The following was submitted as part of an evaluation of an unrelated bug by one of our Russian team members, Denis Mikhalikin, <###@###.###>. Since it reports a possible memory leak, I am submitting it as a separate bug:

I found the chain that follows to static variable - appContext.
JFrame <- OffscreenImage <- RepaintManager <- Hashtable$Entry <-
Hashtable <- AppContext <- static appConext.
All classes, except AppContext, are in swing library.

I think this unfinalized static chain may lead to this memory leak.
When i cut this link by removing RepaintManager through
RepaintManager.setCurrentManager(null), the number of leaks is
reduced to zero and some times to 1 (not yet gc`ed, i think) .


OffscreenImage removed automaticaly via removeNotify if the component is

JComponent:
"
     // JComponent will inform us when it is no longer valid
     // (via removeNotify) we have no such hook to other components,
     // therefore we don't keep a ref to the Component
     // (indirectly through the Image) by stashing the image.
"

But JFrame is not JComponent.

I think the procedure suggested in swing code for removing
OffscreenImage is dangerous - it postpones the removing
until next call to getOffscreenBuffer - in most cases until
next redrawing. But when all J-components are
removed and no JComponent available, the link
would never be removed.

RepaintManager.setCurrentManager(null) removes OffscreenImage at once.
So i suggest to call this method in removeNotify - same as in JComponent

to remove link.