You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@geronimo.apache.org by Alexei Zakharov <al...@gmail.com> on 2006/03/17 18:21:24 UTC
unit test failures
Hi, community!
I experiment with running Geronimo on various JVM's. The interesting thing
I've encountered is that some of unit tests fail on BEA Jrockit VM. I've got
at least three failures unique to BEA Jrockit 1.4.2_04 VM:
Module: modules/directory
Test: org.apache.geronimo.directory.RunningTest
Result: VM hangs
Module: modules/timer
Test:
org.apache.geronimo.timer.NontransactionalThreadPooledTimerTest.testTasksInUnspecifiedTxContext
Output: expected:<20> but was:<19>
Module: modules/tomcat
Test: org.apache.geronimo.tomcat.JAASSecurityTest.testNotAuthorized
Output: expected:<<!-- Login Page -->> but was:<null>
These tests pass on all Sun VMs. Situation with Jrockit 1.5 is much worse -
more than 60 failures. It may result in overall instability while running
Geronimo on BEA.
Any comments & suggestions?
Alexei Zakharov,
Intel Middleware Product Division
Re: unit test failures [Directory Problem]
Posted by Matt Hogstrom <ma...@hogstrom.org>.
Genender might have an idea.
Alexei Zakharov wrote:
> Hi Kevan,
>
> This is follow-up to the story about BEA VMs. I have analyzed
> org.apache.geronimo.directory.RunningTest (modules/directory) failure
> on BEA JRockit. I've raised the JIRA issue #1805
> (http://issues.apache.org/jira/browse/GERONIMO-1805) with the detailed
> result of my investigation. In brief, IMHO there is some problem with
> LDAP server built in Geronimo. As far as I understand LDAP server is
> not part of Geronimo itself but is an independent component. It is
> part of some "apacheds-core" package. Therefore, it seems the problem
> is not a "pure" Geronimo problem. Do anybody knows from where apacheds
> comes from? Is it a separate Apache project?
>
> This is follow-up to the story about BEA VMs. I have analized
>
> 2006/3/21, Kevan Miller <ke...@gmail.com>:
>
>>
>>On Mar 20, 2006, at 1:19 PM, Alexei Zakharov wrote:
>>
>>Hi Kevan,
>>
>>I was lucky to build it using JRockit 1.4.2 + Maven 1.1 Beta. Therefore,
>>this was a "pure JRockit" experiment.
>>
>>As for dependencies from Sun, AFAIK there are a few non-CORBA places in
>>code with "com.sun" hardcoded. This issue has been rised recently in the
>>list. You may check "Using non-Sun JNDI/RMI service provider" thread for
>>details.
>>
>>Ok, I will continue investigation of these failures. BTW, if you are
>>interested in BEA 1.5 I attach the JRockit 1.5 failure log. Messages are
>>direct JUnit messages. I was using JRockit 1.5.0-b64 win32 + maven 1.0.2 to
>>obtain these results.
>>
>>
>>Hi Alexei,
>>Thanks. Let us know what you find.
>>
>>Looks like the majority of the 1.5 problems are "environmental" something to
>>do with the junit environment on JRockit 1.5. The "kernel is already running
>>this kernel name: geronimo.test" and "runFinalizer" messages seem to imply
>>that you're seeing collisions between Junit tests on different components...
>>
>>--kevan
>>
>>
>>Regards,
>>Alexei
>>
>>2006/3/18, Kevan Miller <ke...@gmail.com>:
>>
>>>Hi Alexei,
>>>Thanks for the information. Are you both building <and> running tests with
>>
>>JRockit? Or are you building w/ Sun and only running tests under JRockit?
>>Would be good to isolate the two. Good chance, however, that it won't make
>>any difference...
>>
>>>
>>>Geronimo currently has direct ties to the Sun 1.4 corba implementation.
>>
>>That's the only JRE restriction that I'm aware of. People have run Geronimo
>>on Sun 1.5 and also on IBM's 1.4 JRE.
>>
>>>
>>>It seems that the tests you describe <ought> to work. There's a reasonable
>>
>>chance that these are Geronimo bugs that are masked by our current runtime.
>>So, it would be great if you could help us uncover the causes of these
>>failures.
>>
>>>
>>>I don't know why JRockit 1.5 would be worse.
>>>
>>>
>>>--kevan
>>>
>>>
>>>
>>>On Mar 17, 2006, at 12:21 PM, Alexei Zakharov wrote:
>>>
>>>
>>>
>>>Hi, community!
