[infinispan-dev] playing as Hudson

[Sanne Grinovero]

Hello,
I'm going to play the role of Hudson today, as the last build
published on http://hudson.jboss.org/hudson/view/Infinispan/ is from 5
months ago.
Isn't there a working CI for Infinispan? I've had some trouble today
with unexpected issues, according to JIRA I could find some issues
related to my problems,
but most of them were solved recently according to JIRA. The tests I
just have run are having a different opinion:

Failed tests:
  testTransactional(org.infinispan.distribution.rehash.ConcurrentOverlappingLeaveTest)
  testWriteSkewWithOnlyPut(org.infinispan.api.mvcc.repeatable_read.WriteSkewTest)
  testEnableJmxStats(org.infinispan.jmx.RpcManagerMBeanTest)
  testTransactional(org.infinispan.distribution.rehash.ConcurrentJoinTest)
  testNonTransactional(org.infinispan.distribution.rehash.ConcurrentJoinTest)
  testonInfinispanDIST(org.infinispan.stress.PutIfAbsentStressTest)

Some comments:

org.infinispan.distribution.rehash.ConcurrentOverlappingLeaveTest
Is NOT failing consistently (it sometimes runs fine); It's
inconsistent even running it alone mvn test
-Dtest=org.infinispan.distribution.rehash.ConcurrentOverlappingLeaveTest

org.infinispan.api.mvcc.repeatable_read.WriteSkewTest
Is failing all the time.

org.infinispan.jmx.RpcManagerMBeanTest is "interesting" as the assert
fails on this message "Expected 1, was 1".
Debugging I see it's the wrong type:
mBeanServer.getAttribute(rpcManager1, "ReplicationCount").equals("1")
fails because getAttribute is returning a "Long", not a String.

org.infinispan.distribution.rehash.ConcurrentJoinTest
This one was the reason for me to rerun all tests, as it's making me
fail several Lucene index tests.
The message "Some caches have not finished rehashing after 8 minutes":
gets me a bit worried :-)

org.infinispan.stress.PutIfAbsentStressTest
This one is also breaking Lucene; we know for sure that it was running
fine as I've tested it several times when Markus fixed the related
issue.

I'm attaching the full reports with stacktraces.
Isn't there a real Hudson running to prevent this? The weird things I
experienced today were killing my brain, as I was looking in the wrong
direction expecting that stuff fixed last week was stil fine.
Sorry for telling that all was fine with Lucene Directory this
morning; I'm going to step through recent versions to identify the
breaking change, it's the only way I can help.

Cheers,
Sanne
-------------- next part --------------
A non-text attachment was scrubbed...
Name: infinispan-failed-tests-report.tar.bz2
Type: application/x-bzip2
Size: 2575 bytes
Desc: not available
Url : http://lists.jboss.org/pipermail/infinispan-dev/attachments/20091111/fb4c8300/attachment-0002.bz2