)
+    StepRangeLen{T}(ref[1]/ref[2], step[1]/step[2], len, offset)
 end
 
 # AbstractFloat constructors (can supply a single number or a 2-tuple
@@ -347,14 +351,13 @@ function steprangelen_hp(::Type{Float64}, ref::F_or_FF,
                          step::F_or_FF, nb::Integer,
                          len::Integer, offset::Integer)
     StepRangeLen(TwicePrecision{Float64}(ref...),
-                 twiceprecision(TwicePrecision{Float64}(step...), nb), Int(len), offset)
+                 twiceprecision(TwicePrecision{Float64}(step...), nb), len, offset)
 end
 
 function steprangelen_hp(::Type{T}, ref::F_or_FF,
                          step::F_or_FF, nb::Integer,
                          len::Integer, offset::Integer) where {T<:IEEEFloat}
-    StepRangeLen{T}(asF64(ref),
-                    asF64(step), Int(len), offset)
+    StepRangeLen{T}(asF64(ref), asF64(step), len, offset)
 end
 
 
@@ -365,30 +368,33 @@ StepRangeLen(ref::TwicePrecision{T}, step::TwicePrecision{T},
 
 # Construct range for rational start=start_n/den, step=step_n/den
 function floatrange(::Type{T}, start_n::Integer, step_n::Integer, len::Integer, den::Integer) where T
+    len = len + 0 # promote with Int
     if len < 2 || step_n == 0
-        return steprangelen_hp(T, (start_n, den), (step_n, den), 0, Int(len), 1)
+        return steprangelen_hp(T, (start_n, den), (step_n, den), 0, len, oneunit(len))
     end
     # index of smallest-magnitude value
-    imin = clamp(round(Int, -start_n/step_n+1), 1, Int(len))
+    L = typeof(len)
+    imin = clamp(round(typeof(len), -start_n/step_n+1), oneunit(L), len)
     # Compute smallest-magnitude element to 2x precision
     ref_n = start_n+(imin-1)*step_n  # this shouldn't overflow, so don't check
     nb = nbitslen(T, len, imin)
-    steprangelen_hp(T, (ref_n, den), (step_n, den), nb, Int(len), imin)
+    steprangelen_hp(T, (ref_n, den), (step_n, den), nb, len, imin)
 end
 
 function floatrange(a::AbstractFloat, st::AbstractFloat, len::Real, divisor::AbstractFloat)
+    len = len + 0 # promote with Int
     T = promote_type(typeof(a), typeof(st), typeof(divisor))
     m = maxintfloat(T, Int)
     if abs(a) <= m && abs(st) <= m && abs(divisor) <= m
         ia, ist, idivisor = round(Int, a), round(Int, st), round(Int, divisor)
         if ia == a && ist == st && idivisor == divisor
             # We can return the high-precision range
-            return floatrange(T, ia, ist, Int(len), idivisor)
+            return floatrange(T, ia, ist, len, idivisor)
         end
     end
     # Fallback (misses the opportunity to set offset different from 1,
     # but otherwise this is still high-precision)
-    steprangelen_hp(T, (a,divisor), (st,divisor), nbitslen(T, len, 1), Int(len), 1)
+    steprangelen_hp(T, (a,divisor), (st,divisor), nbitslen(T, len, 1), len, oneunit(len))
 end
 
