Re: [boost] Boost super_string

From: Jeff Garland (jeff_at_[hidden])
Date: 2006-10-13 08:20:19

Martin Bonner wrote:
> ----Original Message----
> From: boost-bounces_at_[hidden]
> [mailto:boost-bounces_at_[hidden]] On Behalf Of Caleb Epstein Sent:
> 12 October 2006 22:09 To: boost_at_[hidden]
> Subject: Re: [boost] Boost super_string
> 
>> On 10/12/06, Arkadiy Vertleyb <vertleyb_at_[hidden]> wrote:
>>
>>> More importantly, according
>>> to the standard, the basic_string class destructor is not virtual,
>>> and so basic_string is not intended for derivation.
>> Wouldn't this only matter if super_string added data members (it
>> doesn't AFAICT)
> Not relevent.

It is relevant. If you add data members you now want to ensure the subclass 
destructor runs in the case below.  Detailed completely in Effective C++ #14.

>> and was being deleted via pointers to
>> std::basic_string?
> 
> THIS is the relevent point.
> 
>    string* pstr = new super_string;
>    delete pstr;
> 
> is undefined behaviour.  In practise, you get away with it if
> super_string doesnt do anything in its destructor, 

Exactly.

> and if all additional
> members don't do anything in their destructors.  However, formally it is
> ALWAYS undefined behaviour.
> 
> The real question is how often will people create a super_string on the
> heap, and delete it through pointer to string.

No.  The intent is that the above new/delete is supported.  If it isn't then 
it's a really bad idea.

Jeff