Skip to main content

My latest updates

8 replies [Last post]
Anonymous

Hello,

I have incorporated my changes from
branches/users/msokolnikov/Issue_70 in the new more organized way.

They are in the new branch:

http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=940
http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=941
http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=943
http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=948

Unfortunately I had no time to verify the fixes. I will do later.
I am going on vacation until next Wednesday

Thank you,
Maxim

---------------------------------------------------------------------
To unsubscribe, e-mail: meframework-unsubscribe@cqme.dev.java.net
For additional commands, e-mail: meframework-help@cqme.dev.java.net

Reply viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.
Maxim V. Sokolnikov

Vladimir, Alexander,

I have fixed the following issues:
#166: Request.COMPLETE_AND_TERMINATE is not supported
Revision 1083 (http://fisheye4.cenqua.com/changelog/cqme/?cs=1083)
#167: AgentXlet and AgentApplet should be created for the new genetric communication agent
Revision 1143 (http://fisheye4.cenqua.com/changelog/cqme/?cs=1143)

I have tried reuse AgentApplet from JavaTest, but it requires
duplication and modification a lot of classes, therefore I have
created AgentApplet from scratch. It is not completed yet. I have to
complete some minor fixes (improved applet parameter support, right
now only GUI is supported) and update make files. I can do it only after
vacation.

Could you review the changes?

Thank you,
Maxim

Maxim V. Sokolnikov writes:
> Vladimir,
>
> I have started the integration of the generic agent functionality to
> the AgentXlet and AgentApplet wrapper.
> Before integrating of the my updates I want to double check that my
> approach is right.
>
> 1. J2ME TCK Framework defines the com.sun.tck.xlet.agent.BaseXlet,
> which extends AgentMain from the JavaTest. I am going to change the
> superclass of this class to the com.sun.tck.j2me.agent.AgentMain.
> It causes that the BaseXlet and their subclasses will use new generic
> communication framework and old JavaTest agent will not be supported
> anymore.
>
> 2. I will create AgentApplet wrapper for the
> com.sun.tck.j2me.agent.AgentMain.
>
>
> Also I found that PP TCK uses own version of the AgentXlet. As I
> understand it supports additional functionality for inter xlet
> communication.
>
> Do we need to integrate this functionality from PP TCK to the J2ME TCK
> Framework?
>
> Thank you,
> Maxim
>
> Vladimir Sizikov writes:
> > Maxim,
> >
> > Alexander and I spent some time reviewing and fixing the code here and
> > there, and testing, and lots of code style formattings, javadoc
> > updates and typo fixes. :)
> >
> > Eventually, I've commit the latest version into the trunk.
> >
> > The code works, and in our limited testisg, works good.
> > Sitll, there is a lot to done, and some unresolved issues, comments.
> >
> > Please find below my comments:
> >
> > 1. SimplePipe.DEFAULT_NAME is with wrong package name
> >
> > 2. ServerExecutionService javadoc refers to non-public APIs, like
> > RegistrationHandler and setRegistrationHandler()
> >
> > 3. ServerProactiveCommand does the same.
> >
> > 4. Response has public fields!
> >
> > 5. AgentMain, this very important class, has no public class-level
> > javadocs.
> >
> > 6. Agent-Wrappers (Applet, Xlet) - should be updated to use the new
> > base agent, right?
> >
> > 7. Javadoc warnings should be eliminated (at least those that are from
> > your updates).
> >
> > ====================== Previous comments ========================
> >
> > 1. GenericAgentCommand.setClasspath() is GLOBAL while the classpath
> > values are test-local!
> >
> > 2. GenericAgentCommand: NotifiedTestInfo.waitUntilDone() without the
> > timeout. Possibility for blocking forever!
> >
> > 3. GenericAgentCommand: Javadocs mention "ExecutionRequestQueue" few
> > times, but there is no such thing.
> >
> > 4. Why do we need AgentMonitoringService, and all its dependencies
> > (lots)? Can we get away without it?
> >
> > Thanks,
> > --Vladimir
> >
> > On Wed, May 23, 2007 at 08:11:36PM -0700, Maxim Sokolnikov wrote:
> > > Vladimir, Alexander,
> > >
> > > I have fixed this problem in Revision 990. I also fixed other issues.
> > > Could you please review? (http://fisheye4.cenqua.com/changelog/cqme/?cs=990)
> > >
> > > Thank you,
> > > Maxim
> > >
> > >
> > > Vladimir Sizikov writes:
> > > > Hi Maxim,
> > > >
> > > > One major comment I have so far is that the test execution with the
> > > > new agents is very unstable an the moment.
> > > >
> > > > Typically, I cannot execute tests more than once (and sometimes even
> > > > more than one test!). Most of the time I get
> > > > Error. RegistrationHandler.NO_AGENTS_AVAILABLE
> > > >
> > > > This is clearly a showstopper and we cannot commit the new code with
> > > > this issue into the trunk.
> > > >
> > > > Please, take a look as soon as you can.
> > > >
> > > > Both Alexander and I use SATSA TCK as our driving vehicle, in CDC
> > > > mode. We've changed the sources in your branch to be able to run the
> > > > TCK.
> > > >
> > > > Another big issue - sometimes when I start the test run the second
> > > > time, I see LOTS AND LOTS of agents registerations and deregistrations
> > > > (hundreds), with 100% CPU consumed, and the end result is always the
> > > > dreaded RegistrationHandler.NO_AGENTS_AVAILABLE.
> > > >
> > > > Here are some minor (compared to the above) issues:
> > > > 1. GenericAgentCommand.setClasspath() is GLOBAL while the classpath
> > > > values are test-local!
> > > >
> > > > 2. GenericAgentCommand: NotifiedTestInfo.waitUntilDone() without the
> > > > timeout. Possibility for blocking forever!
> > > >
> > > > 3. GenericAgentCommand: Javadocs mention "ExecutionRequestQueue" few
> > > > times,
> > > > but there is no such thing.
> > > >
> > > > 4. Why do we need AgentMonitoringService, and all its dependencies
> > > > (lots)?
> > > > Can we get away without it?
> > > >
> > > > 5. Very non-standard java style in most of the sources. I've mass
> > > > reformatted the source (please take a look and take a note :) ).
> > > >
> > > > 6. Lots of places where Exceptions are just swallowed with TODO
> > > > comment. Very hard to debug problems.
> > > >
> > > > Thanks,
> > > > --Vladimir
> > > >
> > > > On Mon, May 21, 2007 at 07:29:22PM +0400, Alexander Alexeev wrote:
> > > > > Hi Maxim, Vladimir,
> > > > >
> > > > > two more comments:
> > > > >
> > > > > 1. The exception is thrown on JavaTest side when test's execution finished.
> > > > > I've
> > > > > submitted an issue: https://cqme.dev.java.net/issues/show_bug.cgi?id=151
> > > > >
> > > > > 2. Frequently second run still fails with error:
> > > > > Error. RegistrationHandler.NO_AGENTS_AVAILABLE
> > > > >
> > > > > Thanks,
> > > > > Alexander
> > > > >
> > > > > Maxim V. Sokolnikov wrote:
> > > > > >Hello,
> > > > > >
> > > > > >I have incorporated my changes from
> > > > > >branches/users/msokolnikov/Issue_70 in the new more organized way.
> > > > > >
> > > > > >They are in the new branch:
> > > > > >
> > > > > >http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=940
> > > > > >http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=941
> > > > > >http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=943
> > > > > >http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=948
> > > > > >
> > > > > >
> > > > > >Unfortunately I had no time to verify the fixes. I will do later.
> > > > > >I am going on vacation until next Wednesday
> > > > > >
> > > > > >Thank you,
> > > > > >Maxim
> > > > > >
> > > > > >---------------------------------------------------------------------
> > > > > >To unsubscribe, e-mail: meframework-unsubscribe@cqme.dev.java.net
> > > > > >For additional commands, e-mail: meframework-help@cqme.dev.java.net
> > > > > >
> > > > >

---------------------------------------------------------------------
To unsubscribe, e-mail: meframework-unsubscribe@cqme.dev.java.net
For additional commands, e-mail: meframework-help@cqme.dev.java.net

Alexander Alexeev

Hi Maxim, Vladimir,

two more comments:

1. The exception is thrown on JavaTest side when test's execution finished. I've
submitted an issue: https://cqme.dev.java.net/issues/show_bug.cgi?id=151

2. Frequently second run still fails with error:
Error. RegistrationHandler.NO_AGENTS_AVAILABLE

Thanks,
Alexander

Maxim V. Sokolnikov wrote:
> Hello,
>
> I have incorporated my changes from
> branches/users/msokolnikov/Issue_70 in the new more organized way.
>
> They are in the new branch:
>
> http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=940
> http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=941
> http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=943
> http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=948
>
>
> Unfortunately I had no time to verify the fixes. I will do later.
> I am going on vacation until next Wednesday
>
> Thank you,
> Maxim
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: meframework-unsubscribe@cqme.dev.java.net
> For additional commands, e-mail: meframework-help@cqme.dev.java.net
>

---------------------------------------------------------------------
To unsubscribe, e-mail: meframework-unsubscribe@cqme.dev.java.net
For additional commands, e-mail: meframework-help@cqme.dev.java.net

Vladimir Sizikov

Hi Maxim,

One major comment I have so far is that the test execution with the
new agents is very unstable an the moment.