 function (:)(start::T, step::T, stop::T) where T<:Union{Float16,Float32,Float64}
@@ -407,7 +413,7 @@ function (:)(start::T, step::T, stop::T) where T<:Union{Float16,Float32,Float64}
                     rem(den, start_d) == 0 && rem(den, step_d) == 0      # check lcm overflow
                 start_n = round(Int, start*den)
                 step_n = round(Int, step*den)
-                len = max(0, div(den*stop_n - stop_d*start_n + step_n*stop_d, step_n*stop_d))
+                len = max(0, Int(div(den*stop_n - stop_d*start_n + step_n*stop_d, step_n*stop_d)))
                 # Integer ops could overflow, so check that this makes sense
                 if isbetween(start, start + (len-1)*step, stop + step/2) &&
                         !isbetween(start, start + len*step, stop)
@@ -418,6 +424,7 @@ function (:)(start::T, step::T, stop::T) where T<:Union{Float16,Float32,Float64}
         end
     end
     # Fallback, taking start and step literally
+    # n.b. we use Int as the default length type for IEEEFloats
     lf = (stop-start)/step
     if lf < 0
         len = 0
@@ -436,6 +443,7 @@ step(r::StepRangeLen{T,TwicePrecision{T},TwicePrecision{T}}) where {T<:AbstractF
 step(r::StepRangeLen{T,TwicePrecision{T},TwicePrecision{T}}) where {T} = T(r.step)
 
 function range_start_step_length(a::T, st::T, len::Integer) where T<:Union{Float16,Float32,Float64}
+    len = len + 0 # promote with Int
     start_n, start_d = rat(a)
     step_n, step_d = rat(st)
     if start_d != 0 && step_d != 0 &&
@@ -474,31 +482,38 @@ end
 
 function getindex(r::StepRangeLen{T,<:TwicePrecision,<:TwicePrecision}, s::OrdinalRange{S}) where {T, S<:Integer}
     @boundscheck checkbounds(r, s)
+    len = length(s)
+    L = typeof(len)
+    sstep = step_hp(s)
+    rstep = step_hp(r)
     if S === Bool
-        if length(s) == 0
-            return StepRangeLen(r.ref, r.step, 0, 1)
-        elseif length(s) == 1
+        #rstep *= one(sstep)
+        if len == 0
+            return StepRangeLen{T}(first(r), rstep, zero(L), oneunit(L))
+        elseif len == 1
             if first(s)
-                return StepRangeLen(r.ref, r.step, 1, 1)
+                return StepRangeLen{T}(first(r), rstep, oneunit(L), oneunit(L))
             else
-                return StepRangeLen(r.ref, r.step, 0, 1)
+                return StepRangeLen{T}(first(r), rstep, zero(L), oneunit(L))
             end
-        else # length(s) == 2
-            return StepRangeLen(r[2], step(r), 1, 1)
+        else # len == 2
+            return StepRangeLen{T}(last(r), step(r), oneunit(L), oneunit(L))
         end
     else
-        soffset = 1 + round(Int, (r.offset - first(s))/step(s))
-        soffset = clamp(soffset, 1, length(s))
-        ioffset = first(s) + (soffset-1)*step(s)
-        if step(s) == 1 || length(s) < 2
-            newstep = r.step
+        soffset = round(L, (r.offset - first(s))/sstep + 1)
+        soffset = clamp(soffset, oneunit(L), len)
+        ioffset = L(first(s) + (soffset - oneunit(L)) * sstep)
+        if sstep == 1 || len < 2
+            newstep = rstep #* one(sstep)
         else
-            newstep = twiceprecision(r.step*step(s), nbitslen(T, length(s), soffset))
+            newstep = rstep * sstep
+            newstep = twiceprecision(newstep, nbitslen(T, len, soffset))
         end
+        soffset = max(oneunit(L), soffset)
         if ioffset == r.offset
-            return StepRangeLen(r.ref, newstep, length(s), max(1,soffset))
+            return StepRangeLen{T}(r.ref, newstep, len, soffset)
         else
-            return StepRangeLen(r.ref + (ioffset-r.offset)*r.step, newstep, length(s), max(1,soffset))
+            return StepRangeLen{T}(r.ref + (ioffset-r.offset)*rstep, newstep, len, soffset)
         end
     end
 end
@@ -509,30 +524,30 @@ end
 /(r::StepRangeLen{<:Real,<:TwicePrecision}, x::Real) =
     StepRangeLen(r.ref/x, twiceprecision(r.step/x, nbitslen(r)), length(r), r.offset)
 
