Re: [boost] Re: SmartPtr (Loki) - auto_ptr/move c'tor issue

From: David Abrahams (dave_at_[hidden])
Date: 2003-01-30 23:12:23

"David B. Held" <dheld_at_[hidden]> writes:

> "David Abrahams" <dave_at_[hidden]> wrote in message
> news:uadhicf3d.fsf_at_boost-consulting.com...
>> [...]
>> What about the extra size due to the common compiler feature, "no
>> empty base optimization in the presence of MI?"  That was another
>> problem solved by using a chain of policy templates.
>> [...]
>
> True, but so far, it's only been an unsolvable problem on one compiler.
> Of course, if anyone wants to try out the code on something other than
> bcc 5.5, gcc 3.0, or vc 6.5, I would definitely like to see if it *is* a
> problem on other compilers.  

It will be a problem on gcc-2.95.x. Also ALL versions of Metrowerks,
vc++, Intel for windows, and any other compiler that is trying to be
object-layout-compatible with VC on windows (which is most of them).

Also many other compilers on other platforms have this issue, because
it wasn't seen as a high-priority optimization in the early days and
vendors are very reluctant to break object code compatibility.  It's
not every day you get to start over from scratch with a new ABI like
GCC did.

> And frankly, VC++ is beginning to annoy me greatly.  There's a large
> number of small features that have to be worked around or outright
> sacrificed for it.  I guess I'm curious to see if 7.1 has the same
> size problem, but I don't have access to that yet.

Of course it's object-code compatible, so object layout had to be
preserved.

-- 
                       David Abrahams
   dave_at_[hidden] * http://www.boost-consulting.com
Boost support, enhancements, training, and commercial distribution