Re: shared_ptr implementation without template friends

From: gn_at_[hidden]
Date: 2000-08-01 04:59:07

> 
> The reinterpre_cast is ugly and causes undefined behaviour since
> shared_ptr 
> isn't a POD, but I'd bet it will work in practice.
> 
> What is proposed is bugware (ie, having the wrong behaviour 
(invoking
> undefined 
> behaviour) because someone else had the wrong behaviour (not 
handling 
> standard C++)). Bugware is generaly bad; here I think it is the 
wrong 
> solution because another simplier and working solution is already in
> use: 
> make everything public.
> 

Gee, what a lively discussion - I hadn't expected that :-)

I'm fully aware that my proposal is ugly, but the code would be 
encapsulated by a macro that effectivly stated that this code would 
only be compiled when using a non-conforming (and therefore already 
buggy) compiler. Plus, the external interface of the class would be 
the one we originally designed, without having to sacrifice 
encapsulation.

#ifdef _MSC_VER

:-) Now we already know which compiler we're targeting, so we can as 
well prove that the "undefined behaviour code" does produce 
predictable results with that one compiler. In my case, the 
reinterpret_cast could only fail if the compiler created different 
class layouts for different instantiations of "shared_ptr", which it 
doesn't do because the rest of the code alreaday guarantees 
compatibility between the pointer types.

#endif

Cheerio, Gernot