Hi, No, I don't really mind using the "long way". It is just that I was surprised that the casts were commented out, since I kinda liked the way that they mirrored shared_ptrs casting functions. Using those casts are slightly more legible than doing it the long way. But if you feel that they don't belong in the library, then I accept your judgement. Lars Ion Gaztañaga wrote:
Lars Hagström escribió:
Yes, but it has to be safer and more correct to use static_pointer_casts instead of reinterpret_cast. In my particular situation I am unfortunately not able to use templates, so I do need to do "ugly" pointer stuff. (I am porting an existing shared-memory solution to boost::interprocess).
I've uncommented the static_pointer_cast et al functions from offset_ptr.hpp, and they appear to do their work beautifully.
Lars
Sorry for the late reply. Originally I wrote static_pointer_cast & co. functions to mimic shared_ptr. I even used them as generic casting function for code where the pointer type is templated.
However, those functions need an explicit template argument so ADL was not activated (say, like when you use an unqualified swap call). In the end I decided to comment them to prevent users using them (I should have erased them...).
You can achieve the same using:
offset_ptr(static_cast<MyType*>(my_offset_ptr.get()));
without any performance hit, so I didn't consider them essential (ok, you have to type more ;-)). Is there any reason why you prefer static_pointer_cast instead of the code above? I have no problem to uncomment the code if you find it useful.
Regards,
Ion _______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users