Hi Mathias, This sounds rather like you've got a dangling pointer somewhere in your app. The OSG uses reference counting for all the internal scene graph objects which is specifically to avoid dangling pointers and to leverage this safely you are best to use ref_ptr<> to all OSG objects.
Lots has been discussed about ref_ptr<> over the years so have a search through the archives. Robert. On Thu, Feb 3, 2011 at 7:12 PM, Mathias Buhr <[email protected]> wrote: > Hi, > > I'm experiencing a strange segfault with an inherited osg::Drawable. > Unfortunately I'm currently unable to provide an example demonstrating this. > The application is a clustered rendering system and for various reasons I > have to (un-)load scenes during runtime. These scenes are implemented as > plugins (shared objects). The application implements an "unrolled" frame() > loop and (un-)loading these scenes happens after the call to > updateTraversal(). > The whole system is generally working but a newly created scene is somehow > problematic. The scenes consists only of a reimplemented osg::Drawable where > basically only drawImplementation() is overwritten. This drawable renders a > simple wire cube and display lists are disabled (so drawImplementation() is > called every frame). Rendering this scene works fine. > > Unloading and deletion of the shared object and its related data works > almost fine but the application crashes in unref() of a drawable during the > next cull traversal. Notice the "a" as I'm not entirely sure which drawable > is the problem (afaik there is only one). > The really strange thing is that the destructor of my drawable isn't called. > If I additionally call ref() on this object the segfault disappears but the > destructor still won't get called. The problem seems to be independent of > the threading mode. I've tried singlethreaded and could verify (via > debugger) that no threads where running during the removal of the sceneroot > of the shared object. > > How is it possible that the next cull traversal reaches this drawable when > it is removed from my scene root before the call to renderingTraversals()? > Is there anything obvious missing? > Is there anything happening in the rendering backend that I should be aware > of? > I haven't build OSG in debug mode but probably I'll try this next. Maybe > this will provide some useful insights. > > I appreciate any idea on this. > Best regards and thanks in advance > Mathias Buhr > _______________________________________________ > 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