>>>
>>>I experiment with running Geronimo on various JVM's. The interesting thing
>>
>>I've encountered is that some of unit tests fail on BEA Jrockit VM. I've got
>>at least three failures unique to BEA Jrockit 1.4.2_04 VM:
>>
>>>Module: modules/directory
>>>Test: org.apache.geronimo.directory.RunningTest
>>>Result: VM hangs
>>>
>>>Module: modules/timer
>>>Test:
>>
>>org.apache.geronimo.timer.NontransactionalThreadPooledTimerTest.testTasksInUnspecifiedTxContext
>>
>>>Output: expected:<20> but was:<19>
>>>
>>>Module: modules/tomcat
>>>Test:
>>
>>org.apache.geronimo.tomcat.JAASSecurityTest.testNotAuthorized
>>
>>>Output: expected:<<!-- Login Page -->> but was:<null>
>>>
>>>These tests pass on all Sun VMs. Situation with Jrockit 1.5 is much worse
>>
>>- more than 60 failures. It may result in overall instability while running
>>Geronimo on BEA.
>>
>>>Any comments & suggestions?
>>>
>>>Alexei Zakharov,
>>>Intel Middleware Product Division
>>>
>>>
>>
>>
>><bea-1.5_geronimo-1.0.html>
>>
>>
>
>
> --
> Alexei Zakharov,
> Intel Middleware Product Division
>
>
>
Re: unit test failures
Posted by Alexei Zakharov <al...@gmail.com>.
Hi Jacek & all,
I've discovered that the most recent Jrockit VM + ApacheDS 1.0 RC1
will completely solve the problem with the
org.apache.geronimo.directory.RunningTest hang on Jrockit (see the
discussion at DIRECTORY-607 for details). Moreover, I've seen some
other posts with signs of dissatisfaction with the outdated version of
Apache DS currently being used. So it seems that we need to move
Geronimo to ApacheDS 1.0 anyway. I can start working on the
corresponding patch. Any comments & objections?
2006/4/14, Alexei Zakharov <al...@gmail.com>:
> Hi Jacek,
>
> > It seems not to be a problem any more since we can be ensured you'll
> > back up our efforts, won't you? ;)
>
> Ok, I will try. :)
> Currently, ApacheDS guys argue about this bug: does it really ApacheDS
> bug? You may see the discussion at
> http://issues.apache.org/jira/browse/DIRSERVER-607
>
>
> 2006/4/11, Jacek Laskowski <ja...@laskowski.net.pl>:
> > On 4/11/06, Alexei Zakharov <al...@gmail.com> wrote:
> >
> > > However, they use different API since ApacheDS 1.0
> > > RC1, some packages are renamed. Some efforts will be required to
> > > rewrite the Geronimo portion of the code.
> >
> > Hi Alexei,
> >
> > It seems not to be a problem any more since we can be ensured you'll
> > back up our efforts, won't you? ;)
> >
> > Do you happen to know when they're going to release the fixed version?
> > Any estimates?
--
Alexei Zakharov,
Intel Middleware Product Division
Re: unit test failures
Posted by Alexei Zakharov <al...@gmail.com>.
Hi Jacek,
> It seems not to be a problem any more since we can be ensured you'll
> back up our efforts, won't you? ;)
Ok, I will try. :)
Currently, ApacheDS guys argue about this bug: does it really ApacheDS
bug? You may see the discussion at
http://issues.apache.org/jira/browse/DIRSERVER-607
2006/4/11, Jacek Laskowski <ja...@laskowski.net.pl>:
> On 4/11/06, Alexei Zakharov <al...@gmail.com> wrote:
>
> > However, they use different API since ApacheDS 1.0
> > RC1, some packages are renamed. Some efforts will be required to
> > rewrite the Geronimo portion of the code.
>
> Hi Alexei,
>
> It seems not to be a problem any more since we can be ensured you'll
> back up our efforts, won't you? ;)
>
> Do you happen to know when they're going to release the fixed version?
> Any estimates?
>
> Jacek
--
Alexei Zakharov,
Intel Middleware Product Division
Re: unit test failures
Posted by Jacek Laskowski <ja...@laskowski.net.pl>.
On 4/11/06, Alexei Zakharov <al...@gmail.com> wrote:
> However, they use different API since ApacheDS 1.0
> RC1, some packages are renamed. Some efforts will be required to
> rewrite the Geronimo portion of the code.
Hi Alexei,
It seems not to be a problem any more since we can be ensured you'll
back up our efforts, won't you? ;)
Do you happen to know when they're going to release the fixed version?
Any estimates?
Jacek
--
Jacek Laskowski
http://www.laskowski.org.pl
Re: unit test failures
Posted by Alexei Zakharov <al...@gmail.com>.
Hi Kevan,
> FYI: Apache DS is here -- http://directory.apache.org/subprojects/
> apacheds/index.html
The bug is reproducible on their latest stable build also. I think it
is best to wait until ApacheDS team fix the bug and incorporate their
patched build then. However, they use different API since ApacheDS 1.0
RC1, some packages are renamed. Some efforts will be required to
rewrite the Geronimo portion of the code.
2006/4/5, Alexei Zakharov <al...@gmail.com>:
> Kevan,
>
> > FYI: Apache DS is here -- http://directory.apache.org/subprojects/
>
> Thanks! I will try their recent builds.
