Posted by gouessej on Mar 07, 2014; 10:14am
URL: https://forum.jogamp.org/Java3D-stereo-tp4029914p4031791.html

babor wrote

Finally, I am aware of this change in setGraphicsConfiguration in Java 7, but it should not influence stereo on Canvas3D as it does not influence stereo on GLCanvas in JOGL (and they both extend java.awt.Canvas).

Maybe it would be possible to write a dirty kludge to work around this known regression in AWT but I'm against this option because it isn't suitable on the long term, it would probably be difficult to maintain on all supported platforms (especially under Mac) and this bug should be fixed in AWT itself. Moreover, the fact that GLCanvas and Canvas3D both extend java.awt.Canvas doesn't mean that they don't work differently. Canvas3D was initially written not with JOGL in mind as Java3D had to stay compatible with several renderers. If you really want to avoid this kind of trouble in the future, rather use another scenegraph whose canvases and windows directly extend JOGL canvases. The poorly implemented Java3D canvases won't be rewritten and we won't provide some NEWT compatible Java3D canvases which means that you'll be concerned by all regressions in AWT affecting its canvases based on it or on Swing. Xerxes just did the right thing.

I imagine that this is typically not the kind of answer you would like to get but I still plan to write detailed migration guides to help developers to switch to more viable scenegraphs on the long term and I can help you. Some of us try to keep Java3D alive as is, without changing its public API but some of its design flaws prevent us from making it evolve without risking to break it. You might have to live with Java3D and undergo some remaining AWT bugs or to switch to another scenegraph (which is doable but not without any risk).