Typically, I cannot execute tests more than once (and sometimes even
more than one test!). Most of the time I get
Error. RegistrationHandler.NO_AGENTS_AVAILABLE

This is clearly a showstopper and we cannot commit the new code with
this issue into the trunk.

Please, take a look as soon as you can.

Both Alexander and I use SATSA TCK as our driving vehicle, in CDC
mode. We've changed the sources in your branch to be able to run the
TCK.

Another big issue - sometimes when I start the test run the second
time, I see LOTS AND LOTS of agents registerations and deregistrations
(hundreds), with 100% CPU consumed, and the end result is always the
dreaded RegistrationHandler.NO_AGENTS_AVAILABLE.

Here are some minor (compared to the above) issues:
1. GenericAgentCommand.setClasspath() is GLOBAL while the classpath
values are test-local!

2. GenericAgentCommand: NotifiedTestInfo.waitUntilDone() without the
timeout. Possibility for blocking forever!

3. GenericAgentCommand: Javadocs mention "ExecutionRequestQueue" few
times,
but there is no such thing.

4. Why do we need AgentMonitoringService, and all its dependencies
(lots)?
Can we get away without it?

5. Very non-standard java style in most of the sources. I've mass
reformatted the source (please take a look and take a note :) ).

6. Lots of places where Exceptions are just swallowed with TODO
comment. Very hard to debug problems.

Thanks,
--Vladimir

On Mon, May 21, 2007 at 07:29:22PM +0400, Alexander Alexeev wrote:
> Hi Maxim, Vladimir,
>
> two more comments:
>
> 1. The exception is thrown on JavaTest side when test's execution finished.
> I've
> submitted an issue: https://cqme.dev.java.net/issues/show_bug.cgi?id=151
>
> 2. Frequently second run still fails with error:
> Error. RegistrationHandler.NO_AGENTS_AVAILABLE
>
> Thanks,
> Alexander
>
> Maxim V. Sokolnikov wrote:
> >Hello,
> >
> >I have incorporated my changes from
> >branches/users/msokolnikov/Issue_70 in the new more organized way.
> >
> >They are in the new branch:
> >
> >http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=940
> >http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=941
> >http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=943
> >http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=948
> >
> >
> >Unfortunately I had no time to verify the fixes. I will do later.
> >I am going on vacation until next Wednesday
> >
> >Thank you,
> >Maxim
> >
> >---------------------------------------------------------------------
> >To unsubscribe, e-mail: meframework-unsubscribe@cqme.dev.java.net
> >For additional commands, e-mail: meframework-help@cqme.dev.java.net
> >
>

---------------------------------------------------------------------
To unsubscribe, e-mail: meframework-unsubscribe@cqme.dev.java.net
For additional commands, e-mail: meframework-help@cqme.dev.java.net

Maxim Sokolnikov

Vladimir, Alexander,

I have fixed this problem in Revision 990. I also fixed other issues.
Could you please review? (http://fisheye4.cenqua.com/changelog/cqme/?cs=990)

