[c++] Re: Reverted QPID-1879 work for now

[Gordon Sim <gs...@redhat.com>]

Andrew Stitcher wrote:
> The changes I made for QPID-1879 were causing the cluster code to fail
> erratically, so I've reverted nearly all of them.
> 
> One change that I have left in is a change to the linkage of the
> cluster_test unit test program. This change seems to cause a test
> failure:
> 
> ../../../qpid-git/cpp/src/tests/cluster_test.cpp(716): error in
> "queueDurabilityPropagationToNewbie": check durable_query.getDurable()
> == true failed [0 != 1]
> 
> I'm so puzzled that adding force linking shared libraries that should
> not be linked and get loaded anyway can cause this sort of problem that
> I've left that change there for the moment as I think it might actually
> indicate a real issue that might have been hidden by the strange linkage
> before.

I believe this is down to the fact that the failing test was running one 
of the clustered nodes in-process and with the link change the cluster 
module was not loaded in this case and thus there was no replication to 
the second node.

I've altered that test to have both nodes forked (in which case the 
modules are loaded normally).

---------------------------------------------------------------------
Apache Qpid - AMQP Messaging Implementation
Project:      http://qpid.apache.org
Use/Interact: mailto:dev-subscribe@qpid.apache.org

Re: [c++] Re: Reverted QPID-1879 work for now

[Andrew Stitcher <as...@redhat.com>]

On Wed, 2009-06-03 at 17:17 +0100, Gordon Sim wrote:

> > I'm so puzzled that adding force linking shared libraries that should
> > not be linked and get loaded anyway can cause this sort of problem that
> > I've left that change there for the moment as I think it might actually
> > indicate a real issue that might have been hidden by the strange linkage
> > before.
> 
> I believe this is down to the fact that the failing test was running one 
> of the clustered nodes in-process and with the link change the cluster 
> module was not loaded in this case and thus there was no replication to 
> the second node.
> 
> I've altered that test to have both nodes forked (in which case the 
> modules are loaded normally).

Thank you. Do you think this might be an issue with the other failing
cluster tests too?

I guess I'll rebase my change and try it again.

Andrew





---------------------------------------------------------------------
Apache Qpid - AMQP Messaging Implementation
Project:      http://qpid.apache.org
Use/Interact: mailto:dev-subscribe@qpid.apache.org