|
Boost Users : |
Subject: Re: [Boost-users] [Serialization] shared_ptr to const object
From: Robert Ramey (ramey_at_[hidden])
Date: 2009-12-21 14:14:03
Steven Watanabe wrote:
> AMDG
>
> elizabeta petreska wrote:
>> I just want to say that I managed to solve this problem without
>> creating a temporary , thanks to this thread :
>> http://old.nabble.com/(de)serializing-map%3Cconst-A*%3E-to5538474.html#a5596607<http://old.nabble.com/%28de%29serializing-map%3Cconst-A*%3E-to5538474.html#a5596607>
>>
>> And here it is what I have got in my problem :
>>
>> boost::shared_ptr<const TT> jj(new TT(5));
>>
>> {
>> boost::archive::xml_oarchive oa(ofs);
>> oa & BOOST_SERIALIZATION_NVP(jj);
>> }
>>
>> boost::shared_ptr<const TT> rjj;
>>
>> {
>> void* xptr=&rjj;
>> boost::shared_ptr<TT> *xxptr;
>> xxptr=static_cast< boost::shared_ptr<TT> * > (xptr);
>> boost::archive::xml_iarchive ia(ifs);
>> ia & BOOST_SERIALIZATION_NVP(*xxptr);
>> }
>>
>
> This is undefined behavior. I would strongly advise against it
> if there is a reasonable alternative.
>
> In Christ,
> Steven Watanabe
Note that for better or worse, the serialization library does a lot of this
kind
of thing internally. If fact, a number of key facilities (e.g. export, usage
of DLLS, etc depend
on behavior which is technically undefined, but necessary to implement them.
Just
getting them to work at all was a huge challenge. I doubt that the above
is any more risky than many usages of the library itself.
Robert Ramey
Boost-users list run by williamkempf at hotmail.com, kalb at libertysoft.com, bjorn.karlsson at readsoft.com, gregod at cs.rpi.edu, wekempf at cox.net