JBMICROCONT-74 - Testcases for redeployments us...| JBoss.org Content Archive (Read Only)

15. Re: JBMICROCONT-74 - Testcases for redeployments using JVMTI

clebert.suconic Apr 17, 2006 11:00 AM (in response to clebert.suconic)

Ah ok...

JBossMemoryTestCase could be considered abstract. I didn't specify the keyword abstract though but I could do it.

I guess this is okay then.

16. Re: JBMICROCONT-74 - Testcases for redeployments using JVMTI

adrian.brock Apr 18, 2006 7:26 AM (in response to clebert.suconic)

This is broken:
http://cruisecontrol.jboss.com/cc/artifacts/microcontainer-head-testsuite/20060418063626/results/TEST-org.jboss.test.memorytests.ClassInfoMemoryTestCase.xml

17. Re: JBMICROCONT-74 - Testcases for redeployments using JVMTI

bkeh12 Apr 18, 2006 11:10 AM (in response to clebert.suconic)

Hi,Adrain
you configer is

<'property name="os.name" value="Linux"/>
<'property name="os.arch" value="i386"/>
<'property name="os.version" value="2.6.9-22.0.1.ELsmp"/>
<'property name="sun.boot.library.path" value="/opt/j2sdk1.4.2_07/jre/lib/i386"/>

"clebert.suconic@jboss.com" wrote:

also, it's important to understand that this tool can only be used on a JVM 5. (needs JVMTI support)

thank

18. Re: JBMICROCONT-74 - Testcases for redeployments using JVMTI

clebert.suconic Apr 18, 2006 11:18 AM (in response to clebert.suconic)

Chung,

This happens on cruisecontrol, which is an automatic building system. I should have taken care and make sure this would run there, it was my mistake... my bad.

Anyway, JVMTIInterface has a check fo the library being loaded.

 static {
 try {
 System.loadLibrary("jbossAgent");
 }
 catch (Throwable e) {
 isLoaded=false;
 log.error("The DLL couldn't be loaded, you won't be able to use any JVMTIInterface feature",e);
 }
 }

This exception handling was broken before. ( I was using Exception e)

So, if jbossAgent.dll is not available, the test will just use a WeakReference, System.GC() and a timeout to make sure the reference is gone, and the test should work fine now.

You will see a bunch of jvmti.isActive on the super Class and on the test itself.

19. Re: JBMICROCONT-74 - Testcases for redeployments using JVMTI

clebert.suconic Apr 18, 2006 2:30 PM (in response to clebert.suconic)

Fixed:

http://cruisecontrol.jboss.com/cc/artifacts/microcontainer-head-testsuite/20060418133856/results/org/jboss/test/memorytests/ClassInfoMemoryTestCase.html

20. Re: JBMICROCONT-74 - Testcases for redeployments using JVMTI

bkeh12 Apr 20, 2006 2:30 AM (in response to clebert.suconic)

Hi,Adrian

Why (sample1) use MC's Javassist reflect model to test is perfectly right ?

sample 1:
 public class ClassInfoMemoryTestCase extends JBossMemoryTestCase
 {
 .........

 //Class simpleBeanClass = loader.loadClass("org.jboss.test.classinfo.support.SimpleBean");
 //Class simpleBeanInterface = loader.loadClass("org.jboss.test.classinfo.support.SimpleInterface");
 ClassInfo cinfo = getClassInfo("org.jboss.test.classinfo.support.SimpleBean",loader);
 cinfo=null;

 //Test if the reference on ClassInfo is not too weak
 JVMTIInterface jvmti = new JVMTIInterface();
 if (jvmti.isActive())
 {
 jvmti.forceGC();
 assertEquals(1,jvmti.getAllObjects(ClassInfo.class).length);

 }

 cinfo = getClassInfo("org.jboss.test.classinfo.support.SimpleBean",loader);

 ..........

 protected TypeInfoFactory getTypeInfoFactory()
 {
 //return new IntrospectionTypeInfoFactory();
 return new JavassistTypeInfoFactory();

 }
 }


Testsuite: org.jboss.test.memorytests.ClassInfoMemoryTestCase
Tests run: 1, Failures: 1, Errors: 0, Time elapsed: 0.218 sec
------------- Standard Output ---------------
+++oldloader =sun.misc.Launcher$AppClassLoader@16930e2
+++newloader =java.net.FactoryURLClassLoader@1f14ceb
------------- ---------------- ---------------

Testcase: testSimpleBeanClassInfo took 0.218 sec
 FAILED
expected:<1> but was:<0>
junit.framework.AssertionFailedError: expected:<1> but was:<0>
 at org.jboss.test.memorytests.ClassInfoMemoryTestCase.testSimpleBeanClassInfo(ClassInfoMemoryTestCase.java:75)


sample 2:
 public class ClassInfoMemoryTestCase extends JBossMemoryTestCase
 {
 .........