-StepRangeLen{T,R,S}(r::StepRangeLen{T,R,S}) where {T<:AbstractFloat,R<:TwicePrecision,S<:TwicePrecision} = r
+StepRangeLen{T,R,S,L}(r::StepRangeLen{T,R,S,L}) where {T<:AbstractFloat,R<:TwicePrecision,S<:TwicePrecision,L} = r
 
-StepRangeLen{T,R,S}(r::StepRangeLen) where {T<:AbstractFloat,R<:TwicePrecision,S<:TwicePrecision} =
-    _convertSRL(StepRangeLen{T,R,S}, r)
+StepRangeLen{T,R,S,L}(r::StepRangeLen) where {T<:AbstractFloat,R<:TwicePrecision,S<:TwicePrecision,L} =
+    _convertSRL(StepRangeLen{T,R,S,L}, r)
 
 StepRangeLen{Float64}(r::StepRangeLen) =
-    _convertSRL(StepRangeLen{Float64,TwicePrecision{Float64},TwicePrecision{Float64}}, r)
+    _convertSRL(StepRangeLen{Float64,TwicePrecision{Float64},TwicePrecision{Float64},Int}, r)
 StepRangeLen{T}(r::StepRangeLen) where {T<:IEEEFloat} =
-    _convertSRL(StepRangeLen{T,Float64,Float64}, r)
+    _convertSRL(StepRangeLen{T,Float64,Float64,Int}, r)
 
 StepRangeLen{Float64}(r::AbstractRange) =
-    _convertSRL(StepRangeLen{Float64,TwicePrecision{Float64},TwicePrecision{Float64}}, r)
+    _convertSRL(StepRangeLen{Float64,TwicePrecision{Float64},TwicePrecision{Float64},Int}, r)
 StepRangeLen{T}(r::AbstractRange) where {T<:IEEEFloat} =
-    _convertSRL(StepRangeLen{T,Float64,Float64}, r)
+    _convertSRL(StepRangeLen{T,Float64,Float64,Int}, r)
 
-function _convertSRL(::Type{StepRangeLen{T,R,S}}, r::StepRangeLen{<:Integer}) where {T,R,S}
-    StepRangeLen{T,R,S}(R(r.ref), S(r.step), length(r), r.offset)
+function _convertSRL(::Type{StepRangeLen{T,R,S,L}}, r::StepRangeLen{<:Integer}) where {T,R,S,L}
+    StepRangeLen{T,R,S,L}(R(r.ref), S(r.step), L(length(r)), L(r.offset))
 end
 
-function _convertSRL(::Type{StepRangeLen{T,R,S}}, r::AbstractRange{<:Integer}) where {T,R,S}
-    StepRangeLen{T,R,S}(R(first(r)), S(step(r)), length(r))
+function _convertSRL(::Type{StepRangeLen{T,R,S,L}}, r::AbstractRange{<:Integer}) where {T,R,S,L}
+    StepRangeLen{T,R,S,L}(R(first(r)), S(step(r)), L(length(r)))
 end
 
-function _convertSRL(::Type{StepRangeLen{T,R,S}}, r::AbstractRange{U}) where {T,R,S,U}
+function _convertSRL(::Type{StepRangeLen{T,R,S,L}}, r::AbstractRange{U}) where {T,R,S,L,U}
     # if start and step have a rational approximation in the old type,
     # then we transfer that rational approximation to the new type
     f, s = first(r), step(r)
@@ -546,17 +561,17 @@ function _convertSRL(::Type{StepRangeLen{T,R,S}}, r::AbstractRange{U}) where {T,
                 rem(den, start_d) == 0 && rem(den, step_d) == 0
             start_n = round(Int, f*den)
             step_n = round(Int, s*den)
-            return floatrange(T, start_n, step_n, length(r), den)
+            return floatrange(T, start_n, step_n, L(length(r)), den)
         end
     end
-    __convertSRL(StepRangeLen{T,R,S}, r)
+    return __convertSRL(StepRangeLen{T,R,S,L}, r)
 end
 
