After a few more tests, I found out that the regression in quality shown in my previous message seems to come both from a lower antialiasing quality
and a smaller depth size. If, in the
getBestConfiguration method of
JoglPipeline class, I replace
caps.setNumSamples(2) call by
caps.setNumSamples(4) and
caps.setDepthBits(gct.getDepthSize()) call by
caps.setDepthBits(24), then I get similar results to Java3D 1.6.0-pre10.
I don't mind to keep
caps.setNumSamples(2) call unchanged (the quality difference isn't huge compared to
caps.setNumSamples(4)), but still, I wonder why
caps.setDepthBits(16) call (16 is the default value returned by
gct.getDepthSize()) leads to a different result under Mac OS X with Java3D 1.6.0-pre11. Was the requested value of
gct.getDepthSize() not respected in 1.6.0-pre10, and the made modification now respects the requested value?
Can I set depth size to 24 or 32 without any risk that the 3D configuration fails? Should I call
isGraphicsConfigSupported or just trust
getBestConfiguration implementation?
Thanks for your suggestions
Emmanuel Puybaret