JSVGCanvas and other fixes

[Bill Haneman <Bi...@ireland.sun.com>]

Hi all:

I think that I am ready to make a judgement call and
commit my current set of changes to JSVGCanvas.

I have not convinced myself that it is 100% safe, but
I have been hammering on it pretty hard this afternoon
and the only errors it's throwing seem to occur when
the parser thread is interrupted, and those errors occur
very seldom (I only was able to make it happen twice).  
I trap them now and just throw an InterruptedException, which 
is perhaps a cheat but since it only occurs when loads are
interrupted it seems OK.

I have backgrounded the offscreen buffer repaint again,
more in line with Stephane's intentions.  As I said, I 
am not sure it's 100% bulletproof but it is certainly
much more robust than it was.  I justify the uncertainty by noting that
the viewer is now subjectively much more performant than it was
with the very conservative fixes of earlier today.

If anyone can make it crash in a threaded area (e.g.
stacktrace starts in run()) please send me the trace.
Also please give me your assessments of stability if you
think we should either back out the threading changes or
devote more time to checking them.

Thanks,

Bill

------
Bill Haneman +353 1 8199279
Ireland Desktop Engineering
Sun Microsystems Ireland Ltd.