Yes thanks! the tests don't deadlock anymore, they just fail consistently:

Failed tests:
testForceCommitXid(org.infinispan.tx.recovery.admin.SimpleCacheRecoveryAdminTest):
expected:<false> but was:<true>
 timeBasedTakeOffline(org.infinispan.xsite.offline.OfflineStatusTest):
expected:<false> but was:<true>
 indexWasStored(org.infinispan.loaders.decorators.BatchAsyncCacheStoreTest):
expected:<false> but was:<true>
 testTwoNamedCachesSameNode(org.infinispan.tx.TransactionsSpanningDistributedCachesTest):
Timed out waiting for rebalancing to complete on node
TransactionsSpanningDistributedCachesTest-NodeS-50037, expected member
list is [TransactionsSpanningDistributedCachesTest-NodeS-50037,
TransactionsSpanningDistributedCachesTest-NodeT-374], current member
list is [TransactionsSpanningDistributedCachesTest-NodeS-50037]!

Tests run: 1640, Failures: 4, Errors: 0, Skipped: 0

Not perfect, but a nice progress.

Cheers,
Sanne


On 11 October 2012 12:47, Manik Surtani <manik@jboss.org> wrote:

Sanne, does this fix the test suite for you?  For whatever reason, this

issue appears on your machine more often than anyone else's … :)

On 3 Oct 2012, at 12:56, Dan Berindei <dan.berindei@gmail.com> wrote:

Thanks for the workaround, Bela!

I've included it in https://github.com/infinispan/infinispan/pull/1358.

On Wed, Oct 3, 2012 at 9:32 AM, Bela Ban <bban@redhat.com> wrote:

On 10/2/12 6:02 PM, Dan Berindei wrote:

The members should be updated in viewAccepted(). When you connect a

channel, then viewAccepted() will be called *before* JChannel.connect()

returns, so the main thread won't have a chance of accessing members

before that.

Ok, if viewAccepted() is guaranteed to be called before connect()

returns

then you're right, the latch is superfluous. We only need to call

viewAccepted() ourselves if the channel was injected (which we already

do).

In this particular case, however, we don't get a viewAccepted() call or

an

exception from connect(), so I'm not sure it would have been better

(easier

to debug) without the latch.

This is due to https://issues.jboss.org/browse/JGRP-1522, which was

introduced in 3.2.0; when this is fixed, you will get a viewAccepted().

The workaround is to set NAKACK{2}.become_server_queue_size=0.

--

Bela Ban, JGroups lead (http://www.jgroups.org)

_______________________________________________

infinispan-dev mailing list

infinispan-dev@lists.jboss.org

https://lists.jboss.org/mailman/listinfo/infinispan-dev

_______________________________________________

infinispan-dev mailing list

infinispan-dev@lists.jboss.org

https://lists.jboss.org/mailman/listinfo/infinispan-dev

--

Manik Surtani

manik@jboss.org

twitter.com/maniksurtani

Platform Architect, JBoss Data Grid

http://red.ht/data-grid

_______________________________________________

infinispan-dev mailing list

infinispan-dev@lists.jboss.org

https://lists.jboss.org/mailman/listinfo/infinispan-dev

_______________________________________________
infinispan-dev mailing list
infinispan-dev@lists.jboss.org
https://lists.jboss.org/mailman/listinfo/infinispan-dev