maria-developers team mailing list archive

[Alexander Barkov <bar@xxxxxxxxxxx>]

Hi Sergei,


On 08/08/2016 03:17 PM, Sergei Golubchik wrote:
> Hi, Alexander!
>
> On Aug 03, Alexander Barkov wrote:
> > diff --git a/sql/item_sum.cc b/sql/item_sum.cc
> > index adf48f6..03600b7 100644
> > --- a/sql/item_sum.cc
> > +++ b/sql/item_sum.cc
> > @@ -1762,7 +1762,7 @@ double Item_sum_std::val_real()
> >   {
> >     DBUG_ASSERT(fixed == 1);
> >     double nr= Item_sum_variance::val_real();
> > -  DBUG_ASSERT(nr >= 0.0);
> > +  DBUG_ASSERT(nr >= 0.0 || my_isinf(nr));
> >     return sqrt(nr);
> >   }
>
> is it ok to do sqrt(inf)? what will it be? 0? inf? nan?
> may be this should return NULL in inf case?

sqrt(inf) returns inf.

man sqrt:

If x is positive infinity, positive infinity is returned.

'inf' is further replaced to '0' in Item::send().



Note, SUM in a similar situation (i.e. overflow) also returns 0:

SELECT SUM(f) FROM (SELECT 1.7976931348623157e+308 AS f UNION SELECT 
1e308 AS f) t2;

Item_sum_sum::val_real() returns 0, which is replaced to '0' by 
Item::send().



> Regards,
> Sergei
> Chief Architect MariaDB
> and security@xxxxxxxxxxx