Posted by gouessej on Oct 18, 2010; 7:10pm
URL: https://forum.jogamp.org/New-tutorial-for-JOGL-in-Eclipse-RCP-application-tp1670638p1727531.html

Wade Walker wrote

It's hard to be sure. With open-source you can never really tell what's going to work in the long term, unfortunately. The SWT GLCanvas was around long before GEF 3D (we've been using it at work since 2006 or so), so I'm not really worried that GEF 3D failing will affect it.

I see what you mean but I spoke to some members of the Eclipse Foundation and it is clear that the OpenGL binding of SWT is not actively maintained. I don't say that it has to be thrown into the trashbin but we have to be aware of this problem.

Wade Walker wrote

Since SWT is what all of Eclipse is based on, and I'm writing apps based on Eclipse RCP, it seemed like a good bet to use the SWT GLCanvas instead of the AWT one.

SWT is a good bet on my view, I write Eclipse RCP applications too but it does not mean that SWT GLCanvas is a good bet. I'm aware of the weakness of SWT, especially on Linux, I have to mix SWT and AWT in some situations including the drag and drop because of known bugs never fixed on Linux.

Wade Walker wrote

And for science/engineering apps like I write, I need to be able to count on all the window messages working correctly with the canvas inside an SWT application framework, which seemed more likely with SWT GLCanvas than with AWT or NEWT. But I haven't tried using those other options in a complex app -- they might be just as good or better than SWT.

JOGL GLCanvas is fast and reliable even in Eclipse RCP applications when using JOGL 1.1.1a (there are still locking and focus problems with JOGL 2.0), NEWT is a bit faster in some situations but there are some bugs on Linux and I don't know how it works on Mac OS X. The SWT_AWT helper is very useful and works really fine, I only had to add a fix for Linux that will be included in Java 1.7.

Wade Walker wrote

I can't claim that there'll never be a problem with the way I'm doing things in these tutorials. But I can say that I've used this way at work for about five years on a very complex engineering application, and I haven't run into an insurmountable problem so far

Did you measure the memory footprint?