Thomas Watson | 5 Oct 23:45
Picon
Favicon

Equinox contribution for next Helios integration build.

The map file has been updated for the following Bug changes:
+ Bug 288783. [ds] SCRCommandProvider does not format help quite right (FIXED)
+ Bug 289384. Filter does not allow !, | or & as first character of attr names (FIXED)
+ Bug 290534. ServiceRegistry doesn't remove BundleContextImpl keys from publishedServicesByContext HashMap when list of services becomes empty (FIXED)
+ Bug 290661. SWT & Launcher fragments using 'HEAD' as qualifier (FIXED)

The following projects have changed:
org.eclipse.equinox.ds
org.eclipse.equinox.launcher.motif.solaris.sparc
org.eclipse.osgi.tests
org.eclipse.osgi

Tom

Chun Ling Li | 6 Oct 04:01
Picon
Favicon

AUTO: Chun Ling Li is out of the office (returning 10/09/2009)

I am out of the office until 10/09/2009.

China National Holiday. Urgent call: 13641329869.


Note: This is an automated response to your message "[equinox-dev] Equinox contribution for next Helios integration build." sent on 10/6/09 5:45:37.

This is the only notification you will receive while this person is away.

Michael Furtak | 6 Oct 19:51

Use of Thread Context ClassLoader considered harmful?

Hi all,

I am using Apache XMLBeans 2.4 within Equinox 3.5, and am running into
problems. Specifically, I (and others [1,2] - references at bottom) have
noticed that a specific part of the type resolution process breaks when
run in an OSGi environment.

After some investigation, I have found that XMLBeans is creating its
SchemaTypeLoader objects with the current thread's context
classloader[3]. My experience has been that any time a library does this
and then needs to load classes that are in a different OSGi bundle, the
classes can not be resolved.

So I suppose my questions are: Could Equinox do more to prevent this
kind of use-case from breaking down? Alternatively, for compatibility
with an OSGi environment, is using a TCCL considered harmful? If so,
what should I suggest in my XMLBeans ticket as an OSGi-compatible
alternative?

Thanks,
-Mike Furtak

1)
http://markmail.org/message/vx2vdg7yhb4l3ej6#query:+page:1+mid:vx2vdg7yh
b4l3ej6+state:results

2)
http://mail-archives.apache.org/mod_mbox/ws-axis-user/200701.mbox/%3C8d2
a81130701171341p55f95b70o7f144ea7d941639a@...%3E

3)
http://grepcode.com/file/repo1.maven.org$maven2-+O6RNhJB37wziNMXjqAYNGzQI0aq7523 <at> public.gmane.org$xmlb
eans <at> 2.4.0 <at> org$apache$xmlbeans$impl$schema$SchemaTypeLoaderImpl.java#Sch
emaTypeLoaderImpl.getContextTypeLoader%28%29 (Line 131)

 

 
THIS MESSAGE IS INTENDED FOR THE USE OF THE PERSON TO WHOM IT IS ADDRESSED. IT MAY CONTAIN INFORMATION THAT IS
PRIVILEGED, CONFIDENTIAL AND EXEMPT FROM DISCLOSURE UNDER APPLICABLE LAW. If you are not the intended
recipient, your use of this message for any purpose is strictly prohibited. If you have received this
communication in error, please delete the message and notify the sender so that we may correct our records.
Thomas Watson | 6 Oct 20:14
Picon
Favicon

Re: Use of Thread Context ClassLoader considered harmful?

In Equinox we have something called buddy class loading (not OSGi specified). It may be used to help your situation. See http://wiki.eclipse.org/index.php/Context_Class_Loader_Enhancements

HTH

Tom



"Michael Furtak" ---10/06/2009 12:53:27 PM---Hi all,


From:

"Michael Furtak" <mfurtak-hKw+ElpZSXI@public.gmane.org>

To:

"Equinox development mailing list" <equinox-dev-j9T/66MeVpFAfugRpC6u6w@public.gmane.org>

Date:

10/06/2009 12:53 PM

Subject:

[equinox-dev] Use of Thread Context ClassLoader considered harmful?



