The following bug has been reported and fixed:
450523: [Profiles] Performance bottleneck and exceptions in StereotypeElementListener
The Master test builds are now operational, and the tests for plugins/ shows 23 failures:
The extra-tests are still green
It seems that the test build is not « completely » frozen ; only “super-slow” (It’s been running for 1h30 in my computer, and only ~30% of the tests have been covered so far). I have a lot of exceptions related to the new Stereotype listener:
!ENTRY org.eclipse.emf.transaction 4 45 2014-11-07 12:53:26.607
!MESSAGE Uncaught exception during post-commit listener notifications
java.lang.NullPointerException
at org.eclipse.papyrus.uml.tools.listeners.StereotypeElementListener.handleFilteredNotification(StereotypeElementListener.java:157)
at org.eclipse.papyrus.uml.tools.listeners.StereotypeElementListener.resourceSetChanged(StereotypeElementListener.java:88)
at org.eclipse.emf.transaction.impl.TransactionalEditingDomainImpl$1.run(TransactionalEditingDomainImpl.java:781)
at org.eclipse.emf.transaction.impl.TransactionalEditingDomainImpl.runExclusive(TransactionalEditingDomainImpl.java:328)
at org.eclipse.emf.transaction.impl.TransactionalEditingDomainImpl.postcommit(TransactionalEditingDomainImpl.java:771)
at org.eclipse.emf.transaction.impl.TransactionalEditingDomainImpl.deactivate(TransactionalEditingDomainImpl.java:543)
at org.eclipse.emf.transaction.impl.TransactionImpl.close(TransactionImpl.java:712)
at org.eclipse.emf.transaction.impl.TransactionImpl.commit(TransactionImpl.java:474)
at org.eclipse.emf.workspace.AbstractEMFOperation.redo(AbstractEMFOperation.java:463)
at org.eclipse.core.commands.operations.DefaultOperationHistory.doRedo(DefaultOperationHistory.java:357)
at org.eclipse.core.commands.operations.DefaultOperationHistory.redo(DefaultOperationHistory.java:1110)
at org.eclipse.papyrus.commands.CheckedOperationHistory.redo(CheckedOperationHistory.java:220)
at org.eclipse.gmf.runtime.diagram.ui.parts.DiagramCommandStack.redo(DiagramCommandStack.java:366)
at org.eclipse.papyrus.sysml.diagram.internalblock.tests.utils.TestUtils.defaultExecutionTest(TestUtils.java:742)
at org.eclipse.papyrus.sysml.diagram.internalblock.tests.utils.TestUtils.setEncapsulationDeleteConnectorTest(TestUtils.java:591)
at org.eclipse.papyrus.sysml.diagram.internalblock.tests.set.encapsulation.AbstractSetEncapsulationDeleteConnectorTest.setEncapsulationDeleteConnectorWithTargetProperty(AbstractSetEncapsulationDeleteConnectorTest.java:114)
And I had already noticed some (apparently major) performances issues with this new Listener (Although I wasn’t sure enough to raise a bug). The
filter is invoked on absolutely all model notifications, so it is really important that extra care is brought to its performances (Especially, it shouldn’t have to load any resource, and it should never browse the model tree – only follow forward references.
Not sure why it is so slow exactly, I haven’t investigated)
I would like to remind everyone that the Gerrit build only checks that everything compiles properly (plugins/ + extraplugins/), but currently doesn’t run any
test. Thus, once your contribution has been accepted, please verify that the tests are still green:
With the build failures for the last 3 days, a lot of contributions have been merged without being automatically tested, and we have a lot of regressions (Hopefully
minor ones):
- Luna-Tests
has 9 failures, mainly Memory Leaks
- Luna-Extra-Tests
is still green
- Master-Tests
is frozen and cannot complete (And shortly before the failure, showed 5 new regressions)
- Master-Extra-Tests
is frozen and cannot complete
Frozen builds are difficult to debug online, so you probably need to run the test suites locally (~30-40 minutes). I will run a complete test suite and hopefully
be able to give some more precise information regarding the failures.