Fix rintf rounding (bug 2548)

[Joseph S. Myers]

On Wed, 22 Feb 2012, Andreas Jaeger wrote:

> The same code snippets you remove for flt-32 are alos in the dbl-64 code. 
> Are they really needed in that case?

No, the same reasoning applies to say that all the versions of this code 
can be simplified in the same way.  I've made a note to clean them up 
later (I'll file a bug for that if I don't get to it soon) - it's just 
that unlike the float case I'm not aware of an actual bug caused by the 
unnecessary code (but I didn't try to construct one beyond the testcases I 
added for double) and testing on more platforms (probably MIPS as well as 
x86_64) will be needed to cover all the different variants of the code.

> The patch itself looks fine to me,

Thanks, I've committed these two patches.

-- 
Joseph S. Myers
joseph@codesourcery.com