>
> 2006/4/5, Kevan Miller <ke...@gmail.com>:
> >
> > On Apr 5, 2006, at 7:58 AM, Alexei Zakharov wrote:
> >
> > > Hi Kevan,
> > >
> > > This is follow-up to the story about BEA VMs. I have analyzed
> > > org.apache.geronimo.directory.RunningTest (modules/directory) failure
> > > on BEA JRockit. I've raised the JIRA issue #1805
> > > (http://issues.apache.org/jira/browse/GERONIMO-1805) with the detailed
> > > result of my investigation. In brief, IMHO there is some problem with
> > > LDAP server built in Geronimo. As far as I understand LDAP server is
> > > not part of Geronimo itself but is an independent component. It is
> > > part of some "apacheds-core" package. Therefore, it seems the problem
> > > is not a "pure" Geronimo problem. Do anybody knows from where apacheds
> > > comes from? Is it a separate Apache project?
> >
> > Hi Alexei,
> > Thanks! I haven't looked at the Jira, yet, but will soon...
> >
> > FYI: Apache DS is here -- http://directory.apache.org/subprojects/
> > apacheds/index.html
> >
> > --kevan
> >
> >
> > >
> > > This is follow-up to the story about BEA VMs. I have analized
> > >
> > > 2006/3/21, Kevan Miller <ke...@gmail.com>:
> > >>
> > >>
> > >> On Mar 20, 2006, at 1:19 PM, Alexei Zakharov wrote:
> > >>
> > >> Hi Kevan,
> > >>
> > >> I was lucky to build it using JRockit 1.4.2 + Maven 1.1 Beta.
> > >> Therefore,
> > >> this was a "pure JRockit" experiment.
> > >>
> > >> As for dependencies from Sun, AFAIK there are a few non-CORBA
> > >> places in
> > >> code with "com.sun" hardcoded. This issue has been rised recently
> > >> in the
> > >> list. You may check "Using non-Sun JNDI/RMI service provider"
> > >> thread for
> > >> details.
> > >>
> > >> Ok, I will continue investigation of these failures. BTW, if you are
> > >> interested in BEA 1.5 I attach the JRockit 1.5 failure log.
> > >> Messages are
> > >> direct JUnit messages. I was using JRockit 1.5.0-b64 win32 + maven
> > >> 1.0.2 to
> > >> obtain these results.
> > >>
> > >>
> > >> Hi Alexei,
> > >> Thanks. Let us know what you find.
> > >>
> > >> Looks like the majority of the 1.5 problems are "environmental"
> > >> something to
> > >> do with the junit environment on JRockit 1.5. The "kernel is
> > >> already running
> > >> this kernel name: geronimo.test" and "runFinalizer" messages seem
> > >> to imply
> > >> that you're seeing collisions between Junit tests on different
> > >> components...
> > >>
> > >> --kevan
> > >>
> > >>
> > >> Regards,
> > >> Alexei
> > >>
> > >> 2006/3/18, Kevan Miller <ke...@gmail.com>:
> > >>>
> > >>> Hi Alexei,
> > >>> Thanks for the information. Are you both building <and> running
> > >>> tests with
> > >> JRockit? Or are you building w/ Sun and only running tests under
> > >> JRockit?
> > >> Would be good to isolate the two. Good chance, however, that it
> > >> won't make
> > >> any difference...
> > >>>
> > >>>
> > >>> Geronimo currently has direct ties to the Sun 1.4 corba
> > >>> implementation.
> > >> That's the only JRE restriction that I'm aware of. People have run
> > >> Geronimo
> > >> on Sun 1.5 and also on IBM's 1.4 JRE.
> > >>>
> > >>>
> > >>> It seems that the tests you describe <ought> to work. There's a
> > >>> reasonable
> > >> chance that these are Geronimo bugs that are masked by our current
> > >> runtime.
> > >> So, it would be great if you could help us uncover the causes of
> > >> these
> > >> failures.
> > >>>
> > >>>
> > >>> I don't know why JRockit 1.5 would be worse.
> > >>>
> > >>>
> > >>> --kevan
> > >>>
> > >>>
> > >>>
> > >>> On Mar 17, 2006, at 12:21 PM, Alexei Zakharov wrote:
> > >>>
> > >>>
> > >>>
> > >>> Hi, community!