 //Class simpleBeanClass = loader.loadClass("org.jboss.test.classinfo.support.SimpleBean");
 //Class simpleBeanInterface = loader.loadClass("org.jboss.test.classinfo.support.SimpleInterface");
 ClassInfo cinfo = getClassInfo("org.jboss.test.classinfo.support.SimpleBean",loader);
 //cinfo=null;

 //Test if the reference on ClassInfo is not too weak
 //JVMTIInterface jvmti = new JVMTIInterface();
 //if (jvmti.isActive())
 //{
 // jvmti.forceGC();
 // assertEquals(1,jvmti.getAllObjects(ClassInfo.class).length);

 //}

 //cinfo = getClassInfo("org.jboss.test.classinfo.support.SimpleBean",loader);

 ..........

 protected TypeInfoFactory getTypeInfoFactory()
 {
 //return new IntrospectionTypeInfoFactory();
 return new JavassistTypeInfoFactory();

 }


Testsuite: org.jboss.test.memorytests.ClassInfoMemoryTestCase
Tests run: 1, Failures: 0, Errors: 0, Time elapsed: 0.266 sec
------------- Standard Output ---------------
+++oldloader =sun.misc.Launcher$AppClassLoader@16930e2
+++newloader =java.net.FactoryURLClassLoader@1f14ceb
------------- ---------------- ---------------
Testcase: testSimpleBeanClassInfo took 0.266 sec

sample 3:
 public class ClassInfoMemoryTestCase extends JBossMemoryTestCase
 {
 .........

 //Class simpleBeanClass = loader.loadClass("org.jboss.test.classinfo.support.SimpleBean");
 //Class simpleBeanInterface = loader.loadClass("org.jboss.test.classinfo.support.SimpleInterface");
 ClassInfo cinfo = getClassInfo("org.jboss.test.classinfo.support.SimpleBean",loader);
 cinfo=null;

 //Test if the reference on ClassInfo is not too weak
 JVMTIInterface jvmti = new JVMTIInterface();
 if (jvmti.isActive())
 {
 jvmti.forceGC();
 assertEquals(1,jvmti.getAllObjects(ClassInfo.class).length);

 }

 cinfo = getClassInfo("org.jboss.test.classinfo.support.SimpleBean",loader);

 ..........

 protected TypeInfoFactory getTypeInfoFactory()
 {
 return new IntrospectionTypeInfoFactory();
 //return new JavassistTypeInfoFactory();

 }


Testsuite: org.jboss.test.memorytests.ClassInfoMemoryTestCase
Tests run: 1, Failures: 0, Errors: 0, Time elapsed: 0.266 sec
------------- Standard Output ---------------
+++oldloader =sun.misc.Launcher$AppClassLoader@16930e2
+++newloader =java.net.FactoryURLClassLoader@1f14ceb
------------- ---------------- ---------------
Testcase: testSimpleBeanClassInfo took 0.266 sec

21. Re: JBMICROCONT-74 - Testcases for redeployments using JVMTI

adrian.brock Apr 20, 2006 7:12 AM (in response to clebert.suconic)

The prototype javassist stuff doesn't understand scoped classloading.
It just uses the default classpool impl from javassist.
http://www.jboss.com/index.html?module=bb&op=viewtopic&t=77746

22. Re: JBMICROCONT-74 - Testcases for redeployments using JVMTI

bkeh12 Apr 20, 2006 9:14 AM (in response to clebert.suconic)

Hi,adrian

I need to do the homework to understand well you said.

thank

23. Re: JBMICROCONT-74 - Testcases for redeployments using JVMTI

bkeh12 Apr 20, 2006 10:53 AM (in response to clebert.suconic)

Hi,adrian

Forgives me to ask a stupid question.
Is org.jboss.reflect.plugins.ClassInfoImpl$UnknownClassInfo Strong Reference ?
Can it produce memory leaks on redeployment ?


 //Test if the reference on ClassInfo is not too weak
 JVMTIInterface jvmti = new JVMTIInterface();
 if (jvmti.isActive())
 {
 jvmti.forceGC();
 assertEquals(1,jvmti.getAllObjects(ClassInfo.class).length);
 //is org.jboss.reflect.plugins.ClassInfoImpl$UnknownClassInfo Strong Reference ?

 }


public class ClassInfoImpl extends InheritableAnnotationHolder implements ClassInfo
{
 .......

 /** Marker for generation */
 static final ClassInfo UNKNOWN_CLASS = new UnknownClassInfo();
 .......

 static class UnknownClassInfo implements ClassInfo //static inner class
 {
 .......

 }
}

thank

24. Re: JBMICROCONT-74 - Testcases for redeployments using JVMTI

clebert.suconic Apr 20, 2006 11:54 PM (in response to clebert.suconic)

This is about your three samples. three messages ago.

More specifically in your first sample.

The assertion is not about a memory leak. It's more if the reference is not too weak, havng something being released sooner than expected.

During my tests I didn't find that situation happening. What was your change that made this happen. This was not clear from your example.

25. Re: JBMICROCONT-74 - Testcases for redeployments using JVMTI

bkeh12 Apr 21, 2006 2:53 AM (in response to clebert.suconic)

The assertion is not about a memory leak. It's more if the reference is not too weak, havng something being released sooner than expected.

the ClassInfoImpl$UnknownClassInfo can be released.
Sorry this is my mistake...:-)