Re: [gl4java-usergroup] Java Binding Mess . . . or . . . Why GL4Java AND JoGL currently SUCK

[Kenneth B. R. <kbr...@al...>]

> There are 2 killer bugs right now for the use of GL4Java in medical
> imaging software.  They are, the inability to safely add and remove
> GL enabled canvases from a parent, and the inability to work on
> multiple heterogeneous monitors.  These are not problems in the java
> side code of GL4Java,
> the problem is in the actual JVM source code.  My company has the
> source code to the JRE.  We can compare it to the SWT source.  We
> even see the problem.  My boss is done waiting though.  So we made
> our own binding under SWT.

Could you please provide more details on where exactly you think
the bug is in the JRE?

> JoGL STILL has the add and remove bug.  Although to their credit,
> the code is FAR more clean, and GlueGen is something I think EVERY
> developer should be looking at right now, and not just for GL.  I did,
> however, have hopes that since Ken was one of the developers the
> internal machinery at Sun or JavaSoft would fix those problems.  They
> did not.

We didn't have time to stress test this before JavaOne. Now that
that rush is over we can do so. As I've mentioned in another mail
to this mailing list, there was already some amount of planning
ahead in the Jogl sources for this case; see roughly line 156 in
src/net/java/games/jogl/impl/windows/WindowsOnscreenGLContext.java
and roughly line 142 in
src/net/java/games/jogl/impl/x11/X11OnscreenGLContext.java, in
particular the checking for JAWT_LOCK_SURFACE_CHANGED. I don't
know whether that will be sufficient; in particular I think if a
canvas is being animated and it's removed from its parent it will
probably start returning JAWT_LOCK_ERROR, which will cause the
GLException a few lines earlier to be thrown.

I think we might need to override something like dispose() in
GLCanvas to get earlier notifications that the component has been
removed from the hierarchy; I'll try to ask the AWT developers
about this this week. (I work on the JVM and am not an AWT
expert.)

> I hope JoGL or GL4Java will rectify this insane situation in
> the very near future.  If you need my help or our binding code,
> just let me know

Yes, please do help us out. Jogl is a community effort which is
why it's open source on a collaborative site. If you can provide
a test case illustrating an error, help fix the bug, show where
in the JRE any underlying bug is, or any or all of the above,
we'd appreciate your help.

-Ken