From: Johan Holmberg <johan556@...>
Date: 2006-10-07T05:22:05+09:00
Subject: Re: Bug: right shift (>>) for Bignum?

Sorry for the mail here on Ruby-talk.
My intention was to post this on ruby-core,
where I suppose it belongs.

/Johan Holmberg


On 10/6/06, Johan Holmberg <johan556@gmail.com> wrote:
> Hi!
>
> I think I have stumbled on a bug in the right shift method on Bignums.
> An expression like
>
>     (1 - 2**i) >> (i+1)
>
> normally gives a value of -1 (i.e. minus one). But when the shift
> amount is near a multiple of 32 (or 64 on 64-bit architectures)
> it suddenly gives a value of 0 (i.e. zero) instead.
>
> A unit test demonstrating the problem is given below.
> On my IMac running OS-X I get an error when i=31.
> I am using Ruby 1.8.5.
>
> --------------------------------
> require "test/unit"
>
> class TC_bignum_rshift_bug < Test::Unit::TestCase
>   def test_rshift_bug
>     (1..100).each do |i|
>       assert_equal( -1,
>                     (1 - 2**i) >> (i+1),
>                     "shift steps: #{i}" )
>     end
>   end
> end
> --------------------------------
>
> Regards,
> Johan Holmberg
>
>