You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@tapestry.apache.org by Nourredine <no...@atos.net> on 2011/09/21 11:31:43 UTC

[Tapestry 5.2] Per-thread scope services not thread safe ?

Hi,

Since Tapestry 5.2, it seems that per-thread scope services are not
thread-safe (some of our integration tests executed in parallel fail since
Tapestry 5.2).

We do not obtain a new instance of the service for each request/thread as
expected. 
That leads to strange behaviour when you use the Environmental services as
data's container for example.

I also noticed that the synchronization of some resources (used by
per-thread scope services) is effective only for jdk 1.5 (see
PerthreadmanagerImpl's constructor and JDUtils.JDK_1_5). But this aspect
does not explain everything : the pb occurs with JDK 1.5 too.

Has someone already experimented that pb? should I raise a JIRA? (I can
provide a sample that reproduces the issue)

Nourredine.

--
View this message in context: http://tapestry.1045711.n5.nabble.com/Tapestry-5-2-Per-thread-scope-services-not-thread-safe-tp4825735p4825735.html
Sent from the Tapestry - Dev mailing list archive at Nabble.com.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@tapestry.apache.org
For additional commands, e-mail: dev-help@tapestry.apache.org


Re: [Tapestry 5.2] Per-thread scope services not thread safe ?

Posted by Martin Strand <do...@gmail.com>.
On Thu, 22 Sep 2011 11:03:02 +0200, Nourredine  
<no...@atos.net> wrote:

> Thanks to Martin, I realized that my sample was not relevant and that the
> conditioned lock around the holder, in PerthreadManagerImpl, is not a
> problem and maybe, "in the worst case", only redundant.

Indeed, PerThreadServiceCreator somewhat looks like a race bug and might  
deserve a comment explaining why it's not:

if (perThreadValue.exists())
    return perThreadValue.get();
  return perThreadValue.set(delegate.createObject());

Eventhough it's not obvious at a first glance, I believe this really is  
thread safe since the ThreadLocal in PerthreadManagerImpl#holder holds the  
only mutable object involved here.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@tapestry.apache.org
For additional commands, e-mail: dev-help@tapestry.apache.org


Re: [Tapestry 5.2] Per-thread scope services not thread safe ?

Posted by Nourredine <no...@atos.net>.
Hi,

Finally, it was a false alarm. See the discussion on the jira :
https://issues.apache.org/jira/browse/TAP5-1660

Thanks to Martin, I realized that my sample was not relevant and that the
conditioned lock around the holder, in PerthreadManagerImpl, is not a
problem and maybe, "in the worst case", only redundant. 

My original issue is probably related to a side effect from our code since
the tapestry 5.2 migration (the issue does not occur when the page pool is
enabled).

By the way, it's really a pleasure to read the tapestry source code and
discuss with the community.

Nourredine.

--
View this message in context: http://tapestry.1045711.n5.nabble.com/Tapestry-5-2-Per-thread-scope-services-not-thread-safe-tp4825735p4829505.html
Sent from the Tapestry - Dev mailing list archive at Nabble.com.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@tapestry.apache.org
For additional commands, e-mail: dev-help@tapestry.apache.org


Re: [Tapestry 5.2] Per-thread scope services not thread safe ?

Posted by Nourredine <no...@atos.net>.
The related JIRA : https://issues.apache.org/jira/browse/TAP5-1660

--
View this message in context: http://tapestry.1045711.n5.nabble.com/Tapestry-5-2-Per-thread-scope-services-not-thread-safe-tp4825735p4826258.html
Sent from the Tapestry - Dev mailing list archive at Nabble.com.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@tapestry.apache.org
For additional commands, e-mail: dev-help@tapestry.apache.org


Re: [Tapestry 5.2] Per-thread scope services not thread safe ?

Posted by Nourredine <no...@atos.net>.
Thiago,

Thanks for answering and clarifying my previous sentence . What I meant is
that , in fine, the proxy is expected to use a new instance of the service
for each thread/request.

I'll raise a jira!

Nourredine.






--
View this message in context: http://tapestry.1045711.n5.nabble.com/Tapestry-5-2-Per-thread-scope-services-not-thread-safe-tp4825735p4826193.html
Sent from the Tapestry - Dev mailing list archive at Nabble.com.

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@tapestry.apache.org
For additional commands, e-mail: dev-help@tapestry.apache.org