-function __convertSRL(::Type{StepRangeLen{T,R,S}}, r::StepRangeLen{U}) where {T,R,S,U}
-    StepRangeLen{T,R,S}(R(r.ref), S(r.step), length(r), r.offset)
+function __convertSRL(::Type{StepRangeLen{T,R,S,L}}, r::StepRangeLen{U}) where {T,R,S,L,U}
+    StepRangeLen{T,R,S,L}(R(r.ref), S(r.step), L(length(r)), L(r.offset))
 end
-function __convertSRL(::Type{StepRangeLen{T,R,S}}, r::AbstractRange{U}) where {T,R,S,U}
-    StepRangeLen{T,R,S}(R(first(r)), S(step(r)), length(r))
+function __convertSRL(::Type{StepRangeLen{T,R,S,L}}, r::AbstractRange{U}) where {T,R,S,L,U}
+    StepRangeLen{T,R,S,L}(R(first(r)), S(step(r)), L(length(r)))
 end
 
 function sum(r::StepRangeLen)
@@ -567,7 +582,7 @@ function sum(r::StepRangeLen)
     np, nn = l - r.offset, r.offset - 1  # positive, negative
     # To prevent overflow in sum(1:n), multiply its factors by the step
     sp, sn = sumpair(np), sumpair(nn)
-    W = widen(Int)
+    W = widen(typeof(l))
     Δn = W(sp[1]) * W(sp[2]) - W(sn[1]) * W(sn[2])
     s = r.step * Δn
     # Add in contributions of ref
@@ -603,19 +618,20 @@ function +(r1::StepRangeLen{T,R}, r2::StepRangeLen{T,R}) where T where R<:TwiceP
         imid = r1.offset
         ref = r1.ref + r2.ref
     else
-        imid = round(Int, (r1.offset+r2.offset)/2)
+        imid = round(typeof(len), (r1.offset+r2.offset)/2)
         ref1mid = _getindex_hiprec(r1, imid)
         ref2mid = _getindex_hiprec(r2, imid)
         ref = ref1mid + ref2mid
     end
     step = twiceprecision(r1.step + r2.step, nbitslen(T, len, imid))
-    StepRangeLen{T,typeof(ref),typeof(step)}(ref, step, len, imid)
+    StepRangeLen{T,typeof(ref),typeof(step),typeof(len)}(ref, step, len, imid)
 end
 
 ## LinRange
 
 # For Float16, Float32, and Float64, this returns a StepRangeLen
 function range_start_stop_length(start::T, stop::T, len::Integer) where {T<:IEEEFloat}
+    len = len + 0 # promote with Int
     len < 2 && return _linspace1(T, start, stop, len)
     if start == stop
         return steprangelen_hp(T, start, zero(T), 0, len, 1)
@@ -638,32 +654,35 @@ function range_start_stop_length(start::T, stop::T, len::Integer) where {T<:IEEE
 end
 
 function _linspace(start::T, stop::T, len::Integer) where {T<:IEEEFloat}
+    len = len + 0 # promote with Int
     (isfinite(start) && isfinite(stop)) || throw(ArgumentError("start and stop must be finite, got $start and $stop"))
     # Find the index that returns the smallest-magnitude element
     Δ, Δfac = stop-start, 1
     if !isfinite(Δ)   # handle overflow for large endpoints
-        Δ, Δfac = stop/len - start/len, Int(len)
+        Δ, Δfac = stop/len - start/len, len
     end
     tmin = -(start/Δ)/Δfac            # t such that (1-t)*start + t*stop == 0
-    imin = round(Int, tmin*(len-1)+1) # index approximately corresponding to t
+    L = typeof(len)
+    lenn1 = len - oneunit(L)
+    imin = round(L, tmin*lenn1 + 1) # index approximately corresponding to t
     if 1 < imin < len
         # The smallest-magnitude element is in the interior
-        t = (imin-1)/(len-1)
+        t = (imin - 1)/lenn1
         ref = T((1-t)*start + t*stop)
         step = imin-1 < len-imin ? (ref-start)/(imin-1) : (stop-ref)/(len-imin)
     elseif imin <= 1
-        imin = 1
+        imin = oneunit(L)
         ref = start
-        step = (Δ/(len-1))*Δfac
+        step = (Δ/(lenn1))*Δfac
     else
-        imin = Int(len)
+        imin = len
         ref = stop
-        step = (Δ/(len-1))*Δfac
+        step = (Δ/(lenn1))*Δfac
     end
     if len == 2 && !isfinite(step)
         # For very large endpoints where step overflows, exploit the
         # split-representation to handle the overflow
-        return steprangelen_hp(T, start, (-start, stop), 0, 2, 1)
+        return steprangelen_hp(T, start, (-start, stop), 0, len, oneunit(L))
     end
     # 2x calculations to get high precision endpoint matching while also
     # preventing overflow in ref_hi+(i-offset)*step_hi
@@ -676,23 +695,28 @@ function _linspace(start::T, stop::T, len::Integer) where {T<:IEEEFloat}
     a, b = (start - x1_hi) - x1_lo, (stop - x2_hi) - x2_lo
     step_lo = (b - a)/(len - 1)
     ref_lo = a - (1 - imin)*step_lo
-    steprangelen_hp(T, (ref, ref_lo), (step_hi, step_lo), 0, Int(len), imin)
+    steprangelen_hp(T, (ref, ref_lo), (step_hi, step_lo), 0, len, imin)
 end
 
