Swing Tool Tip Causes Incorrect GLContext

I am seeing an issue where Swing tool tips are causing JOGL to be unable to obtain a valid GLContext.  Before a tool tip is rendered, the proper GLContext is returned and everything works as expected.  Any time after a Swing tool tip is rendered, new GLContext creations will fall back to OpenGL 1.1.  This bug has been reproduced on multiple machines using ATI Radeon HD graphics cards, which are all I have available to test on.

I have included a very simple test case that reliably illustrates the problem tooltipdemo.java and the stack trace of the exception produced by JOGL tooltipstacktrace.txt.

Two things to note:

        1)  This bug does not appear if the "sun.java2d.d3d=false" (or older "sun.java2d.noddraw=true") value is not set, but everything I have read, including the JOGAMP webpage, says this flag is necessary to prevent other issues.  If I am mistaken, please correct me.
        2)  I was not able to reproduce this on Java 1.7.0_07 with everything else the exact same, so there may have been a Java update that contributes to this.

Tested Configurations:

        Windows 7
        ATI Radeon HD 5450 with older (2011) and latest graphics drivers
        Java 1.7.0_67, Java 1.7.0_75, and 1.8.0_40 (but could not be reproduced on 1.7.0_07)
        JOGL 2.1.5 (2.1-b1240-20140311) and 2.3.0 (2.3-b1378-20150311)

To replicate:

        1) Execute the main method in the attached java file tooltipdemo.java.
        2) Hover the mouse over the "Test" button until the tool tip appears.
        3) Click on the "Test" button to attempt to open a GLCanvas.

On 03/13/2015 09:37 PM, flem [via jogamp] wrote: This _was_ true for a long time, hmm.
However, I checked our test environment and jogl-demos,
and in most cases we test and 'demo' w/o any those 'sun.java2d' flags.
Maybe it is time to revisit the issue and squash the recommendation.

However, it would be nice to understand the issue a bit more.

Looking at your demo code, I assume that GLCanvas creation happens
on the AWT-EDT - well, your log file seems to support this.

Hmm, the mixing of Swing LW/HW often exposes issues of incompatibility.
Maybe time to remove above mentioned recommendation?
We would need to gather a bit more cross platform data maybe.

Thank you for your thorough report.

It would be great if you can put all this to a new bug report
and we can track more heuristics there - and come to a conclusion.

~Sven

Thanks for the quick response.  I'll try removing the sun.java2d flags and see if I run into any unexpected behaviors.

Per your request, I've added a bug report: https://jogamp.org/bugzilla/show_bug.cgi?id=1146.  Hopefully you can replicate it easily and figure out what is occurring.