Re: [boost] [1.33.1][serialization] buginlibs\serialization\src\basic_iarchive.cpp and proposed patch

From: JOAQUIN LOPEZ MU?Z (joaquin_at_[hidden])
Date: 2005-09-02 13:16:48

Robert Ramey <ramey <at> rrsd.com> writes:

> 
> Joaquín Mª López Muñoz wrote:
> > The recent updates to Boost.Serialization code in RC_1_33_0
> > can provoke a crash at run-time, as shown for instance
> > in http://tinyurl.com/dwu78 (there are other, compile-time problems
> > that I'm not addressing in tihs post.)
> 
> What are these compile-time problems?

As I'm writting this the report engined seems to be stuck,
so I cannot give you an URL: CW 9.5 for Linux, for
instance, fails to build the lib complanining about
an ambiguous overload with state_saver (if my memory
serves me well.)

> 
> > The problem is located in reset_object_address. The attached
> > patch to libs\serialization\src\basic_iarchive.cpp fixes the
> > problem in my local tests, but as I lack the knowledge to
> > be sure, I'd rather not commit it myself and let it be confirmed
> > to be OK.
> 
> I've put the fix in my local system and will be testing it soon.  I don't 
> see any harm in it but I'm not quite sure it should be necessary either.

Look at it this way: before the updates to Boost.Serialization
code 34h ago, multi_index->test_serialization worked fine,
now it doesn't and I haven't changed anything in my code base, so...

Of course, the possibility exists that I'm using your lib in
a wrong way.

Also be aware that I don't really know what the patch is supposed
to be doing: I've just copypasted from the previous version of
reset_object_address implementation and verified that the
problem went away.

> I 
> looked at the files in multi-index and didn't find any reference to 
> reset_object_address so I'm still a little puzzled.  More information 
> please?
> 

In multi_index_container::load() (boost/multi_index_container.hpp)

The error shows as follows: when loading a previously saved
pointer to a tracked object, the returned pointer is bogus.
Funny thing is that the object pointed to is *not* the
one on which reset_object_address() is used. If you need
more details on the problem as I debugged it, please tell me
so.

Joaquín M López Muñoz
Telefónica, Investigación y Desarrollo


_______________________________________________
Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost