A DESCRIPTION OF THE REGRESSION :
1. Start the SwingSet2 demo.

2. Switch to the GTK LAF.

3. Point at a tab long enough to get a tooltip (or go to the cow example).

REPRODUCIBLE TESTCASE OR STEPS TO REPRODUCE:
Sun's SwingSet2 demo shows the problem as described above.

RELEASE LAST WORKED:
5.0 Update 6

RELEASE TEST FAILS:
mustang-b71

EXPECTED VERSUS ACTUAL BEHAVIOR :
EXPECTED -
I see a yellow tooltip background.
ACTUAL -
I see a gray tooltip background. (This is still the case in b77.)

OBSERVED APPLICATION IMPACT:
Trivial. And Java 6's GTK LAF is *way* better than Java 5's. Good enough that I plan to remove the code from various applications I'm responsible for that currently disable the GTK LAF. I have a bunch of slightly more significant GTK LAF bugs, but none are regressions, and I really wanted a t-shirt. (A trivial related example is that the borders on the GTK LAF's tooltips are wrong, but were wrong in Java 5 too.)

I think I deserve a t-shirt for the pain that http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6320676 has caused me, even if I wasn't first to report that regression. (And, as per my comment on that bug, it's a worse regression than the title implies. I'm sure it's easily skipped over in defect reviews given the title!)

I don't need an Ultra 20 because I already have one. But a t-shirt would be cool.

I promise not to bring it into disrepute like that guy wearing the "JavaOne" t-shirt in one of the "bumfights" DVDs.

Unless I need the money for food.