Currently when we initialize the OpenGL-based pipeline, the default GraphicsConfig is chosen based upon a selection algorithm that finds an X11 visual most appropriate for AWT and our X11-based pipeline (see
makeDefaultConfig() in awt_GraphicsEnv.c). In many cases, the visual selected
by this method is also a good choice for the default visual from which we
construct a GLXFBConfig. In other words, the chosen visual has a 24- or 32-bit
color buffer and a >= 1-bit stencil buffer, so it's suitable for using the
OGL pipeline.

However, this visual may not be the best, or most resource
efficient one from OGL's perspective. For example, it may have depth,
accumulation, multisample, or other auxiliary buffers that we don't use,
and therefore take up VRAM unnecessarily. Likewise, it may only be single-
buffered, even though apps that use BufferStrategy (including possibly Swing
in Mustang) perform best if there is a doublebuffered visual in use. Also,
sometimes the default visual chosen has no stencil buffer, and therefore we
can't use the OGL pipeline for the default GraphicsConfig.

It would be much better if we just avoided the existing selection method
and wrote one specifically tuned for finding the best visual for the OGL
pipeline. GLX offers a convenient mechanism for enumerating all the
FBConfigs (visuals) for a given screen and selecting the most appropriate one
for your application.
###@###.### 2004-12-10 00:06:02 GMT