numbers.jl fails bug Windows

[cosmoc]

Using 32-bit julia 0.2.0 on Windows 7 as installed from julia-58f8444d80

testing fails on numbers.jl

If the following lines are removed from numbers.jl the remaining lines pass. The failing tests are:

@test float64(int128(0)) == 0.0
@test float32(int128(0)) == 0.0f0
@test float64(int128(-1)) == -1.0
@test float32(int128(-1)) == -1.0f0
@test float64(int128(3)) == 3.0
@test float32(int128(3)) == 3.0f0
@test float64(uint128(10121)) == 10121.0
@test float32(uint128(10121)) == 10121.0f0
@test float64(typemin(Int128)) == -2.0^127
@test float32(typemin(Int128)) == -2.0f0^127
@test float64(typemax(Int128)) == 2.0^127
@test float32(typemax(Int128)) == 2.0f0^127
@test float64(typemin(Uint128)) == 0.0
@test float32(typemin(Uint128)) == 0.0f0
@test float64(typemax(Uint128)) == 2.0^128
@test float32(typemax(Uint128)) == 2.0f0^128

@test int128(-2.0^127) == typemin(Int128)
@test float64(uint128(3.7e19)) == 3.7e19
@test float64(uint128(3.7e30)) == 3.7e30

[pao: formatting to delink @test]

[JeffBezanson]

Possible dup of #3081? @vtjnash @loladiro

[Keno]

Quite probable. Can't say without booting up windows though.

[vtjnash]

no, that is a 64-bit only problem. I think this has already been fixed (by switching to BigInt) since it doesn't occur on master, only on the old binaries.