 # range for rational numbers, start = start_n/den, stop = stop_n/den
 # Note this returns a StepRangeLen
 _linspace(::Type{T}, start::Integer, stop::Integer, len::Integer) where {T<:IEEEFloat} = _linspace(T, start, stop, len, one(start))
 function _linspace(::Type{T}, start_n::Integer, stop_n::Integer, len::Integer, den::Integer) where T<:IEEEFloat
+    len = len + 0 # promote with Int
     len < 2 && return _linspace1(T, start_n/den, stop_n/den, len)
-    start_n == stop_n && return steprangelen_hp(T, (start_n, den), (zero(start_n), den), 0, len, 1)
+    L = typeof(len)
+    start_n == stop_n && return steprangelen_hp(T, (start_n, den), (zero(start_n), den), 0, len, oneunit(L))
     tmin = -start_n/(Float64(stop_n) - Float64(start_n))
-    imin = round(Int, tmin*(len-1)+1)
-    imin = clamp(imin, 1, Int(len))
-    ref_num = Int128(len-imin) * start_n + Int128(imin-1) * stop_n
-    ref_denom = Int128(len-1) * den
+    imin = round(typeof(len), tmin*(len-1)+1)
+    imin = clamp(imin, oneunit(L), len)
+    W = widen(L)
+    start_n = W(start_n)
+    stop_n = W(stop_n)
+    ref_num = W(len-imin) * start_n + W(imin-1) * stop_n
+    ref_denom = W(len-1) * den
     ref = (ref_num, ref_denom)
-    step_full = (Int128(stop_n) - Int128(start_n), ref_denom)
-    steprangelen_hp(T, ref, step_full,  nbitslen(T, len, imin), Int(len), imin)
+    step_full = (stop_n - start_n, ref_denom)
+    steprangelen_hp(T, ref, step_full, nbitslen(T, len, imin), len, imin)
 end
 
 # For len < 2
@@ -704,7 +728,7 @@ function _linspace1(::Type{T}, start, stop, len::Integer) where T<:IEEEFloat
         # The output type must be consistent with steprangelen_hp
         if T<:Union{Float32,Float16}
             return StepRangeLen{T}(Float64(start), Float64(start) - Float64(stop), len, 1)
-        else
+        else # T == Float64
             return StepRangeLen(TwicePrecision(start, zero(T)), TwicePrecision(start, -stop), len, 1)
         end
     end
@@ -713,8 +737,8 @@ end
 