> > >>>
> > >>> I experiment with running Geronimo on various JVM's. The
> > >>> interesting thing
> > >> I've encountered is that some of unit tests fail on BEA Jrockit
> > >> VM. I've got
> > >> at least three failures unique to BEA Jrockit 1.4.2_04 VM:
> > >>>
> > >>> Module: modules/directory
> > >>> Test: org.apache.geronimo.directory.RunningTest
> > >>> Result: VM hangs
> > >>>
> > >>> Module: modules/timer
> > >>> Test:
> > >> org.apache.geronimo.timer.NontransactionalThreadPooledTimerTest.testT
> > >> asksInUnspecifiedTxContext
> > >>> Output: expected:<20> but was:<19>
> > >>>
> > >>> Module: modules/tomcat
> > >>> Test:
> > >> org.apache.geronimo.tomcat.JAASSecurityTest.testNotAuthorized
> > >>> Output: expected:<<!-- Login Page -->> but was:<null>
> > >>>
> > >>> These tests pass on all Sun VMs. Situation with Jrockit 1.5 is
> > >>> much worse
> > >> - more than 60 failures. It may result in overall instability
> > >> while running
> > >> Geronimo on BEA.
> > >>> Any comments & suggestions?
> > >>>
> > >>> Alexei Zakharov,
> > >>> Intel Middleware Product Division
> > >>>
> > >>>
> > >>
> > >>
> > >> <bea-1.5_geronimo-1.0.html>
> > >>
> > >>
> > >
> > >
> > > --
> > > Alexei Zakharov,
> > > Intel Middleware Product Division
> >
> >
>
>
> --
> Alexei Zakharov,
> Intel Middleware Product Division
>
--
Alexei Zakharov,
Intel Middleware Product Division
Re: unit test failures
Posted by Alexei Zakharov <al...@gmail.com>.
Kevan,
> FYI: Apache DS is here -- http://directory.apache.org/subprojects/
Thanks! I will try their recent builds.
2006/4/5, Kevan Miller <ke...@gmail.com>:
>
> On Apr 5, 2006, at 7:58 AM, Alexei Zakharov wrote:
>
> > Hi Kevan,
> >
> > This is follow-up to the story about BEA VMs. I have analyzed
> > org.apache.geronimo.directory.RunningTest (modules/directory) failure
> > on BEA JRockit. I've raised the JIRA issue #1805
> > (http://issues.apache.org/jira/browse/GERONIMO-1805) with the detailed
> > result of my investigation. In brief, IMHO there is some problem with
> > LDAP server built in Geronimo. As far as I understand LDAP server is
> > not part of Geronimo itself but is an independent component. It is
> > part of some "apacheds-core" package. Therefore, it seems the problem
> > is not a "pure" Geronimo problem. Do anybody knows from where apacheds
> > comes from? Is it a separate Apache project?
>
> Hi Alexei,
> Thanks! I haven't looked at the Jira, yet, but will soon...
>
> FYI: Apache DS is here -- http://directory.apache.org/subprojects/
> apacheds/index.html
>
> --kevan
>
>
> >
> > This is follow-up to the story about BEA VMs. I have analized
> >
> > 2006/3/21, Kevan Miller <ke...@gmail.com>:
> >>
> >>
> >> On Mar 20, 2006, at 1:19 PM, Alexei Zakharov wrote:
> >>
> >> Hi Kevan,
> >>
> >> I was lucky to build it using JRockit 1.4.2 + Maven 1.1 Beta.
> >> Therefore,
> >> this was a "pure JRockit" experiment.
> >>
> >> As for dependencies from Sun, AFAIK there are a few non-CORBA
> >> places in
> >> code with "com.sun" hardcoded. This issue has been rised recently
> >> in the
> >> list. You may check "Using non-Sun JNDI/RMI service provider"
> >> thread for
> >> details.
> >>
> >> Ok, I will continue investigation of these failures. BTW, if you are
> >> interested in BEA 1.5 I attach the JRockit 1.5 failure log.
> >> Messages are
> >> direct JUnit messages. I was using JRockit 1.5.0-b64 win32 + maven
> >> 1.0.2 to
> >> obtain these results.
> >>
> >>
> >> Hi Alexei,
> >> Thanks. Let us know what you find.
> >>
> >> Looks like the majority of the 1.5 problems are "environmental"
> >> something to
> >> do with the junit environment on JRockit 1.5. The "kernel is
> >> already running
> >> this kernel name: geronimo.test" and "runFinalizer" messages seem
> >> to imply
> >> that you're seeing collisions between Junit tests on different
> >> components...
> >>
> >> --kevan
> >>
> >>
> >> Regards,
> >> Alexei
> >>
> >> 2006/3/18, Kevan Miller <ke...@gmail.com>:
> >>>
> >>> Hi Alexei,
> >>> Thanks for the information. Are you both building <and> running
> >>> tests with
> >> JRockit? Or are you building w/ Sun and only running tests under
> >> JRockit?
> >> Would be good to isolate the two. Good chance, however, that it
> >> won't make
> >> any difference...
> >>>
> >>>
> >>> Geronimo currently has direct ties to the Sun 1.4 corba
> >>> implementation.
> >> That's the only JRE restriction that I'm aware of. People have run
> >> Geronimo
> >> on Sun 1.5 and also on IBM's 1.4 JRE.
