FinalizableReferenceQueue still leaks

[gissuebot]

From gili.tzabari on December 25, 2008 12:51:41

Issue 227 is closed but the problem was not really fixed.

Stuart's comment #8 explains what remains to be fixed: https://code.google.com/p/google-guice/issues/detail?id=227&can=1&start=200#c8

Original issue: http://code.google.com/p/google-guice/issues/detail?id=288

[gissuebot]

From limpbizkit on December 30, 2008 14:03:08

I've started working on a fix, but it might not be worthwhile... http://groups.google.com/group/google-guice-dev/browse_thread/thread/28caab16fc75ad25

[gissuebot]

From limpbizkit on December 30, 2008 15:39:59

(No comment was entered for this change.)

Labels: Priority-High Milestone-Release2.0

[gissuebot]

From gili.tzabari on December 30, 2008 16:10:12

1. Is there a way to use SoftReferences without a dedicated thread (perhaps similar
   to how WeakHashMap works)?
2. Another possibility (one that has been brought up before) is exposing an explicit
   method for shutting down the Thread. Developers would invoke this method from
   ServletContextListener.contextDestroyed()

[gissuebot]

From limpbizkit on December 30, 2008 21:03:52

1. yes
2. not necessary. The thread improved speed and memory usage by vacuuming out the collected elements. But
   it's not strictly necessary.

[gissuebot]

From mcculls on December 31, 2008 14:19:55

Any thoughts on the experimental concurrent reference map from JSR166? http://viewvc.jboss.org/cgi- bin/viewvc.cgi/jbosscache/experimental/jsr166/src/jsr166y/ConcurrentReferenceHashMap.
java?view=markup

it doesn't use a thread, but still supports the concurrent map API.

[gissuebot]

From gili.tzabari on December 31, 2008 14:32:44

Very cool! Doug Lea wrote the thing so I am willing to bet that it's pretty solid ;)
There is also the fact that it's scheduled for inclusion in Java7.

[gissuebot]

From limpbizkit on December 31, 2008 15:42:52

Unfortunately, although it's extremely handy, Doug Lea's ConcurrentReferenceHashMap doesn't have the one
method we need, whose signature would resemble either this
  public void putIfAbsent(K key, Function<K, V> keyToValue)
or perhaps this (which is what ReferenceCache uses):
  public abstract V create(K)

[gissuebot]

From gili.tzabari on December 31, 2008 16:13:22

Jesse,

I suppose you could always file a RFE with Doug and see what he says. Alternatively,
would it be possible to store Function in place of the actual value?

[gissuebot]

From dhanji on January 03, 2009 22:45:32

I wanted to add that awesome method to CHM itself, but there didn't seem to be much enthusiasm when I tried.

Afaik it is not solely Doug Lea's RefMap, it was developed by Red Hat et al., in parallel to a somewhat better
implementation that Google open sourced (viz. ReferenceCache). Unfortunately for us, the wrong one made it
into the JDK =(

[gissuebot]

From gili.tzabari on January 03, 2009 22:50:35

I didn't get the impression that it was too late to make changes to their
implementation. Did you file a formal RFE with them regarding this issue (do they
even have a public bug tracking system)? If so please link to it so I can run it by
my contacts at Sun. There is always room to apply more pressure :)

[gissuebot]

From crazyboblee on January 04, 2009 10:19:58

Nothing has made it into the JDK. Doug hasn't even actually looked at the RedHat impl (which doesn't actually
work). I already sent Doug a much better impl that shares code w/ CHM.

[gissuebot]

From dmdabbs on March 16, 2009 09:57:43

// Comment 11 by crazyboblee, Jan 04, 2009
// Nothing has made it into the JDK. Doug hasn't even actually looked at the RedHat
impl (which doesn't actually
//work). I already sent Doug a much better impl that shares code w/ CHM.

Is this better impl available anywhere?

[gissuebot]

From crazyboblee on March 16, 2009 10:03:53

Yes. Check out MapMaker in Google Collections
( https://code.google.com/p/google-collections/ ).

[gissuebot]

From limpbizkit on May 03, 2009 15:49:41

I believe this should be fixed. At a minimum, we now respect a security manager that refuses to let us spawn
threads.

Status: Fixed

[gissuebot]

From syvalta@yahoo.com on October 06, 2009 07:32:52

It seems to me that this issue is not fixed. From
com.google.inject.internal.Finalizer there is still the context class loader
reference, which has been discussed in the related issues.