commons-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Phil Steitz (JIRA)" <j...@apache.org>
Subject [jira] Commented: (POOL-157) GenericObjectPool.setFactory(...) does not destroy idle pool objects with their original factory object.
Date Wed, 10 Feb 2010 13:24:28 GMT

    [ https://issues.apache.org/jira/browse/POOL-157?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12831981#action_12831981
] 

Phil Steitz commented on POOL-157:
----------------------------------

This is the result of the change in pool 1.5 to move factory methods outside of synchronized
scope to avoid deadlocks.  The obvious fix for this would be to move the call to destroy above
the factory reassignment.  This would, however, violate the lock order principle above.  


I am curious as to the use case for resetting the factory on an active pool.   Why not just
close the pool and create a new one with the new factory?  Personally, I have always thought
that the factory should be an immutable property of the pool.   I am interested in the understanding
the use case.

I propose to resolve this by
* deprecating the setter - to be removed in 2.0
* adding private destroy(PoolableObjectFactory, Collection) and passing it the old factory

> GenericObjectPool.setFactory(...) does not destroy idle pool objects with their original
factory object.
> --------------------------------------------------------------------------------------------------------
>
>                 Key: POOL-157
>                 URL: https://issues.apache.org/jira/browse/POOL-157
>             Project: Commons Pool
>          Issue Type: Bug
>            Reporter: David Hu
>            Priority: Minor
>
> When setting a new object factory, the existing idle poolable objects are not destroyed
by the same factory that created them.
> 1453	    public void setFactory(PoolableObjectFactory factory) throws IllegalStateException
{
> 1454	        List toDestroy = new ArrayList();
> 1455	        synchronized (this) {
> 1456	            assertOpen();
> 1457	            if(0 < getNumActive()) {
> 1458	                throw new IllegalStateException("Objects are already active");
> 1459	            } else {
> 1460	                toDestroy.addAll(_pool);
> 1461	                _numInternalProcessing = _numInternalProcessing + _pool._size;
> 1462	                _pool.clear();
> 1463	            }
> 1464	            _factory = factory;
> 1465	        }
> 1466	        destroy(toDestroy); // <----- indirectly calls _factory.destroy(...)
when _factory has already been replaced by the new factory.
> 1467	    }

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.


Mime
View raw message