> >>>
> >>>
> >>> It seems that the tests you describe <ought> to work. There's a
> >>> reasonable
> >> chance that these are Geronimo bugs that are masked by our current
> >> runtime.
> >> So, it would be great if you could help us uncover the causes of
> >> these
> >> failures.
> >>>
> >>>
> >>> I don't know why JRockit 1.5 would be worse.
> >>>
> >>>
> >>> --kevan
> >>>
> >>>
> >>>
> >>> On Mar 17, 2006, at 12:21 PM, Alexei Zakharov wrote:
> >>>
> >>>
> >>>
> >>> Hi, community!
> >>>
> >>> I experiment with running Geronimo on various JVM's. The
> >>> interesting thing
> >> I've encountered is that some of unit tests fail on BEA Jrockit
> >> VM. I've got
> >> at least three failures unique to BEA Jrockit 1.4.2_04 VM:
> >>>
> >>> Module: modules/directory
> >>> Test: org.apache.geronimo.directory.RunningTest
> >>> Result: VM hangs
> >>>
> >>> Module: modules/timer
> >>> Test:
> >> org.apache.geronimo.timer.NontransactionalThreadPooledTimerTest.testT
> >> asksInUnspecifiedTxContext
> >>> Output: expected:<20> but was:<19>
> >>>
> >>> Module: modules/tomcat
> >>> Test:
> >> org.apache.geronimo.tomcat.JAASSecurityTest.testNotAuthorized
> >>> Output: expected:<<!-- Login Page -->> but was:<null>
> >>>
> >>> These tests pass on all Sun VMs. Situation with Jrockit 1.5 is
> >>> much worse
> >> - more than 60 failures. It may result in overall instability
> >> while running
> >> Geronimo on BEA.
> >>> Any comments & suggestions?
> >>>
> >>> Alexei Zakharov,
> >>> Intel Middleware Product Division
> >>>
> >>>
> >>
> >>
> >> <bea-1.5_geronimo-1.0.html>
> >>
> >>
> >
> >
> > --
> > Alexei Zakharov,
> > Intel Middleware Product Division
>
>
--
Alexei Zakharov,
Intel Middleware Product Division
Re: unit test failures
Posted by Kevan Miller <ke...@gmail.com>.
On Apr 5, 2006, at 7:58 AM, Alexei Zakharov wrote:
> Hi Kevan,
>
> This is follow-up to the story about BEA VMs. I have analyzed
> org.apache.geronimo.directory.RunningTest (modules/directory) failure
> on BEA JRockit. I've raised the JIRA issue #1805
> (http://issues.apache.org/jira/browse/GERONIMO-1805) with the detailed
> result of my investigation. In brief, IMHO there is some problem with
> LDAP server built in Geronimo. As far as I understand LDAP server is
> not part of Geronimo itself but is an independent component. It is
> part of some "apacheds-core" package. Therefore, it seems the problem
> is not a "pure" Geronimo problem. Do anybody knows from where apacheds
> comes from? Is it a separate Apache project?
Hi Alexei,
Thanks! I haven't looked at the Jira, yet, but will soon...
FYI: Apache DS is here -- http://directory.apache.org/subprojects/
apacheds/index.html
--kevan
>
> This is follow-up to the story about BEA VMs. I have analized
>
> 2006/3/21, Kevan Miller <ke...@gmail.com>:
>>
>>
>> On Mar 20, 2006, at 1:19 PM, Alexei Zakharov wrote:
>>
>> Hi Kevan,
>>
>> I was lucky to build it using JRockit 1.4.2 + Maven 1.1 Beta.
>> Therefore,
>> this was a "pure JRockit" experiment.
>>
>> As for dependencies from Sun, AFAIK there are a few non-CORBA
>> places in
>> code with "com.sun" hardcoded. This issue has been rised recently
>> in the
>> list. You may check "Using non-Sun JNDI/RMI service provider"
>> thread for
>> details.
>>
>> Ok, I will continue investigation of these failures. BTW, if you are
>> interested in BEA 1.5 I attach the JRockit 1.5 failure log.
>> Messages are
>> direct JUnit messages. I was using JRockit 1.5.0-b64 win32 + maven
>> 1.0.2 to
>> obtain these results.
>>
>>
>> Hi Alexei,
>> Thanks. Let us know what you find.
>>
>> Looks like the majority of the 1.5 problems are "environmental"
>> something to
>> do with the junit environment on JRockit 1.5. The "kernel is
>> already running
>> this kernel name: geronimo.test" and "runFinalizer" messages seem
>> to imply
>> that you're seeing collisions between Junit tests on different
>> components...
>>
>> --kevan
>>
>>
>> Regards,
>> Alexei
>>
>> 2006/3/18, Kevan Miller <ke...@gmail.com>:
>>>
>>> Hi Alexei,
>>> Thanks for the information. Are you both building <and> running
>>> tests with
>> JRockit? Or are you building w/ Sun and only running tests under
>> JRockit?