Thank you,
Maxim

Vladimir Sizikov writes:
> Hi Maxim,
>
> One major comment I have so far is that the test execution with the
> new agents is very unstable an the moment.
>
> Typically, I cannot execute tests more than once (and sometimes even
> more than one test!). Most of the time I get
> Error. RegistrationHandler.NO_AGENTS_AVAILABLE
>
> This is clearly a showstopper and we cannot commit the new code with
> this issue into the trunk.
>
> Please, take a look as soon as you can.
>
> Both Alexander and I use SATSA TCK as our driving vehicle, in CDC
> mode. We've changed the sources in your branch to be able to run the
> TCK.
>
> Another big issue - sometimes when I start the test run the second
> time, I see LOTS AND LOTS of agents registerations and deregistrations
> (hundreds), with 100% CPU consumed, and the end result is always the
> dreaded RegistrationHandler.NO_AGENTS_AVAILABLE.
>
> Here are some minor (compared to the above) issues:
> 1. GenericAgentCommand.setClasspath() is GLOBAL while the classpath
> values are test-local!
>
> 2. GenericAgentCommand: NotifiedTestInfo.waitUntilDone() without the
> timeout. Possibility for blocking forever!
>
> 3. GenericAgentCommand: Javadocs mention "ExecutionRequestQueue" few
> times,
> but there is no such thing.
>
> 4. Why do we need AgentMonitoringService, and all its dependencies
> (lots)?
> Can we get away without it?
>
> 5. Very non-standard java style in most of the sources. I've mass
> reformatted the source (please take a look and take a note :) ).
>
> 6. Lots of places where Exceptions are just swallowed with TODO
> comment. Very hard to debug problems.
>
> Thanks,
> --Vladimir
>
> On Mon, May 21, 2007 at 07:29:22PM +0400, Alexander Alexeev wrote:
> > Hi Maxim, Vladimir,
> >
> > two more comments:
> >
> > 1. The exception is thrown on JavaTest side when test's execution finished.
> > I've
> > submitted an issue: https://cqme.dev.java.net/issues/show_bug.cgi?id=151
> >
> > 2. Frequently second run still fails with error:
> > Error. RegistrationHandler.NO_AGENTS_AVAILABLE
> >
> > Thanks,
> > Alexander
> >
> > Maxim V. Sokolnikov wrote:
> > >Hello,
> > >
> > >I have incorporated my changes from
> > >branches/users/msokolnikov/Issue_70 in the new more organized way.
> > >
> > >They are in the new branch:
> > >
> > >http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=940
> > >http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=941
> > >http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=943
> > >http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=948
> > >
> > >
> > >Unfortunately I had no time to verify the fixes. I will do later.
> > >I am going on vacation until next Wednesday
> > >
> > >Thank you,
> > >Maxim
> > >
> > >---------------------------------------------------------------------
> > >To unsubscribe, e-mail: meframework-unsubscribe@cqme.dev.java.net
> > >For additional commands, e-mail: meframework-help@cqme.dev.java.net
> > >
> >

---------------------------------------------------------------------
To unsubscribe, e-mail: meframework-unsubscribe@cqme.dev.java.net
For additional commands, e-mail: meframework-help@cqme.dev.java.net

Vladimir Sizikov

Maxim,

Alexander and I spent some time reviewing and fixing the code here and
there, and testing, and lots of code style formattings, javadoc
updates and typo fixes. :)

Eventually, I've commit the latest version into the trunk.

The code works, and in our limited testisg, works good.
Sitll, there is a lot to done, and some unresolved issues, comments.

Please find below my comments:

