Arkady Arkright wrote:
Done some more tests. The problem (using FS) exists on the TAG grid, OSGRID and 3rd Rock - so it looks like it's common to all OpenSim derived grids. Furthermore, the sphere 'resolves' when the camera viewpoint is moved in any direction, so it looks as if OS grids in general are not fully rezzing child prims until the camera moves.
It doesn't make any sense for this to be server-side; we only have full updates and terse (movement, rotation) updates. When you first see the object, that's a full update. We only tell the viewer that it's a sphere and the dimensions, etc. It's not like the server sends down a mesh shape for the sphere. From the server side, either you should see the prim or not see it at all. Also we don't treat child prims any differently than root prims.
I'm not sure why you're seeing it fine on SL but not OpenSim grids. The camera movement does cause a recalc of which prims to show, but doesn't change (at the server end) how they are shown.
Perhaps the problem isn't related to the prims them all but something more to do with the initial camera position, or something like that. There could be a server-side difference in how the avatar is initialized, but I would think the viewer initializes the starting camera position, so there's nothing for the server to notify the viewer about at initialization time.
Regarding the child prim difference, if you take one of the "good" objects, shift-copy it, unlink the prims and relink them in the other order, does it become a "bad" object?
I think you've focused the problem enough now that I can do some tests and hopefully reproduce it.
When you first asked, I thought you were referring to the slightly
flattened sphere in your "good" image, which still isn't quite a smooth round sphere, just close enough and what we're used to for years. But the poorly resolved one is very