>> Would be good to isolate the two. Good chance, however, that it
>> won't make
>> any difference...
>>>
>>>
>>> Geronimo currently has direct ties to the Sun 1.4 corba
>>> implementation.
>> That's the only JRE restriction that I'm aware of. People have run
>> Geronimo
>> on Sun 1.5 and also on IBM's 1.4 JRE.
>>>
>>>
>>> It seems that the tests you describe <ought> to work. There's a
>>> reasonable
>> chance that these are Geronimo bugs that are masked by our current
>> runtime.
>> So, it would be great if you could help us uncover the causes of
>> these
>> failures.
>>>
>>>
>>> I don't know why JRockit 1.5 would be worse.
>>>
>>>
>>> --kevan
>>>
>>>
>>>
>>> On Mar 17, 2006, at 12:21 PM, Alexei Zakharov wrote:
>>>
>>>
>>>
>>> Hi, community!
>>>
>>> I experiment with running Geronimo on various JVM's. The
>>> interesting thing
>> I've encountered is that some of unit tests fail on BEA Jrockit
>> VM. I've got
>> at least three failures unique to BEA Jrockit 1.4.2_04 VM:
>>>
>>> Module: modules/directory
>>> Test: org.apache.geronimo.directory.RunningTest
>>> Result: VM hangs
>>>
>>> Module: modules/timer
>>> Test:
>> org.apache.geronimo.timer.NontransactionalThreadPooledTimerTest.testT
>> asksInUnspecifiedTxContext
>>> Output: expected:<20> but was:<19>
>>>
>>> Module: modules/tomcat
>>> Test:
>> org.apache.geronimo.tomcat.JAASSecurityTest.testNotAuthorized
>>> Output: expected:<<!-- Login Page -->> but was:<null>
>>>
>>> These tests pass on all Sun VMs. Situation with Jrockit 1.5 is
>>> much worse
>> - more than 60 failures. It may result in overall instability
>> while running
>> Geronimo on BEA.
>>> Any comments & suggestions?
>>>
>>> Alexei Zakharov,
>>> Intel Middleware Product Division
>>>
>>>
>>
>>
>> <bea-1.5_geronimo-1.0.html>
>>
>>
>
>
> --
> Alexei Zakharov,
> Intel Middleware Product Division
Re: unit test failures
Posted by Alexei Zakharov <al...@gmail.com>.
Hi Kevan,
This is follow-up to the story about BEA VMs. I have analyzed
org.apache.geronimo.directory.RunningTest (modules/directory) failure
on BEA JRockit. I've raised the JIRA issue #1805
(http://issues.apache.org/jira/browse/GERONIMO-1805) with the detailed
result of my investigation. In brief, IMHO there is some problem with
LDAP server built in Geronimo. As far as I understand LDAP server is
not part of Geronimo itself but is an independent component. It is
part of some "apacheds-core" package. Therefore, it seems the problem
is not a "pure" Geronimo problem. Do anybody knows from where apacheds
comes from? Is it a separate Apache project?
This is follow-up to the story about BEA VMs. I have analized
2006/3/21, Kevan Miller <ke...@gmail.com>:
>
>
> On Mar 20, 2006, at 1:19 PM, Alexei Zakharov wrote:
>
> Hi Kevan,
>
> I was lucky to build it using JRockit 1.4.2 + Maven 1.1 Beta. Therefore,
> this was a "pure JRockit" experiment.
>
> As for dependencies from Sun, AFAIK there are a few non-CORBA places in
> code with "com.sun" hardcoded. This issue has been rised recently in the
> list. You may check "Using non-Sun JNDI/RMI service provider" thread for
> details.
>
> Ok, I will continue investigation of these failures. BTW, if you are
> interested in BEA 1.5 I attach the JRockit 1.5 failure log. Messages are
> direct JUnit messages. I was using JRockit 1.5.0-b64 win32 + maven 1.0.2 to
> obtain these results.
>
>
> Hi Alexei,
> Thanks. Let us know what you find.
>
> Looks like the majority of the 1.5 problems are "environmental" something to
> do with the junit environment on JRockit 1.5. The "kernel is already running
> this kernel name: geronimo.test" and "runFinalizer" messages seem to imply
> that you're seeing collisions between Junit tests on different components...
>
> --kevan
>
>
> Regards,
> Alexei
>
> 2006/3/18, Kevan Miller <ke...@gmail.com>:
> >
> > Hi Alexei,
> > Thanks for the information. Are you both building <and> running tests with
> JRockit? Or are you building w/ Sun and only running tests under JRockit?
> Would be good to isolate the two. Good chance, however, that it won't make
> any difference...
> >
> >
> > Geronimo currently has direct ties to the Sun 1.4 corba implementation.