Hi all,

I am using Apache XMLBeans 2.4 within Equinox 3.5, and am running into
problems. Specifically, I (and others [1,2] - references at bottom) have
noticed that a specific part of the type resolution process breaks when
run in an OSGi environment.

After some investigation, I have found that XMLBeans is creating its
SchemaTypeLoader objects with the current thread's context
classloader[3]. My experience has been that any time a library does this
and then needs to load classes that are in a different OSGi bundle, the
classes can not be resolved.

So I suppose my questions are: Could Equinox do more to prevent this
kind of use-case from breaking down? Alternatively, for compatibility
with an OSGi environment, is using a TCCL considered harmful? If so,
what should I suggest in my XMLBeans ticket as an OSGi-compatible
alternative?


Thanks,
-Mike Furtak


1)
http://markmail.org/message/vx2vdg7yhb4l3ej6#query:+page:1+mid:vx2vdg7yh
b4l3ej6+state:results

2)
http://mail-archives.apache.org/mod_mbox/ws-axis-user/200701.mbox/%3C8d2
a81130701171341p55f95b70o7f144ea7d941639a-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org%3E

3)
http://grepcode.com/file/repo1.maven.org$maven2 <at> org.apache.xmlbeans$xmlb
eans <at> 2.4.0 <at> org$apache$xmlbeans$impl$schema$SchemaTypeLoaderImpl.java#Sch
emaTypeLoaderImpl.getContextTypeLoader%28%29 (Line 131)




THIS MESSAGE IS INTENDED FOR THE USE OF THE PERSON TO WHOM IT IS ADDRESSED. IT MAY CONTAIN INFORMATION THAT IS PRIVILEGED, CONFIDENTIAL AND EXEMPT FROM DISCLOSURE UNDER APPLICABLE LAW. If you are not the intended recipient, your use of this message for any purpose is strictly prohibited. If you have received this communication in error, please delete the message and notify the sender so that we may correct our records.
_______________________________________________
equinox-dev mailing list
equinox-dev-j9T/66MeVpFAfugRpC6u6w@public.gmane.org
https://dev.eclipse.org/mailman/listinfo/equinox-dev


Michael Furtak | 6 Oct 21:00

RE: Use of Thread Context ClassLoader consideredharmful?

Hi Tom,

 

The Buddy manifest adjustments do indeed resolve the issue. Thank you for that excellent link.

 

Out of curiosity, are you aware of any effort to make this solution a part of the OSGi spec?

 

Thanks again,

-Mike

 

 

 
 
 
 
THIS MESSAGE IS INTENDED FOR THE USE OF THE PERSON TO WHOM IT IS ADDRESSED. IT MAY CONTAIN INFORMATION THAT IS PRIVILEGED, CONFIDENTIAL AND EXEMPT FROM DISCLOSURE UNDER APPLICABLE LAW. If you are not the intended recipient, your use of this message for any purpose is strictly prohibited. If you have received this communication in error, please delete the message and notify the sender so that we may correct our records.
 
 
 

From: equinox-dev-bounces-j9T/66MeVpFAfugRpC6u6w@public.gmane.org [mailto:equinox-dev-bounces-j9T/66MeVpFAfugRpC6u6w@public.gmane.org] On Behalf Of Thomas Watson
Sent: Tuesday, October 06, 2009 2:14 PM
To: Equinox development mailing list
Subject: Re: [equinox-dev] Use of Thread Context ClassLoader consideredharmful?

 

In Equinox we have something called buddy class loading (not OSGi specified). It may be used to help your situation. See http://wiki.eclipse.org/index.php/Context_Class_Loader_Enhancements

HTH

Tom



"Michael Furtak" ---10/06/2009 12:53:27 PM---Hi all,


From:


"Michael Furtak" <mfurtak-hKw+ElpZSXI@public.gmane.org>


To:


"Equinox development mailing list" <equinox-dev-j9T/66MeVpFAfugRpC6u6w@public.gmane.org>


Date:


10/06/2009 12:53 PM


Subject:


[equinox-dev] Use of Thread Context ClassLoader considered harmful?




