Hi Chirs, I think it's a plausible story :-)
I have now merged and submitted the fix to svn/trunk. Robert. On 22 February 2013 16:33, Chris Hanson <[email protected]> wrote: > Well, remember, this isn't even my code. This is all core OSG code, in the > databasepager. So, I'm not claiming to be an expert in what it does, or was > intended to do. I can probably try to track down where it happened, roughly, > but I don't think I left breadcrumbs after finding the actual cause. > > This was an already valid LightPointNode that was loaded by the FLT loader, > called from the databasepager. For reasons I didn't totally understand, I > believe as part of osgUtil optmization, this loaded node was cloned prior to > being inserted into the live graph, and the clone was what actually ended up > in the graph. At this point, I can't recall if the bbox of the source node > was actually already computed and valid, or if in fact the problem was that > it was not, but that the copy constructor was not setting the max/min values > to FLT_MAX which would then trigger a later computation of the bbox. My > memory suggests it was the latter -- the bbox values were uninitialized and > therefore didn't trigger a recompute when they needed to. > > Does that make sense? > _______________________________________________ osg-submissions mailing list [email protected] http://lists.openscenegraph.org/listinfo.cgi/osg-submissions-openscenegraph.org