> That's the only JRE restriction that I'm aware of. People have run Geronimo
> on Sun 1.5 and also on IBM's 1.4 JRE.
> >
> >
> > It seems that the tests you describe <ought> to work. There's a reasonable
> chance that these are Geronimo bugs that are masked by our current runtime.
> So, it would be great if you could help us uncover the causes of these
> failures.
> >
> >
> > I don't know why JRockit 1.5 would be worse.
> >
> >
> > --kevan
> >
> >
> >
> > On Mar 17, 2006, at 12:21 PM, Alexei Zakharov wrote:
> >
> >
> >
> > Hi, community!
> >
> > I experiment with running Geronimo on various JVM's. The interesting thing
> I've encountered is that some of unit tests fail on BEA Jrockit VM. I've got
> at least three failures unique to BEA Jrockit 1.4.2_04 VM:
> >
> > Module: modules/directory
> > Test: org.apache.geronimo.directory.RunningTest
> > Result: VM hangs
> >
> > Module: modules/timer
> > Test:
> org.apache.geronimo.timer.NontransactionalThreadPooledTimerTest.testTasksInUnspecifiedTxContext
> > Output: expected:<20> but was:<19>
> >
> > Module: modules/tomcat
> > Test:
> org.apache.geronimo.tomcat.JAASSecurityTest.testNotAuthorized
> > Output: expected:<<!-- Login Page -->> but was:<null>
> >
> > These tests pass on all Sun VMs. Situation with Jrockit 1.5 is much worse
> - more than 60 failures. It may result in overall instability while running
> Geronimo on BEA.
> > Any comments & suggestions?
> >
> > Alexei Zakharov,
> > Intel Middleware Product Division
> >
> >
>
>
> <bea-1.5_geronimo-1.0.html>
>
>
--
Alexei Zakharov,
Intel Middleware Product Division
Re: unit test failures
Posted by Kevan Miller <ke...@gmail.com>.
On Mar 20, 2006, at 1:19 PM, Alexei Zakharov wrote:
> Hi Kevan,
>
> I was lucky to build it using JRockit 1.4.2 + Maven 1.1 Beta.
> Therefore, this was a "pure JRockit" experiment.
>
> As for dependencies from Sun, AFAIK there are a few non-CORBA
> places in code with "com.sun" hardcoded. This issue has been rised
> recently in the list. You may check "Using non-Sun JNDI/RMI service
> provider" thread for details.
>
> Ok, I will continue investigation of these failures. BTW, if you
> are interested in BEA 1.5 I attach the JRockit 1.5 failure log.
> Messages are direct JUnit messages. I was using JRockit 1.5.0-b64
> win32 + maven 1.0.2 to obtain these results.
Hi Alexei,
Thanks. Let us know what you find.
Looks like the majority of the 1.5 problems are "environmental"
something to do with the junit environment on JRockit 1.5. The
"kernel is already running this kernel name: geronimo.test" and
"runFinalizer" messages seem to imply that you're seeing collisions
between Junit tests on different components...
--kevan
>
> Regards,
> Alexei
>
> 2006/3/18, Kevan Miller <ke...@gmail.com>:
> Hi Alexei,
> Thanks for the information. Are you both building <and> running
> tests with JRockit? Or are you building w/ Sun and only running
> tests under JRockit? Would be good to isolate the two. Good chance,
> however, that it won't make any difference...
>
>
> Geronimo currently has direct ties to the Sun 1.4 corba
> implementation. That's the only JRE restriction that I'm aware of.
> People have run Geronimo on Sun 1.5 and also on IBM's 1.4 JRE.
>
>
> It seems that the tests you describe <ought> to work. There's a
> reasonable chance that these are Geronimo bugs that are masked by
> our current runtime. So, it would be great if you could help us
> uncover the causes of these failures.
>
>
> I don't know why JRockit 1.5 would be worse.
>
>
> --kevan
>
> On Mar 17, 2006, at 12:21 PM, Alexei Zakharov wrote:
>
>> Hi, community!
>>
>> I experiment with running Geronimo on various JVM's. The
>> interesting thing I've encountered is that some of unit tests fail
>> on BEA Jrockit VM. I've got at least three failures unique to BEA
>> Jrockit 1.4.2_04 VM:
>>
>> Module: modules/directory
>> Test: org.apache.geronimo.directory.RunningTest
>> Result: VM hangs
>>
>> Module: modules/timer
>> Test:
>> org.apache.geronimo.timer.NontransactionalThreadPooledTimerTest.testT
>> asksInUnspecifiedTxContext
>> Output: expected:<20> but was:<19>
>>
>> Module: modules/tomcat
>> Test: org.apache.geronimo.tomcat.JAASSecurityTest.testNotAuthorized
>> Output: expected:<<!-- Login Page -->> but was:<null>
>>
>> These tests pass on all Sun VMs. Situation with Jrockit 1.5 is
>> much worse - more than 60 failures. It may result in overall
>> instability while running Geronimo on BEA.
