[jnet] Threading issues
Reinhard Pötz <reinhard <at> apache.org>
2010-02-05 10:11:37 GMT
Moving over this thread to dev.
Carsten, any idea what could go wrong here? And, once I asked you why
you used an InheritableThreadLocal instead of a ThreadLocal to store the
list but can't remember. Can you explain it to me again? 
--
Reinhard Pötz Managing Director, {Indoqa} GmbH
http://www.indoqa.com/en/people/reinhard.poetz/
Member of the Apache Software Foundation
Apache Cocoon Committer, PMC member reinhard <at> apache.org
________________________________________________________________________
Reinhard Pötz wrote:
> Gabriel Gruber wrote:
>> Hi Joe,
>>
>> well, I don't have a replayable scenario yet (f.i. testcase), but it
>> seems that the exception comes after heavy use of cocoon pages within
>> our application.
>> I suspect that cocoon:/ protocol usage AND/OR inheritence of sitemap
>> calling (we "extend" cocoon-forms block to override specific resources)
>> might be the root of problem. But that is only a guess...
>>
>> We are using latest trunk versions of servlet-service-framework and jnet
>> subprojects. I know that Grek made some changes in cocoon servlet
>> service framework in summer 2009, which went into the 1.2 release, which
>> we are using (since then no changes have been comitted).
>>
>> The problems reported here (exact same problem we have)
>>
>> http://jira.dspace.org/jira/browse/DS-253?page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel
>
> I had a look at the JNet code and haven't found a cause for your
> problems. The list of URLStreamHandlerFactory objects is stored in a
> thread local.
>
> I propose that you add extensive logging to
> org.apache.cocoon.jnet.URLHandlerFactoryCollector,
> org.apache.cocoon.jnet.DynamicURLStreamHandlerFactory and
> org.apache.cocoon.jnet.URLStreamHandlerFactoryInstaller
>
> Then create a special logging target for all org.apache.cocoon.jnet
> messages and make sure that you also log the tread ids.
>
> Then try to reproduce your problems and you should get a hint from the
> logs what could be the cause of empty lists.
>
> Maybe you can also provide a patch with the logging statements.
It might also be interesting to see, whether the exceptions changes (or
disappears) if you change line 42 of
o.a.c.jnet.DynamicURLStreamHandlerFactory to
list = Collections.synchronizedList(new
LinkedList<URLStreamHandlerFactory>());
Steven and I experimented with linked lists today and didn't find a way
to produce a NPE as long as only a single thread accesses it. Those
exceptions only occur in the case of multiple threads modifying it
without synchronization.
So I expect that this change will just lead to an
IndexOutOfBoundException ...
--
--
Reinhard Pötz Managing Director, {Indoqa} GmbH
http://www.indoqa.com/en/people/reinhard.poetz/
Member of the Apache Software Foundation
Apache Cocoon Committer, PMC member reinhard <at> apache.org
________________________________________________________________________
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe <at> cocoon.apache.org
For additional commands, e-mail: users-help <at> cocoon.apache.org
RSS Feed