Re: [boost] BigInt divide-by-zero handling

From: Sohail Somani (s.somani_at_[hidden])
Date: 2006-05-17 14:25:50

> -----Original Message-----
> From: boost-bounces_at_[hidden] 
> [mailto:boost-bounces_at_[hidden]] On Behalf Of Gerhard Wesp
> Sent: Wednesday, May 17, 2006 9:40 AM
> To: boost_at_[hidden]
> Subject: Re: [boost] BigInt divide-by-zero handling
> 
> On Wed, May 17, 2006 at 09:22:58AM -0700, Sohail Somani wrote:
> > Do an assertion in debug mode, U.B. in release. This is 
> most sensible
> > for a type that is an extension of a built-in type.
> 
> Seems OK also.  I just happen to think that whereever 
> behaviour *can* be
> defined, it should (and do it as the ints whereever it is defined what
> they do!).  There's already too much U.B. in C++.
> 
> A division by zero exception in an arbitrary size integer 
> package would
> not incur any measurable overhead, I think.

Heres another possible justification:

template<typename IntType>
T make_T(IntType a)
{
...
..b/a..
}

This function behaves differently for IntType=int or IntType=BigInt.