[PATCH, i386]: Fix target/35944, [4.3, 4.4 regression] wrong result for MOD with kind=10 for some array argument values

[Uros Bizjak]

FX wrote:

> Here are two testcases that should cover both the real(kind=10) and
> also the standard float and double floating point types:
>   

FX, thanks again for your help!

I have committed following patch in your name:

2008-04-16  Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>

        PR target/35944
        * gfortran.dg/pr35944-1.f90: New test.
        * gfortran.dg/pr35944-2.f90: Ditto.

The patch was regression tested on x86_64-pc-linux-gnu {,-m32}.

Thanks,
Uros.