Hi all,

I am using Apache XMLBeans 2.4 within Equinox 3.5, and am running into
problems. Specifically, I (and others [1,2] - references at bottom) have
noticed that a specific part of the type resolution process breaks when
run in an OSGi environment.

After some investigation, I have found that XMLBeans is creating its
SchemaTypeLoader objects with the current thread's context
classloader[3]. My experience has been that any time a library does this
and then needs to load classes that are in a different OSGi bundle, the
classes can not be resolved.

So I suppose my questions are: Could Equinox do more to prevent this
kind of use-case from breaking down? Alternatively, for compatibility
with an OSGi environment, is using a TCCL considered harmful? If so,
what should I suggest in my XMLBeans ticket as an OSGi-compatible
alternative?


Thanks,
-Mike Furtak


1)
http://markmail.org/message/vx2vdg7yhb4l3ej6#query:+page:1+mid:vx2vdg7yh
b4l3ej6+state:results

2)
http://mail-archives.apache.org/mod_mbox/ws-axis-user/200701.mbox/%3C8d2
a81130701171341p55f95b70o7f144ea7d941639a-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org%3E

3)
http://grepcode.com/file/repo1.maven.org$maven2-+O6RNhJB37wziNMXjqAYNA@public.gmane.organs$xmlb
eans <at> 2.4.0 <at> org$apache$xmlbeans$impl$schema$SchemaTypeLoaderImpl.java#Sch
emaTypeLoaderImpl.getContextTypeLoader%28%29 (Line 131)




THIS MESSAGE IS INTENDED FOR THE USE OF THE PERSON TO WHOM IT IS ADDRESSED. IT MAY CONTAIN INFORMATION THAT IS PRIVILEGED, CONFIDENTIAL AND EXEMPT FROM DISCLOSURE UNDER APPLICABLE LAW. If you are not the intended recipient, your use of this message for any purpose is strictly prohibited. If you have received this communication in error, please delete the message and notify the sender so that we may correct our records.
_______________________________________________
equinox-dev mailing list
equinox-dev-j9T/66MeVpFAfugRpC6u6w@public.gmane.org
https://dev.eclipse.org/mailman/listinfo/equinox-dev

Thomas Watson | 6 Oct 21:36
Picon
Favicon

RE: Use of Thread Context ClassLoader consideredharmful?

You may have noticed that the wiki is rather detailed in describing the buddy class loading design. This was because the Equinox team prepared the design as input for a future OSGi specification. The design works for about 80% of the cases, but it does have some disadvantages.

1) There is no good way to force the ContextFinder across all threads in the system. If some code unsets the context class loader for a thread then the context finder will no longer take effect.
2) ContextFinder has issues when used with Class.forName(String, boolean, ClassLoader). Class.forName will cause the specified class loader to be the initiating class loader. The VM will cache all Class objects for the initiating class loader, any time that class loader is asked for a particular class name again it must return the same Class object as previously returned. This causes issues if multiple versions of a package need to be supported.
3) There were other issues, but they are not fresh in my mind ;-)

At any rate, this proposal fell out of the specification for OSGi R4 V4.2 core specification. OSGi could decide to tackle this issue again in the next specification, but I am not confident the solution will end up being buddy class loading.

Tom



"Michael Furtak" ---10/06/2009 02:00:33 PM---Hi Tom,


From:

"Michael Furtak" <mfurtak-hKw+ElpZSXI@public.gmane.org>

To:

"Equinox development mailing list" <equinox-dev-j9T/66MeVpFAfugRpC6u6w@public.gmane.org>

Date:

10/06/2009 02:00 PM

Subject:

RE: [equinox-dev] Use of Thread Context ClassLoader consideredharmful?



Hi Tom,

The Buddy manifest adjustments do indeed resolve the issue. Thank you for that excellent link.

Out of curiosity, are you aware of any effort to make this solution a part of the OSGi spec?

Thanks again,
-Mike





THIS MESSAGE IS INTENDED FOR THE USE OF THE PERSON TO WHOM IT IS ADDRESSED. IT MAY CONTAIN INFORMATION THAT IS PRIVILEGED, CONFIDENTIAL AND EXEMPT FROM DISCLOSURE UNDER APPLICABLE LAW. If you are not the intended recipient, your use of this message for any purpose is strictly prohibited. If you have received this communication in error, please delete the message and notify the sender so that we may correct our records.



