We have been running into a problem in Java3D for over a year now, and
we are looking for a better solution to the problem than we are
currently using. The
problem is this: we have a database of polygons that make up the
earth's surface, but just for land, not for the oceans. There is no
easy way using the data to
separate out the coastline, so we can't really generate the water
ourselves.
So, what we want to do is render a high-resolution blue sphere, and then
render the land on top of the sphere. The two parts are coplanar (there
is no
elevation data in our dataset), so we clearly need to do something to
get around the Z-fighting problem. We have tried a variety of things,
but none of them
seem to work:
1.The correct solution would be to use a DecalGroup (since that is
what it was created for in the first place!!!) Unfortunately, there is
a known bug in
Java3D (BugId 4476109) that messes up the depth buffer when using a
DecalGroup on Nvidia cards. We
submitted this bug report a year ago, and it has not yet been
resolved.
2.Our second attempt was to use an OrderedGroup instead of a
DecalGroup, setting setDepthBufferEnable(false) and
setDepthBufferWriteEnable(false). This had the exact same problem
as the DecalGroup attempt.
3.The third attempt was to use the polygonOffset to try to bump the
Z-buffer value of the land to prevent Z-fighting. This was CLOSE to
working, but
there was still Z-fighting in certain spots. And besides, the
Java3D archives are full of messages saying "be careful with
polygonOffset. It works
differently on different graphics cards"
4.The only way we were able to get it working (and I use that term
loosely) is to set a transparency attribute on EVERY object in the
scene. So, we are
effectively disabling the depth buffer and rendering every object
using alpha blending. Which means that you have some objects rendered
in front of
other objects even though they should be behind them, and even the
objects themselves are sometimes drawn backwards, with parts of the
object
incorrectly in front of other parts.
So, what I'm looking for is (a) a brilliant solution to the problem, (b)
a kludgy solution that works better than what we have, or at least (c) a
better feel for
when (and more importantly, IF) the bug mentioned above will ever be
fixed.
Thanks,
-Roy Duffy
[EMAIL PROTECTED]
===========================================================================
To unsubscribe, send email to [EMAIL PROTECTED] and include in the body
of the message "signoff JAVA3D-INTEREST". For general help, send email to
[EMAIL PROTECTED] and include in the body of the message "help".
