[icedtea-web] not closeable javaws (and maybe more)

[Jie Kang]

----- Original Message -----
> On 02/17/2015 11:57 AM, Jiri Vanek wrote:
> > On 02/17/2015 11:00 AM, Jiri Vanek wrote:
> >> Hi!
> >>
> >> Head, for already some time javaws apps needs to be closed by signal kill.
> >> For long time I thought this is soem accident or whatever, but it is not.
> >> Today I tracked it to guilty changeset of
> >> http://icedtea.classpath.org/hg/icedtea-web/rev/6f4c1d501560
> >>
> >> Looking for the fix now, but in case I fail, this will spare some time to
> >> the followr,
> >>
> >>   J.
> >
> > Ok, so the caus eis clear, ServiceExecutor do not run as daemon, and so is
> > preventing jvm to stop.
> >
> > I'm wondering Why I dont see this issue from plugin side...
> >
> > Anyway - for javaws - both ok and failed run - where to place call to
> >
> > diff -r a8baec8d9d21 netx/net/sourceforge/jnlp/cache/ResourceTracker.java
> > --- a/netx/net/sourceforge/jnlp/cache/ResourceTracker.java    Fri Feb 13
> > 12:48:24 2015 +0100
> > +++ b/netx/net/sourceforge/jnlp/cache/ResourceTracker.java    Tue Feb 17
> > 11:57:01 2015 +0100
> > @@ -34,6 +34,7 @@
> >   import java.util.List;
> >   import java.util.concurrent.ExecutorService;
> >   import java.util.concurrent.Executors;
> > +import java.util.concurrent.TimeUnit;
> >
> >   import net.sourceforge.jnlp.DownloadOptions;
> >   import net.sourceforge.jnlp.Version;
> > @@ -633,4 +634,15 @@
> >       interface Filter<T> {
> >           public boolean test(T t);
> >       }
> > +
> > +    public static void shutdDownThreadPool() throws InterruptedException{
> > +        threadPool.shutdown();
> > +        if (!threadPool.awaitTermination(5, TimeUnit.SECONDS)) {
> > +            OutputController.getLogger().log("Executor did not terminate
> > in the specified time.");
> > +            List<Runnable> droppedTasks = threadPool.shutdownNow();
> > +            OutputController.getLogger().log("Executor was abruptly shut
> > down. " + droppedTasks.size() + " tasks will not be executed.");
> > +        }
> > +    }
> > +
> > +
> >   }
> >
> >
> > ?
> 
> One more thought, attached patch fixed the issue for me:
> 
> It does only thing - it changes threads used by ExecutorService to daemons.
> 
> Well, I dont know whot Iahve done - I dont know if ExecutorService is still
> keeping its purpose when its threads are daemons.
> 
> On contrary, documentation is silent, and usage of custom ThreadFactory is
> supported....
> As negativism, original code (before patch 6f4c1d501560) was not using
> daemons. But was not recycling threads...
> 
> 
> Thoughts?

Hello,

I strongly prefer a solution that uses threadPool.shutdown() since it is the proper way to stop the service and generally much more predictable/safer solution.

How about using a ResourceTracker Factory to create ResourceTrackers and keep a list of them there. Then you can call the shutdown on all the ResourceTrackers when IT-W exits.


Regards,


> 
> J.
> 
> 
> 

-- 

Jie Kang

OpenJDK Team - Software Engineering Intern