2011/9/16 Kenta Murata <muraken / gmail.com>:
> On Friday, September 16, 2011 at 02:14 , Marc-Andre Lafortune wrote:
>> The patch has been committed as r33282
>
> I cannot find tests for the commit r33282. Please tell me where the tests=
 are.
> If you write the tests in RubySpec, please describe the commit hashes of =
the corresponding
> commits of RubySpec in the commit message of the commit of CRuby.
>
>> Would it be ok if `Float("3.0e-31").to_s =3D=3D "3.0000000000000003e-31"=
`?
>
> I've objected to the behavior. Please look at the rejected my report.
> http://redmine.ruby-lang.org/issues/4656
>
> I think it is not a bug, but is a specification problem.
> The issue is related to what is to_s.

I reverted r33282 because
* Ruby keeps it as platform dependent or simply use -ffloat-store (/fp:prec=
ise)
* amd64 won't get this issue because compilers for amd64 uses SSE2 to
  calculate floating point numbers instead of x87 FPU.
* this change won't fix the issue under -O
* this commit has no test for the changed behavior

--=20
NARUSE, Yui =A0<naruse / airemix.jp>