undefined behavior in GMP 5.1.2

[Vincent Lefevre]

On 2013-09-23 10:05:35 +0200, Vincent Lefevre wrote:
> On 2013-09-23 07:31:05 +0200, Marc Glisse wrote:
> > On Mon, 23 Sep 2013, Vincent Lefevre wrote:
> > 
> > >On 2013-09-21 11:57:42 +0200, Marc Glisse wrote:
> > >>>>mul.c:81:31: runtime error: signed integer overflow:
> > >>>>-5260204364771764878 + -5260204364771764878 cannot be represented in
> > >>>>type 'long'
> > >>
> > >>We don't check for underflow in mpf multiplication, I didn't touch that.
> > >>People should use mpfr ;-)
> > >
> > >Shouldn't the test be removed since it is known to be buggy?
> > 
> > I didn't study the test closely, I was waiting to see if someone had an
> > opinion on it.
> 
> Actually the test triggering the integer overflow is
> 
>     { "%10.2Ff",    "0.25", "      0.25" },
> 
> If there's an underflow in mpf there, that seems to be a real bug.

This is not due to underflow. This one is due to the call to
mpn_get_d with a negative exp, so that the bug is the same as
the one already reported.

-- 
Vincent Lefèvre <vincent at vinc17.net> - Web: <http://www.vinc17.net/>
100% accessible validated (X)HTML - Blog: <http://www.vinc17.net/blog/>
Work: CR INRIA - computer arithmetic / AriC project (LIP, ENS-Lyon)