 ### Numeric utilities
 
-# Approximate x with a rational representation. Guaranteed to return,
-# but not guaranteed to return a precise answer.
+# Approximate x with a rational representation as a pair of Int values.
+# Guaranteed to return, but not guaranteed to return a precise answer.
 # https://en.wikipedia.org/wiki/Continued_fraction#Best_rational_approximations
 function rat(x)
     y = x
@@ -722,7 +746,7 @@ function rat(x)
     b = c = 0
     m = maxintfloat(narrow(typeof(x)), Int)
     while abs(y) <= m
-        f = trunc(Int,y)
+        f = trunc(Int, y)
         y -= f
         a, c = f*a + c, a
         b, d = f*b + d, b
diff --git a/test/ranges.jl b/test/ranges.jl
index ad8554d14ea4a..d4ff83b81fe2d 100644
--- a/test/ranges.jl
+++ b/test/ranges.jl
@@ -652,8 +652,8 @@ end
     @test broadcast(-, T(1):2:6, 0.3) === T(1)-0.3:2:5-0.3
     is_unsigned = T <: Unsigned
     is_unsigned && @test length(broadcast(-, T(1):3, 2)) === length(T(1)-2:T(3)-2)
-    @test broadcast(-, T(1):3) == -T(1):-1:-T(3) broken=is_unsigned
-    @test broadcast(-, 2, T(1):3) == T(1):-1:-T(1) broken=is_unsigned
+    @test broadcast(-, T(1):3) == -T(1):-T(1):-T(3)
+    @test broadcast(-, 2, T(1):3) == T(1):-T(1):-T(1)
 end
 @testset "operations between ranges and arrays" for T in (Int, UInt, Int128)
     @test all(([T(1):5;] + (T(5):-1:1)) .=== T(6))
@@ -1108,10 +1108,11 @@ end
     # repr/show should display the range nicely
     # to test print_range in range.jl
     replrepr(x) = repr("text/plain", x; context=IOContext(stdout, :limit=>true, :displaysize=>(24, 80)))
+    nb = Sys.WORD_SIZE
     @test replrepr(1:4) == "1:4"
     @test repr("text/plain", 1:4) == "1:4"
     @test repr("text/plain", range(1, stop=5, length=7)) == "1.0:0.6666666666666666:5.0"
-    @test repr("text/plain", LinRange{Float64}(1,5,7)) == "7-element LinRange{Float64}:\n 1.0,1.66667,2.33333,3.0,3.66667,4.33333,5.0"
+    @test repr("text/plain", LinRange{Float64}(1,5,7)) == "7-element LinRange{Float64, Int$nb}:\n 1.0,1.66667,2.33333,3.0,3.66667,4.33333,5.0"
     @test repr(range(1, stop=5, length=7)) == "1.0:0.6666666666666666:5.0"
     @test repr(LinRange{Float64}(1,5,7)) == "range(1.0, stop=5.0, length=7)"
     @test replrepr(0:100.) == "0.0:1.0:100.0"
@@ -1119,7 +1120,7 @@ end
     # only examines spacing of the left and right edges of the range, sufficient
     # to cover the designated screen size.
     @test replrepr(range(0, stop=100, length=10000)) == "0.0:0.010001000100010001:100.0"
-    @test replrepr(LinRange{Float64}(0,100, 10000)) == "10000-element LinRange{Float64}:\n 0.0,0.010001,0.020002,0.030003,0.040004,…,99.95,99.96,99.97,99.98,99.99,100.0"
+    @test replrepr(LinRange{Float64}(0,100, 10000)) == "10000-element LinRange{Float64, Int$nb}:\n 0.0,0.010001,0.020002,0.030003,0.040004,…,99.95,99.96,99.97,99.98,99.99,100.0"
 
     @test sprint(show, UnitRange(1, 2)) == "1:2"
     @test sprint(show, StepRange(1, 2, 5)) == "1:2:5"
@@ -2076,3 +2077,7 @@ end
     @test r[0:2] == r[0]:r[2]
     @test r[0:1:2] == r[0]:1:r[2]
 end
+
+@test length(range(1, 100, length=big(100)^100)) == big(100)^100
+@test length(range(big(1), big(100)^100, length=big(100)^100)) == big(100)^100
+@test length(0 * (1:big(100)^100)) == big(100)^100