From: equinox-dev-bounces-j9T/66MeVpFAfugRpC6u6w@public.gmane.org [mailto:equinox-dev-bounces <at> eclipse.org] On Behalf Of Thomas Watson
Sent: Tuesday, October 06, 2009 2:14 PM
To: Equinox development mailing list
Subject: Re: [equinox-dev] Use of Thread Context ClassLoader consideredharmful?

In Equinox we have something called buddy class loading (not OSGi specified). It may be used to help your situation. See http://wiki.eclipse.org/index.php/Context_Class_Loader_Enhancements

HTH

Tom



"Michael Furtak" ---10/06/2009 12:53:27 PM---Hi all,


From:

"Michael Furtak" <mfurtak-hKw+ElpZSXI@public.gmane.org>

To:

"Equinox development mailing list" <equinox-dev-j9T/66MeVpHF2uMehF1BdA@public.gmane.orgg>

Date:

10/06/2009 12:53 PM

Subject:

[equinox-dev] Use of Thread Context ClassLoader considered harmful?




Hi all,

I am using Apache XMLBeans 2.4 within Equinox 3.5, and am running into
problems. Specifically, I (and others [1,2] - references at bottom) have
noticed that a specific part of the type resolution process breaks when
run in an OSGi environment.

After some investigation, I have found that XMLBeans is creating its
SchemaTypeLoader objects with the current thread's context
classloader[3]. My experience has been that any time a library does this
and then needs to load classes that are in a different OSGi bundle, the
classes can not be resolved.

So I suppose my questions are: Could Equinox do more to prevent this
kind of use-case from breaking down? Alternatively, for compatibility
with an OSGi environment, is using a TCCL considered harmful? If so,
what should I suggest in my XMLBeans ticket as an OSGi-compatible
alternative?


Thanks,
-Mike Furtak


1)
http://markmail.org/message/vx2vdg7yhb4l3ej6#query:+page:1+mid:vx2vdg7yh
b4l3ej6+state:results

2)
http://mail-archives.apache.org/mod_mbox/ws-axis-user/200701.mbox/%3C8d2
a81130701171341p55f95b70o7f144ea7d941639a-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org%3E

3)
http://grepcode.com/file/repo1.maven.org$maven2-+O6RNhJB37wziNMXjqAYNGzQI0aq7523@public.gmane.org$xmlb
eans <at> 2.4.0 <at> org$apache$xmlbeans$impl$schema$SchemaTypeLoaderImpl.java#Sch
emaTypeLoaderImpl.getContextTypeLoader%28%29 (Line 131)




THIS MESSAGE IS INTENDED FOR THE USE OF THE PERSON TO WHOM IT IS ADDRESSED. IT MAY CONTAIN INFORMATION THAT IS PRIVILEGED, CONFIDENTIAL AND EXEMPT FROM DISCLOSURE UNDER APPLICABLE LAW. If you are not the intended recipient, your use of this message for any purpose is strictly prohibited. If you have received this communication in error, please delete the message and notify the sender so that we may correct our records.
_______________________________________________
equinox-dev mailing list
equinox-dev-j9T/66MeVpFAfugRpC6u6w@public.gmane.org
https://dev.eclipse.org/mailman/listinfo/equinox-dev
_______________________________________________
equinox-dev mailing list
equinox-dev-j9T/66MeVpFAfugRpC6u6w@public.gmane.org
https://dev.eclipse.org/mailman/listinfo/equinox-dev


prof_trg | 8 Oct 22:51
Picon

attach a fragment in runtime


Hello,
I am trying to attach a fragment to its host at runtime, but I have some
issues...
in fact , I uninstall the fragment then I install it again, but as if it is
not detected by the host...

before uninstalling the fragment, I have this line in my console:

29 2/active Host (h:28)
28 2/resolved Fragment (f:29)

after uninstalling and then installing the fragment again :

29 2/active Host (h:28)
30 2/installed Fragment

would someone help me understanding this?
Regard
--

-- 
View this message in context: http://www.nabble.com/attach-a-fragment-in-runtime-tp25810711p25810711.html
Sent from the Equinox - Dev mailing list archive at Nabble.com.

