[hibernate-dev] http://ci.hibernate.org/job/hibernate-orm-master/48/

[Steve Ebersole]

On the new machine I am not able to get tests hooked in properly.  I 
added a "post build hook" to "publish test results", but the build 
fails saying it does not like my fileset definition.  I set it as 
"**/target/test-results/*.xml", but it says it is not able to find any 
tests using that.  Do we maybe need a special plugin for aggregated 
multi-module test result publishing?

Even worse, after enabling these post-build actions the build time has 
actually increased pretty significantly.  It went from approx 21 
minutes to run the full build to the last build taking 38 minutes and 
that does not even include (1) the completion of the test result 
publishing and (2) the snapshot artifact publishing to nexus.  This 
might be the fulcrum point for the discussion about separate triggered 
and nightly builds.


On Thu 25 Apr 2013 06:46:12 PM CDT, Sanne Grinovero wrote:
> I got it to work with a filthy hack: I've created a new virtual
> volume, attached it to the VM and activated it as additional swap
> drive..
>
> This way the build wasn't killed and could complete, however (as
> expected) it took ages to complete.
>
> At least now we know the proper remedy: use a machine with more than
> 2GB or RAM; I had underestimated our needs, luckily it's an easy fix:
> tomorrow I'll transfer the VM into a different host.
>
> http://ci.hibernate.org/job/hibernate-orm-master/72/
>
> Sanne
>
>
>
> On 25 April 2013 23:49, Steve Ebersole <steve at hibernate.org> wrote:
>> I am watching a build I just kicked off right now in the ui console and the
>> sticking point seems to actually be compiling the test classes (unless I am
>> mis-understanding the console).
>>
>> The step where the build hung (just before dieing) was:
>> :hibernate-core:testClasses
>>
>> I have to see what all testClasses involves.  Actual test source compilation
>> was already completed earlier.
>>
>>
>>
>> On Thu 25 Apr 2013 03:32:30 PM CDT, Sanne Grinovero wrote:
>>>
>>> I think it's an environment problem of our CI, I yet have to
>>> understand what this message means but while using less memory could
>>> be a valid workaround, I still suspect the problem is not in
>>> Hibernate's code. Sorry this is going to be a bit slow I need to get a
>>> RH subscription to get proper support.. Suggestion from Linux experts
>>> welcome in the meantime: I've never seen such an error before.
>>>
>>> Sanne
>>>
>>> On 25 Apr 2013 20:53, "Steve Ebersole" <steve at hibernate.org
>>> <mailto:steve at hibernate.org>> wrote:
>>>
>>>      As far as I can tell there are no tests leaving the SessionFactory
>>>      open in hibernate-core.  I have not tested yet with the other
>>>      modules, but from what I can see the CI builds fail at the point
>>>      trying to run these tests.
>>>
>>>      Essentially I added a hook in CustomRunner (in
>>>      AfterClassCallbackHandler to be precise) to check the
>>>      SessionFactoryRegistry to make sure there are no still registered
>>>      SessionFactories after a test class is done.  If there are, it
>>>      logs a message and clears the SessionFactoryRegistry.  This is
>>>      simplistic as we still have other resources that need to be
>>>      checked (ServiceRegistries are used independently from
>>>      SessionFactory quite a bit).
>>>
>>>      This was after I got all tests indicated by Strong to extend
>>>      BaseUnitTestCase that did not previously.  I did have problems
>>>      with
>>>
>>> org.hibernate.metamodel.__source.annotations.xml.mocker.__AbstractMockerTest
>>>
>>>      and subclasses.  Having it extend BaseUnitTestCase with that
>>>      functionality led to assertion failures in 3 tests:
>>>
>>>      BasicMockerTest.__testEntityWithEntityMappingsCo__nfiguration
>>>      OverrideTest.__testPersistenceUnitDefaultsCas__cadePersistInAnnotation
>>>      OverrideTest.__testPersistenceUnitMetadataMet__adataComplete
>>>
>>>
>>>      Which is odd, because those tests don't even deal with a
>>>      SessionFactory.
>>>
>>>      As for checking that created ServiceRegistry instances get cleaned
>>>      up, not sure how we can check that.
>>>
>>>      Another thing we should do is to exclude the metamodel tests
>>>      currently on master.  After that, I am out of ideas...
>>>
>>>      On Thu 25 Apr 2013 05:28:35 AM CDT, Sanne Grinovero wrote:
>>>
>>>          On 25 April 2013 09:04, Strong Liu <stliu at hibernate.org
>>>          <mailto:stliu at hibernate.org>> wrote:
>>>
>>>              Sanne,
>>>
>>>              is it possible that you give me ssh access to the ci server?
>>>
>>>
>>>          Sure; I'm adding Strong, if anyone else wants root access just
>>>          send me
>>>          your public key using your redhat account.
>>>
>>>          Cheers,
>>>          Sanne
>>>
>>