Hi JP, Hi Robert, 2009/5/19 Robert Osfield <[email protected]>
> Hi Vincent, > > Looking at J.P. follow up email it does sound like it could be windows > specific issue. Are you using windows? Yes Is you singleton implemented in a header? Yes.. > If so the just have the declaration in the header, and > keep the implementation entirely within the .cpp to ensure that the > allocation happens single dll. > > Robert. I thinks the problem is here : I've a dll and an exe implementing the same singleton (common sources but 2 compilations) ... so I get two instances, that sounds logical. Now I'm thinking a way to get only one ... maybe an other lib ? In OSG sources this is not a problem : the core is a lib, the plugin is a lib based on the core, and the application uses the core. so we can explain the calls just on one single line... but me I've got a dll using an other one, and the one in question implement the same code than the core using the dll ... I can explain it making a triangle, with the exe, the exe code in dll, and the dll... so the same code is implemented twice ... and so my singleton too... If you think of any solution, I would be very interested to hear it. Thanks. Regards, Vincent. > > > On Tue, May 19, 2009 at 3:33 PM, Vincent Bourdier > <[email protected]> wrote: > > Hi Robert, > > > > I get more info : the singleton is not destroyed, but is called from the > lib > > I did and from the exe itself. When the lib call the singleton, it works > > good. when the exe call the singleton, It does a new call to constructor > ... > > > > Any idea about how to solve that problem ? > > > > Thanks. > > > > Regards, > > Vincent. > > > > 2009/5/19 Robert Osfield <[email protected]> > >> > >> On Tue, May 19, 2009 at 2:24 PM, Vincent Bourdier > >> <[email protected]> wrote: > >> > Hi Robert, > >> > > >> > I think I get it : > >> > > >> > I seems to be due to the osgDB::DynamicLibrary instance I use, which > >> > ref_ptr > >> > unref when leaving the load code, so the destructor of the > >> > RegisterReaderWriterProxy (my equivalent) was called, and it seems to > be > >> > sufficient to force a new call to the Registry constructor. > >> > > >> > Does it sounds right for you ? > >> > >> I can't comment on code I'm not party to. > >> > >> As a general note one shouldn't be able to delete a internally > >> singleton used in my example without doing a instance()=0; Just take > >> a ref_ptr<> to the return won't delete the object as the instance() > >> method keeps around a copy. > >> > >> Robert. > >> _______________________________________________ > >> osg-users mailing list > >> [email protected] > >> > http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org > > > > > > _______________________________________________ > > osg-users mailing list > > [email protected] > > > http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org > > > > > _______________________________________________ > osg-users mailing list > [email protected] > http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org >
_______________________________________________ osg-users mailing list [email protected] http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