1. SimplePipe.DEFAULT_NAME is with wrong package name

2. ServerExecutionService javadoc refers to non-public APIs, like
RegistrationHandler and setRegistrationHandler()

3. ServerProactiveCommand does the same.

4. Response has public fields!

5. AgentMain, this very important class, has no public class-level
javadocs.

6. Agent-Wrappers (Applet, Xlet) - should be updated to use the new
base agent, right?

7. Javadoc warnings should be eliminated (at least those that are from
your updates).

====================== Previous comments ========================

1. GenericAgentCommand.setClasspath() is GLOBAL while the classpath
values are test-local!

2. GenericAgentCommand: NotifiedTestInfo.waitUntilDone() without the
timeout. Possibility for blocking forever!

3. GenericAgentCommand: Javadocs mention "ExecutionRequestQueue" few
times, but there is no such thing.

4. Why do we need AgentMonitoringService, and all its dependencies
(lots)? Can we get away without it?

Thanks,
--Vladimir

On Wed, May 23, 2007 at 08:11:36PM -0700, Maxim Sokolnikov wrote:
> Vladimir, Alexander,
>
> I have fixed this problem in Revision 990. I also fixed other issues.
> Could you please review? (http://fisheye4.cenqua.com/changelog/cqme/?cs=990)
>
> Thank you,
> Maxim
>
>
> Vladimir Sizikov writes:
> > Hi Maxim,
> >
> > One major comment I have so far is that the test execution with the
> > new agents is very unstable an the moment.
> >
> > Typically, I cannot execute tests more than once (and sometimes even
> > more than one test!). Most of the time I get
> > Error. RegistrationHandler.NO_AGENTS_AVAILABLE
> >
> > This is clearly a showstopper and we cannot commit the new code with
> > this issue into the trunk.
> >
> > Please, take a look as soon as you can.
> >
> > Both Alexander and I use SATSA TCK as our driving vehicle, in CDC
> > mode. We've changed the sources in your branch to be able to run the
> > TCK.
> >
> > Another big issue - sometimes when I start the test run the second
> > time, I see LOTS AND LOTS of agents registerations and deregistrations
> > (hundreds), with 100% CPU consumed, and the end result is always the
> > dreaded RegistrationHandler.NO_AGENTS_AVAILABLE.
> >
> > Here are some minor (compared to the above) issues:
> > 1. GenericAgentCommand.setClasspath() is GLOBAL while the classpath
> > values are test-local!
> >
> > 2. GenericAgentCommand: NotifiedTestInfo.waitUntilDone() without the
> > timeout. Possibility for blocking forever!
> >
> > 3. GenericAgentCommand: Javadocs mention "ExecutionRequestQueue" few
> > times,
> > but there is no such thing.
> >
> > 4. Why do we need AgentMonitoringService, and all its dependencies
> > (lots)?
> > Can we get away without it?
> >
> > 5. Very non-standard java style in most of the sources. I've mass
> > reformatted the source (please take a look and take a note :) ).
> >
> > 6. Lots of places where Exceptions are just swallowed with TODO
> > comment. Very hard to debug problems.
> >
> > Thanks,
> > --Vladimir
> >
> > On Mon, May 21, 2007 at 07:29:22PM +0400, Alexander Alexeev wrote:
> > > Hi Maxim, Vladimir,
> > >
> > > two more comments:
> > >
> > > 1. The exception is thrown on JavaTest side when test's execution finished.
> > > I've
> > > submitted an issue: https://cqme.dev.java.net/issues/show_bug.cgi?id=151
> > >
> > > 2. Frequently second run still fails with error:
> > > Error. RegistrationHandler.NO_AGENTS_AVAILABLE
> > >
> > > Thanks,
> > > Alexander
> > >
> > > Maxim V. Sokolnikov wrote:
> > > >Hello,
> > > >
> > > >I have incorporated my changes from
> > > >branches/users/msokolnikov/Issue_70 in the new more organized way.
> > > >
> > > >They are in the new branch:
> > > >
> > > >http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=940
> > > >http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=941
> > > >http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=943
> > > >http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=948
> > > >
> > > >
> > > >Unfortunately I had no time to verify the fixes. I will do later.
> > > >I am going on vacation until next Wednesday
> > > >
> > > >Thank you,
> > > >Maxim
> > > >
> > > >---------------------------------------------------------------------
> > > >To unsubscribe, e-mail: meframework-unsubscribe@cqme.dev.java.net
> > > >For additional commands, e-mail: meframework-help@cqme.dev.java.net
> > > >
> > >

---------------------------------------------------------------------
To unsubscribe, e-mail: meframework-unsubscribe@cqme.dev.java.net
For additional commands, e-mail: meframework-help@cqme.dev.java.net

Maxim V. Sokolnikov

Vladimir,

I have started the integration of the generic agent functionality to
the AgentXlet and AgentApplet wrapper.
Before integrating of the my updates I want to double check that my
approach is right.

1. J2ME TCK Framework defines the com.sun.tck.xlet.agent.BaseXlet,
which extends AgentMain from the JavaTest. I am going to change the
superclass of this class to the com.sun.tck.j2me.agent.AgentMain.
It causes that the BaseXlet and their subclasses will use new generic
communication framework and old JavaTest agent will not be supported
anymore.

2. I will create AgentApplet wrapper for the
com.sun.tck.j2me.agent.AgentMain.

Also I found that PP TCK uses own version of the AgentXlet. As I
understand it supports additional functionality for inter xlet
communication.

Do we need to integrate this functionality from PP TCK to the J2ME TCK
Framework?

