Re: [boost] [variant] Important: alignment issue

From: Douglas Gregor (gregod_at_[hidden])
Date: 2003-11-30 10:13:28

On Sunday 30 November 2003 02:21 am, Eric Friedman wrote:
> My belief is that doing so would be unnecessary, and I know that it
> would result in larger variants. On the other hand, if the problem is
> *not* Borland-specific, then it would be critically important to have
> this fix.

I'ts not unheard of that Borland has a problem with 
alignment_of/type_with_alignment. Both have special implementations for 
Borland, including this little gem:

#ifdef __BORLANDC__
// long double gives an incorrect value of 10 (!)
// unless we do this...
struct long_double_wrapper{ long double ld; };
template<> struct alignment_of<long double>
   : public alignment_of<long_double_wrapper>{};
#endif

I guess it's fine just to use the maximally-aligned type for Borland for now 
(we should revert the change _immediately_ on the main trunk once we have 
branched for release), but we should not do this for other compilers.

        Doug