Thomas Watson | 8 Oct 23:04
Picon
Favicon

Re: attach a fragment in runtime

That console output does not look familiar to me. Is that from the Equinox console? Fragments cannot dynamically detach from a host on uninstall. The host bundle must be refreshed in order for the fragment to be completely flushed from the system. A new version of the fragment will not be allowed to attach to the host until the old version is flushed. This is done by calling PackageAdmin.refreshPackages() method on the host bundle.

HTH.

Tom



prof_trg ---10/08/2009 03:53:57 PM---Hello,


From:

prof_trg <prof.target-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>

To:

equinox-dev-j9T/66MeVpFAfugRpC6u6w@public.gmane.org

Date:

10/08/2009 03:53 PM

Subject:

[equinox-dev] attach a fragment in runtime




Hello,
I am trying to attach a fragment to its host at runtime, but I have some
issues...
in fact , I uninstall the fragment then I install it again, but as if it is
not detected by the host...

before uninstalling the fragment, I have this line in my console:

29 2/active Host (h:28)
28 2/resolved Fragment (f:29)

after uninstalling and then installing the fragment again :

29 2/active Host (h:28)
30 2/installed Fragment

would someone help me understanding this?
Regard
--
View this message in context: http://www.nabble.com/attach-a-fragment-in-runtime-tp25810711p25810711.html
Sent from the Equinox - Dev mailing list archive at Nabble.com.

_______________________________________________
equinox-dev mailing list
equinox-dev-j9T/66MeVpFAfugRpC6u6w@public.gmane.org
https://dev.eclipse.org/mailman/listinfo/equinox-dev


Lynn Gayowski | 9 Oct 15:40
Favicon
Gravatar

Webinar: Equinox above the Cloud - Oct. 13

A reminder that Scott Lewis, Markus Knauer & Harald Kornmayer are giving a webinar:

Equinox above the Cloud – Some Call it Heaven!
October 13, 2009 at 8:00 am PDT / 11:00 am EDT / 3:00 pm GMT
To register email webinar-equinox at eclipse dot org

 

Cloud computing is one of the hottest topics currently discussed in the IT sector. Although there are many different definitions of the term 'Cloud', providing remote services is at the center of Cloud computing. Eclipse is hosting a webinar to demonstrate how "goodies" from different Eclipse projects can be combined to build applications on Cloud infrastructures.

 

The webinar will start with a short introduction to Cloud computing. Using the g-Eclipse framework, a distributed, virtual infrastructure will be set-up. Leveraging OSGi and ECF frameworks, a manageable distributed runtime will be started in order to deploy an application dynamically. The front end for the application will be provided with the support of the RAP application.

 

For more information on this and other Eclipse webinars visit http://live.eclipse.org/.  Special thanks to Adobe for contributing their Adobe Acrobat Connect product to host the webinar.

 

Cheers,

Lynn Gayowski

Marketing Events Manager

Eclipse Foundation, Inc.

P (613) 224-9461 ext. 234

F (613) 224-5172

lynn.gayowski-j9T/66MeVpFAfugRpC6u6w@public.gmane.org

www.eclipse.org

 

 

Kirchev, Lazar | 12 Oct 17:28
Picon
Favicon

Problem with defining a custom adaptor hook in a framework extension

Hello,
 
I am trying to define a custom adaptor hook in a framework extension, but it seems as if the hookconfigurators.properties file in the extending fragment bundle jar is not read. When debugging the HooksRegistry code I found that the hookconfigurators.properties file in the bundle is not returned as a resource by the framework classloader.
 
I have set the osgi.framework.extensions property in the config.ini file, and when starting equinox (I start it with java -jar <equinox.jar>), I add the extension bundle to the classpath. The extension fragment is correctly recognized as such, and gets successfully resolved. The framework and extension bundles are in one and the same folder. I am using 3.5 release. 
 
Any idea what the problem may be?
 
Best regards,
Lazar Kirchev

Gmane