Thank you,
Maxim

Vladimir Sizikov writes:
> Maxim,
>
> Alexander and I spent some time reviewing and fixing the code here and
> there, and testing, and lots of code style formattings, javadoc
> updates and typo fixes. :)
>
> Eventually, I've commit the latest version into the trunk.
>
> The code works, and in our limited testisg, works good.
> Sitll, there is a lot to done, and some unresolved issues, comments.
>
> Please find below my comments:
>
> 1. SimplePipe.DEFAULT_NAME is with wrong package name
>
> 2. ServerExecutionService javadoc refers to non-public APIs, like
> RegistrationHandler and setRegistrationHandler()
>
> 3. ServerProactiveCommand does the same.
>
> 4. Response has public fields!
>
> 5. AgentMain, this very important class, has no public class-level
> javadocs.
>
> 6. Agent-Wrappers (Applet, Xlet) - should be updated to use the new
> base agent, right?
>
> 7. Javadoc warnings should be eliminated (at least those that are from
> your updates).
>
> ====================== Previous comments ========================
>
> 1. GenericAgentCommand.setClasspath() is GLOBAL while the classpath
> values are test-local!
>
> 2. GenericAgentCommand: NotifiedTestInfo.waitUntilDone() without the
> timeout. Possibility for blocking forever!
>
> 3. GenericAgentCommand: Javadocs mention "ExecutionRequestQueue" few
> times, but there is no such thing.
>
> 4. Why do we need AgentMonitoringService, and all its dependencies
> (lots)? Can we get away without it?
>
> Thanks,
> --Vladimir
>
> On Wed, May 23, 2007 at 08:11:36PM -0700, Maxim Sokolnikov wrote:
> > Vladimir, Alexander,
> >
> > I have fixed this problem in Revision 990. I also fixed other issues.
> > Could you please review? (http://fisheye4.cenqua.com/changelog/cqme/?cs=990)
> >
> > Thank you,
> > Maxim
> >
> >
> > Vladimir Sizikov writes:
> > > Hi Maxim,
> > >
> > > One major comment I have so far is that the test execution with the
> > > new agents is very unstable an the moment.
> > >
> > > Typically, I cannot execute tests more than once (and sometimes even
> > > more than one test!). Most of the time I get
> > > Error. RegistrationHandler.NO_AGENTS_AVAILABLE
> > >
> > > This is clearly a showstopper and we cannot commit the new code with
> > > this issue into the trunk.
> > >
> > > Please, take a look as soon as you can.
> > >
> > > Both Alexander and I use SATSA TCK as our driving vehicle, in CDC
> > > mode. We've changed the sources in your branch to be able to run the
> > > TCK.
> > >
> > > Another big issue - sometimes when I start the test run the second
> > > time, I see LOTS AND LOTS of agents registerations and deregistrations
> > > (hundreds), with 100% CPU consumed, and the end result is always the
> > > dreaded RegistrationHandler.NO_AGENTS_AVAILABLE.
> > >
> > > Here are some minor (compared to the above) issues:
> > > 1. GenericAgentCommand.setClasspath() is GLOBAL while the classpath
> > > values are test-local!
> > >
> > > 2. GenericAgentCommand: NotifiedTestInfo.waitUntilDone() without the
> > > timeout. Possibility for blocking forever!
> > >
> > > 3. GenericAgentCommand: Javadocs mention "ExecutionRequestQueue" few
> > > times,
> > > but there is no such thing.
> > >
> > > 4. Why do we need AgentMonitoringService, and all its dependencies
> > > (lots)?
> > > Can we get away without it?
> > >
> > > 5. Very non-standard java style in most of the sources. I've mass
> > > reformatted the source (please take a look and take a note :) ).
> > >
> > > 6. Lots of places where Exceptions are just swallowed with TODO
> > > comment. Very hard to debug problems.
> > >
> > > Thanks,
> > > --Vladimir
> > >
> > > On Mon, May 21, 2007 at 07:29:22PM +0400, Alexander Alexeev wrote:
> > > > Hi Maxim, Vladimir,
> > > >
> > > > two more comments:
> > > >
> > > > 1. The exception is thrown on JavaTest side when test's execution finished.
> > > > I've
> > > > submitted an issue: https://cqme.dev.java.net/issues/show_bug.cgi?id=151
> > > >
> > > > 2. Frequently second run still fails with error:
> > > > Error. RegistrationHandler.NO_AGENTS_AVAILABLE
> > > >
> > > > Thanks,
> > > > Alexander
> > > >
> > > > Maxim V. Sokolnikov wrote:
> > > > >Hello,
> > > > >
> > > > >I have incorporated my changes from
> > > > >branches/users/msokolnikov/Issue_70 in the new more organized way.
> > > > >
> > > > >They are in the new branch:
> > > > >
> > > > >http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=940
> > > > >http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=941
> > > > >http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=943
> > > > >http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=948
> > > > >
> > > > >
> > > > >Unfortunately I had no time to verify the fixes. I will do later.
> > > > >I am going on vacation until next Wednesday
> > > > >
> > > > >Thank you,
> > > > >Maxim
> > > > >
> > > > >---------------------------------------------------------------------
> > > > >To unsubscribe, e-mail: meframework-unsubscribe@cqme.dev.java.net
> > > > >For additional commands, e-mail: meframework-help@cqme.dev.java.net
> > > > >
> > > >

---------------------------------------------------------------------
To unsubscribe, e-mail: meframework-unsubscribe@cqme.dev.java.net
For additional commands, e-mail: meframework-help@cqme.dev.java.net

Alexander Alexeev

Hi Maxim,

just to double check. We need to put all new agent staff in main_agent.jar,
applet_agent.jar and xlet_agent.jar. Legacy stuff for old agent should be
removed from these jars. cdc_agent.jar are fully gone away.

