an email from the mailing list:

On Sat, Jul 18, 2009 at 3:29 PM, Robert Hildebrandt<roberts_katz@gmx.de> wrote:
> Hi,
>
> Some Suggestions (for the case it's all nonsence: sorry - I'm a beginner)
> I spend two days wondering, why my program didn't worked, until I found out,
> that Gio::MemoryInputStream::add_data doesn't copy anything, and
> MemoryInputStream owns the pointer afterwards.
> I thing, that it could be useful for other users if this was documented (I'm
> using glibmm 2.4 Documentation (does anybody know why it's 2.4 and not 2.20
> like the current Version?))
> Also it could be nice if there was a possibility to define the
> GDestroyNotify Function without beeing forced to wrap a C Object
>

It seems that we wrap the add_data() with a std::string argument and a raw pointer argument, but the raw pointer one doesn't provide a way to specify a free DestroyNotify function nor does it document that we (apparently) free it ourselves.  The raw pointer overload should probably be deprecated and wrapped properly.