>> Any comments & suggestions?
>>
>> Alexei Zakharov,
>> Intel Middleware Product Division
>
>
> <bea-1.5_geronimo-1.0.html>
Re: unit test failures
Posted by Alexei Zakharov <al...@gmail.com>.
Hi Kevan,
I was lucky to build it using JRockit 1.4.2 + Maven 1.1 Beta. Therefore,
this was a "pure JRockit" experiment.
As for dependencies from Sun, AFAIK there are a few non-CORBA places in
code with "com.sun" hardcoded. This issue has been rised recently in the
list. You may check "Using non-Sun JNDI/RMI service provider" thread for
details.
Ok, I will continue investigation of these failures. BTW, if you are
interested in BEA 1.5 I attach the JRockit 1.5 failure log. Messages are
direct JUnit messages. I was using JRockit 1.5.0-b64 win32 + maven 1.0.2 to
obtain these results.
Regards,
Alexei
2006/3/18, Kevan Miller <ke...@gmail.com>:
>
> Hi Alexei,
> Thanks for the information. Are you both building <and> running tests with
> JRockit? Or are you building w/ Sun and only running tests under JRockit?
> Would be good to isolate the two. Good chance, however, that it won't make
> any difference...
>
>
> Geronimo currently has direct ties to the Sun 1.4 corba implementation.
> That's the only JRE restriction that I'm aware of. People have run Geronimo
> on Sun 1.5 and also on IBM's 1.4 JRE.
>
>
> It seems that the tests you describe <ought> to work. There's a reasonable
> chance that these are Geronimo bugs that are masked by our current runtime.
> So, it would be great if you could help us uncover the causes of these
> failures.
>
>
> I don't know why JRockit 1.5 would be worse.
>
>
> --kevan
>
> On Mar 17, 2006, at 12:21 PM, Alexei Zakharov wrote:
>
> Hi, community!
>
> I experiment with running Geronimo on various JVM's. The interesting thing
> I've encountered is that some of unit tests fail on BEA Jrockit VM. I've got
> at least three failures unique to BEA Jrockit 1.4.2_04 VM:
>
> Module: modules/directory
> Test: org.apache.geronimo.directory.RunningTest
> Result: VM hangs
>
> Module: modules/timer
> Test:
> org.apache.geronimo.timer.NontransactionalThreadPooledTimerTest.testTasksInUnspecifiedTxContext
> Output: expected:<20> but was:<19>
>
> Module: modules/tomcat
> Test: org.apache.geronimo.tomcat.JAASSecurityTest.testNotAuthorized
> Output: expected:<<!-- Login Page -->> but was:<null>
>
> These tests pass on all Sun VMs. Situation with Jrockit 1.5 is much worse
> - more than 60 failures. It may result in overall instability while running
> Geronimo on BEA.
> Any comments & suggestions?
>
> Alexei Zakharov,
> Intel Middleware Product Division
>
>
>
Re: unit test failures
Posted by Kevan Miller <ke...@gmail.com>.
Hi Alexei,
Thanks for the information. Are you both building <and> running tests
with JRockit? Or are you building w/ Sun and only running tests under
JRockit? Would be good to isolate the two. Good chance, however, that
it won't make any difference...
Geronimo currently has direct ties to the Sun 1.4 corba
implementation. That's the only JRE restriction that I'm aware of.
People have run Geronimo on Sun 1.5 and also on IBM's 1.4 JRE.
It seems that the tests you describe <ought> to work. There's a
reasonable chance that these are Geronimo bugs that are masked by our
current runtime. So, it would be great if you could help us uncover
the causes of these failures.
I don't know why JRockit 1.5 would be worse.
--kevan
On Mar 17, 2006, at 12:21 PM, Alexei Zakharov wrote:
> Hi, community!
>
> I experiment with running Geronimo on various JVM's. The
> interesting thing I've encountered is that some of unit tests fail
> on BEA Jrockit VM. I've got at least three failures unique to BEA
> Jrockit 1.4.2_04 VM:
>
> Module: modules/directory
> Test: org.apache.geronimo.directory.RunningTest
> Result: VM hangs
>
> Module: modules/timer
> Test:
> org.apache.geronimo.timer.NontransactionalThreadPooledTimerTest.testTa
> sksInUnspecifiedTxContext
> Output: expected:<20> but was:<19>
>
> Module: modules/tomcat
> Test: org.apache.geronimo.tomcat.JAASSecurityTest.testNotAuthorized
> Output: expected:<<!-- Login Page -->> but was:<null>
>
> These tests pass on all Sun VMs. Situation with Jrockit 1.5 is much
> worse - more than 60 failures. It may result in overall instability
> while running Geronimo on BEA.
> Any comments & suggestions?
>
> Alexei Zakharov,
> Intel Middleware Product Division