Thanks,
Alexander

Maxim V. Sokolnikov wrote:
> Vladimir,
>
> I have started the integration of the generic agent functionality to
> the AgentXlet and AgentApplet wrapper.
> Before integrating of the my updates I want to double check that my
> approach is right.
>
> 1. J2ME TCK Framework defines the com.sun.tck.xlet.agent.BaseXlet,
> which extends AgentMain from the JavaTest. I am going to change the
> superclass of this class to the com.sun.tck.j2me.agent.AgentMain.
> It causes that the BaseXlet and their subclasses will use new generic
> communication framework and old JavaTest agent will not be supported
> anymore.
>
> 2. I will create AgentApplet wrapper for the
> com.sun.tck.j2me.agent.AgentMain.
>
>
> Also I found that PP TCK uses own version of the AgentXlet. As I
> understand it supports additional functionality for inter xlet
> communication.
>
> Do we need to integrate this functionality from PP TCK to the J2ME TCK
> Framework?
>
> Thank you,
> Maxim
>
> Vladimir Sizikov writes:
> > Maxim,
> >
> > Alexander and I spent some time reviewing and fixing the code here and
> > there, and testing, and lots of code style formattings, javadoc
> > updates and typo fixes. :)
> >
> > Eventually, I've commit the latest version into the trunk.
> >
> > The code works, and in our limited testisg, works good.
> > Sitll, there is a lot to done, and some unresolved issues, comments.
> >
> > Please find below my comments:
> >
> > 1. SimplePipe.DEFAULT_NAME is with wrong package name
> >
> > 2. ServerExecutionService javadoc refers to non-public APIs, like
> > RegistrationHandler and setRegistrationHandler()
> >
> > 3. ServerProactiveCommand does the same.
> >
> > 4. Response has public fields!
> >
> > 5. AgentMain, this very important class, has no public class-level
> > javadocs.
> >
> > 6. Agent-Wrappers (Applet, Xlet) - should be updated to use the new
> > base agent, right?
> >
> > 7. Javadoc warnings should be eliminated (at least those that are from
> > your updates).
> >
> > ====================== Previous comments ========================
> >
> > 1. GenericAgentCommand.setClasspath() is GLOBAL while the classpath
> > values are test-local!
> >
> > 2. GenericAgentCommand: NotifiedTestInfo.waitUntilDone() without the
> > timeout. Possibility for blocking forever!
> >
> > 3. GenericAgentCommand: Javadocs mention "ExecutionRequestQueue" few
> > times, but there is no such thing.
> >
> > 4. Why do we need AgentMonitoringService, and all its dependencies
> > (lots)? Can we get away without it?
> >
> > Thanks,
> > --Vladimir
> >
> > On Wed, May 23, 2007 at 08:11:36PM -0700, Maxim Sokolnikov wrote:
> > > Vladimir, Alexander,
> > >
> > > I have fixed this problem in Revision 990. I also fixed other issues.
> > > Could you please review? (http://fisheye4.cenqua.com/changelog/cqme/?cs=990)
> > >
> > > Thank you,
> > > Maxim
> > >
> > >
> > > Vladimir Sizikov writes:
> > > > Hi Maxim,
> > > >
> > > > One major comment I have so far is that the test execution with the
> > > > new agents is very unstable an the moment.
> > > >
> > > > Typically, I cannot execute tests more than once (and sometimes even
> > > > more than one test!). Most of the time I get
> > > > Error. RegistrationHandler.NO_AGENTS_AVAILABLE
> > > >
> > > > This is clearly a showstopper and we cannot commit the new code with
> > > > this issue into the trunk.
> > > >
> > > > Please, take a look as soon as you can.
> > > >
> > > > Both Alexander and I use SATSA TCK as our driving vehicle, in CDC
> > > > mode. We've changed the sources in your branch to be able to run the
> > > > TCK.
> > > >
> > > > Another big issue - sometimes when I start the test run the second
> > > > time, I see LOTS AND LOTS of agents registerations and deregistrations
> > > > (hundreds), with 100% CPU consumed, and the end result is always the
> > > > dreaded RegistrationHandler.NO_AGENTS_AVAILABLE.
> > > >
> > > > Here are some minor (compared to the above) issues:
> > > > 1. GenericAgentCommand.setClasspath() is GLOBAL while the classpath
> > > > values are test-local!
> > > >
> > > > 2. GenericAgentCommand: NotifiedTestInfo.waitUntilDone() without the
> > > > timeout. Possibility for blocking forever!
> > > >
> > > > 3. GenericAgentCommand: Javadocs mention "ExecutionRequestQueue" few
> > > > times,
> > > > but there is no such thing.
> > > >
> > > > 4. Why do we need AgentMonitoringService, and all its dependencies
> > > > (lots)?
> > > > Can we get away without it?
> > > >
> > > > 5. Very non-standard java style in most of the sources. I've mass
> > > > reformatted the source (please take a look and take a note :) ).
> > > >
> > > > 6. Lots of places where Exceptions are just swallowed with TODO
> > > > comment. Very hard to debug problems.
> > > >
> > > > Thanks,
> > > > --Vladimir
> > > >
> > > > On Mon, May 21, 2007 at 07:29:22PM +0400, Alexander Alexeev wrote:
> > > > > Hi Maxim, Vladimir,
> > > > >
> > > > > two more comments:
> > > > >
> > > > > 1. The exception is thrown on JavaTest side when test's execution finished.
> > > > > I've
> > > > > submitted an issue: https://cqme.dev.java.net/issues/show_bug.cgi?id=151
> > > > >
> > > > > 2. Frequently second run still fails with error:
> > > > > Error. RegistrationHandler.NO_AGENTS_AVAILABLE
> > > > >
> > > > > Thanks,
> > > > > Alexander
> > > > >
> > > > > Maxim V. Sokolnikov wrote:
> > > > > >Hello,
> > > > > >
> > > > > >I have incorporated my changes from
> > > > > >branches/users/msokolnikov/Issue_70 in the new more organized way.
> > > > > >
> > > > > >They are in the new branch:
> > > > > >
> > > > > >http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=940
> > > > > >http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=941
> > > > > >http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=943
> > > > > >http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=948
> > > > > >
> > > > > >
> > > > > >Unfortunately I had no time to verify the fixes. I will do later.
> > > > > >I am going on vacation until next Wednesday
> > > > > >
> > > > > >Thank you,
> > > > > >Maxim
> > > > > >
> > > > > >---------------------------------------------------------------------
> > > > > >To unsubscribe, e-mail: meframework-unsubscribe@cqme.dev.java.net
> > > > > >For additional commands, e-mail: meframework-help@cqme.dev.java.net
> > > > > >
> > > > >
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: meframework-unsubscribe@cqme.dev.java.net
> For additional commands, e-mail: meframework-help@cqme.dev.java.net
>

