Community
Participate
Working Groups
With the RSE M5 candidate, PersistenceTest unit tests are failing. When running the test suite, and following what's going on at the screen, I also noticed that some connections seem duplicated. I think it's very important to check the failures since they might be related to recent changes for lazy loading, and persistence is a very important component of RSE. -----------Enter bugs above this line----------- TM 3.0M5 installation : eclipse-SDK-3.4M5 (I20080207-1530), cdt-4.0.0M5, emf-2.4.0M5 DSF-N20071113, ECF-2.0m4, PHPEclipse-1.2.0-Nightly, Releng.Tools-3.4M5, Subversive-0.7.0m4, SWT-MemMonitor, WR-Retriever-3.0.v20070604 RSE install : RSE-SDK-I20080212-2045, TM-terminal, TM-discovery java.runtime : Sun 1.6.0_02-b06 -Xmx512m -XX:MaxPermSize=128m os.name: : Windows XP 5.1, Service Pack 2 ------------------------------------------------ systemtype : Windows-local, Dstore-win, Dstore-linux targetos : Red Hat Enterprise Linux WS release 4 (Nahant Update 3) targetuname : Linux parser 2.6.9-34.EL #1 i686 athlon i386 GNU/Linux targetvm : Sun Java HotSpot(TM) Client VM (build 1.4.2_12-b03, mixed mode) ------------------------------------------------
After some initial investigation, it looks like I'm getting to the cause of the exception by chance - because I put a breakpoint in SWT.error(). Here is what's happening: The test calls JobManager.join() which runs a nested SWT Event Loop: EventLoopProgressMonitor.runEventLoop() line: 127 EventLoopProgressMonitor.isCanceled() line: 97 JobManager.join(Object, IProgressMonitor) line: 773 PersistenceTest.reload() line: 250 PersistenceTest.testProfilePersistence() line: 87 Inside the event loop, the following exception is happening and passed on to the exception handler - but the exception handler just prints an "Unhandled Exception in Event Loop" to the error log, and the error log is never seen from the unit test! org.eclipse.swt.SWTException: Failed to execute runnable (java.lang.NullPointerException) at org.eclipse.swt.SWT.error(SWT.java:3716) at org.eclipse.swt.SWT.error(SWT.java:3634) at org.eclipse.swt.widgets.Synchronizer.runAsyncMessages(Synchronizer.java:133) at org.eclipse.swt.widgets.Display.runAsyncMessages(Display.java:3737) at org.eclipse.swt.widgets.Display.readAndDispatch(Display.java:3374) at org.eclipse.ui.internal.dialogs.EventLoopProgressMonitor.runEventLoop(EventLoopProgressMonitor.java:123) at org.eclipse.ui.internal.dialogs.EventLoopProgressMonitor.done(EventLoopProgressMonitor.java:82) at org.eclipse.core.internal.jobs.JobManager.join(JobManager.java:783) at org.eclipse.rse.tests.persistence.PersistenceTest.reload(PersistenceTest.java:250) [...] Caused by: java.lang.NullPointerException at org.eclipse.jface.viewers.TreeViewer.getExpanded(TreeViewer.java:174) at org.eclipse.rse.internal.ui.view.SafeTreeViewer.getExpanded(SafeTreeViewer.java:143) at org.eclipse.rse.internal.ui.view.SystemView.moveTreeItem(SystemView.java:1503) at org.eclipse.rse.internal.ui.view.SystemView.moveTreeItems(SystemView.java:1549) at org.eclipse.rse.internal.ui.view.SystemView$ResourceChangedJob.runInUIThread(SystemView.java:2287) at org.eclipse.ui.progress.UIJob$1.run(UIJob.java:94) at org.eclipse.swt.widgets.RunnableLock.run(RunnableLock.java:35) at org.eclipse.swt.widgets.Synchronizer.runAsyncMessages(Synchronizer.java:130) ... 59 more So, the unit test continues with the exception being logged to an unvisible destination; and the result of the test is unexpected. I think the first measure needs to be making the exception visible. This can either be done by registering an error handler with JobManager.join(); and/or by setting the WorkbenchUI.DEBUG flag by means of tracing, such that the "unhandled event loop exception" is not only printed to the Eclipse Log but also to stdout. After that, of course, the reason for the NPE in TreeViewer.getExpanded() needs to be found, because this is what's causing not only this one but many other unit tests to fail. Uwe has proposed changing SafeTreeViewer to swallow the NPE's but I'm not sure this is the right solution... we should better go and find the real cause...
Registering a different kind of event handler seems currently impossible, I created bug 219077 for this. The workaround is to enable Tracing of the Workbench Plugin in the Launch that runs the unit tests: org.eclipse.ui.workbench/debug = true in the .options and start with -debug. I'm changing the checked-in RSECombinedTestSuite setting that flag now.
Created attachment 89834 [details] Patch improving error reporting for event loop exceptions Attached patch is an example for improving error reporting about unhandled exceptions in the event loop, taken from a different test case. Attaching this patch as an example showing how such exceptions are best reported through JUnit.
Committed the patch for improving error reporting: [219069] Improved error reporting for exceptions in event loop while running junit plug-in tests
Continuing investigations on bug 219086, which seems unrelated to the persistence test issues.
Dave can you take this over please. On RSE HEAD, the problem reproducably occurs even with the core RSE plugins activated. Also, the problem does NOT seem to be related to SWT issues in any way. Since bug 219086 has been fixed, I cannot see any SWT exceptions any more, but the assertion still fails.
This took an interesting turn. This is a testcase bug. The problem is that reload of a profile - i.e. a reset of the entire model - does not work and has never worked. This is an item for M6 and goes along with import/export. I've fixed the testcase so that it skips the reload and so it succeeds for now.
Thanks Dave. Do you have a bugzilla number for making reload() work? - I'm wondering whether it might make sense to keep the testcase - including reload() in there but just comment it out for now. I'm using a special pattern for existing test cases that are just temporarily commented out. At some points, all of these comments should be re-enabled and the testcases should run: //-test-disabled-// So again, this is for tests which are correct but known to fail currently due to deficiencies, but should be enabled at some later time.
I released the checkin, but verification FAILED on http://download.eclipse.org/dsdp/tm/downloads/drops/I20080215-1245/index.php Start Eclipse, Window > Show View > Other > RSE Testing Select RSECombinedTestSuite, right-click > Run --> testProfilePersistence fails: junit.framework.AssertionFailedError: expected:<3> but was:<4> [...] at org.eclipse.rse.tests.persistence.PersistenceTest.testProfilePersistence(PersistenceTest.java:97) [...] *** testProfilePersistence(org.eclipse.rse.tests.persistence.PersistenceTest): expected:<3> but was:<4>
I also noticed that running the unit tests from the RSE Test Suites View in my normal workspace DELETED the connections that I had in there. This means loss of data, which I consider a CRITICAL issue even though few people would actually run the unit tests. Please investigate.
This is the problem with reload - I managed to not get rid of them all the first time. It should be fixed now. The reload problem is documented in bug 142720.
