I have solved the problem, but I don't understand why my solution
solved the problem.
There actually were several threads that were not being explicitly
shut down. The reason I didn't look at these as the source of the
problem is that several of them had been around for quite some time
in this code and had never caused problems. Also, none of them are
directly involved with the particular thread which was receiving this
ThreadDeath Error.
These "extra" threads are pretty simple. All but one of them simply
monitor some local caches, waiting to clear their particular cache at
some specified time.
The other thread is the one that was causing problems. This one had
also been around in one form or another for a long time as well, and
it wasn't causing any problems. It used to be a private inner class,
but is now its own class (significant? I can't imagine why.). Also,
it would only get started once (it was a static) and would die when
the applet was reloaded. This was undesirable so I changed the code
so that this thread would get re-created if it had died. This change
appears to be significant. But there is a second significant fact: It
makes a (quick) network connection to the server (only about once per
minute). But it had always done this, even when it hadn't been
causing problem. If I remove that network access, it no longer causes
problems, even with the thread being re-created as needed. Go figure.
To summarize:
1. A statically held thread. I didn't actually test this as a
necessary condition. I'm assuming it's necessary given the JVM's
weird behavior with applets and static references.
2. Gets re-created when it dies.
3. Performs network access at least once before dying and being re-
created.
4. Sun's plugin.
If (1 && 2 && 3 && 4) { Browser will freeze on applet shutdown; }
Btw, to answer Greg's specific questions. The two browsers I tested
were Safari and IE (on Windows). The problem only manifests with
Apple's or Sun's JVM. The Microsoft JVM did not have this problem.
So, as you alluded to, I would suspect the plugin as the "cause" of
the problem. Although it may be that the plugin is behaving more
correctly and is exposing some bug in my code. But I'm still unclear
exactly what that bug is.
Rob
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Java-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/java-dev/email@hidden