---------------------------------------------------------------------
To unsubscribe, e-mail: meframework-unsubscribe@cqme.dev.java.net
For additional commands, e-mail: meframework-help@cqme.dev.java.net

Alexander Alexeev

Hi Maxim,

I see we are talking about different class loadings. You means local
class loading but I means remote class loading. After your fix tests
can't load classes from agent's local classpath (-Djava.class.path
option). Why to prevent tests access classes in the local classpath? For
example distributed tests use comm clients. Now these classes will be
loaded over network although they are already presented in the local
classpath. To avoid this we have to duplicate jars via -classpath argument of
agent.

The issue #140 is talking about classpath created in AgentQueue. Looks
at the method AgentQueue.setClassPath(). Loader created in this
method should not have any parent even system class loader since it should
provide resources from classpath of command only.

Comments about -classpath option:
1. Handling of argument is incorrect. Instead of
} else if (args[index].equals("CLASSPATH_OPTION")) {
this.classpath = args[++index];
should be:
} else if (args[index].equals(CLASSPATH_OPTION)) {
this.classpath = args[++index];
return 2;

2. Displayable help should contain -classpath argument.

Thanks,
Alexander

Maxim V. Sokolnikov wrote:
> Hello,
>
> I have incorporated my changes from
> branches/users/msokolnikov/Issue_70 in the new more organized way.
>
> They are in the new branch:
>
> http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=940
> http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=941
> http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=943
> http://fisheye4.cenqua.com/changelog/~author=msokolnikov/cqme/?cs=948
>
>
> Unfortunately I had no time to verify the fixes. I will do later.
> I am going on vacation until next Wednesday
>
> Thank you,
> Maxim
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: meframework-unsubscribe@cqme.dev.java.net
> For additional commands, e-mail: meframework-help@cqme.dev.java.net
>

---------------------------------------------------------------------
To unsubscribe, e-mail: meframework-unsubscribe@cqme.dev.java.net
For additional commands, e-mail: meframework-help@cqme.dev.java.net