Yea I can't really think of a good solution for this. +1 for reverting, or maybe even better just have the exceptions from RequestLifecycle.end() logged. I believe previously they were ignored.
On 12/02/2013 12:29 PM, Nicolas Filotto wrote:
Hi Nick thank you for your help, see below my comments:
Writing the response into memory would of course work but it is clearly not an option for obvious reasons.
I'm not sure I understand what you mean by "For WS-284 could you still write the response out but then supply an error code later ?" If the "response" is already written I don't see how it would be possible to change the status code (I guess it is what you mean by error code) later knowing that the status code is part of the header which is written before the body that contains the "response". Even if it was possible we would have an inconsistency between the body and the status code which will be seen as a bug for many customers/partners/consultants
So I keep on saying that there is nothing I can do on my side, unless you have other suggestions. If not as it is not possible to improve the code on gatein, I will revert the changes for WS-284 within the context of 285
On Mon, Dec 2, 2013 at 4:24 PM, Nick Scavelli <nscavell@redhat.com> wrote:
Hmm yes, unfortunately the only way to retrieve portlet state which is non transient is to load it from the backend which will definitely require JCR session to be available.
For WS-284 could you still write the response out but then supply an error code later ? So at least you have an error from the server even though the response has been written out. When I looked at this awhile ago - I was playing with writing the response into memory, but that obviously won't scale.
On 12/02/2013 10:53 AM, Nicolas Filotto wrote:
If so it means that WS-285 and WS-284 are simply incompatible. To be able to fix WS-284, we need to call RequestLifeCycle.end() before writing the response but for WS-285 we need to call RequestLifeCycle.end() after writing the response so there is nothing I can do on my side.If at gatein level you cannot improve your mop marshaller to avoid depending on the RequestLifeCycle, I believe that the only way to fix it would be to simply revert the changes done for WS-284
Please check on your side and tell me if the mop marshaller can be improved, if not I will revert WS-284 within the context of WS-285.Thank you in advance
On Mon, Dec 2, 2013 at 2:10 PM, Lucas Ponce <lponce@redhat.com> wrote:
Hello Nicolas,
I've tested your .jar in my scenario and I can't make it work.
I'm still having same Exception.
You can test it from gatein using a simple curl command:
curl -i -H "Accept: application/json" -u root:gtn http://localhost:8080/rest/private/managed-components/mop/portalsites/classic/portal.xml
You get a Exception meanshile a xml is expected.
Thanks,
Lucas
----- Mensaje original -----
> De: "Nicolas Filotto" <nicolas.filotto@exoplatform.com>
> Para: "Trong Tran" <trongtt@gmail.com>> Asunto: Re: [gatein-dev] SessionManagerContainer in exo.kernel.container 2.5.0-Alpha3
> CC: "Lucas Ponce" <lponce@redhat.com>, "gatein-dev" <gatein-dev@lists.jboss.org>
> Enviados: Lunes, 2 de Diciembre 2013 9:45:12
>
> @Trong Great catch !!
> @Lucas Thank you very much for reporting this issue and proposing a PR, I
> really appreciate !
> Could you please try with this jar file and tell me if WS-285 and WS-284?
>
> NB: For the PR, please note that we use the repository "exoplatform" for
> the maintenance branches and "exodev" for the master, so in this particular
> case you were supposed to use exodev as it affects the master, thank you
> very much anyway for your PR.
>
>
> On Fri, Nov 29, 2013 at 9:24 AM, Trong Tran <trongtt@gmail.com> wrote:
>
> > I believe that this problem is related to
> > https://github.com/exoplatform/ws/commit/50a829a047ff91a1f358d9ef2536486dceb320ec
> >
> > It seems to me that RequestLifeCycle.end() is being called sooner than
> > expected in
> > https://github.com/exoplatform/ws/blob/2.4.0-Alpha3/exo.ws.rest.core/src/main/java/org/exoplatform/services/rest/impl/RequestHandlerImpl.java
> >
> >
> >
> > On 29 November 2013 15:15, Nicolas Filotto <
> > nicolas.filotto@exoplatform.com> wrote:
> >
> >> There are many things that have been added between Alpha1 and Alpha3, so
> >> I will need more details. What do you mean exactly by "strange behavior
> >> with some ThreadLocal<> variables"?
> >>
> >> Please note that I don't work on Fridays so I will answer next week once
> >> I will have more details
> >>
> >>
> >> On Thu, Nov 28, 2013 at 8:46 PM, Lucas Ponce <lponce@redhat.com> wrote:
> >>
> >>> Nicolas,
> >>>
> >>> Do you know if there is some change related multithreading between
> >>> kernel 2.5.0-Alpha3 and 2.5.0-Alpha1 ?
> >>>
> >>> Debugging a class I'm seeing some strange behaviour with some
> >>> ThreadLocal<> variables.
> >>>
> >>> Thanks again,
> >>> Lucas
> >>>
> >>> ----- Mensaje original -----
> >>> > De: "Nicolas Filotto" <nicolas.filotto@exoplatform.com>
> >>> > Para: "Lucas Ponce" <lponce@redhat.com>
> >>> > CC: "gatein-dev" <gatein-dev@lists.jboss.org>
> >>> > Enviados: Jueves, 28 de Noviembre 2013 18:08:29
> >>> > Asunto: Re: [gatein-dev] SessionManagerContainer in
> >>> exo.kernel.container 2.6.0-Alpha3
> >>> >
> >>> > If you use the SessionManagerContainer, why don't you have a
> >>> compilation
> >>> > error ?
> >>> >
> >>> >
> >>> > On Thu, Nov 28, 2013 at 4:50 PM, Lucas Ponce <lponce@redhat.com>
> >>> wrote:
> >>> >
> >>> > > Hello Nicolas,
> >>> > >
> >>> > > POMSessionManager.java needs to access to SessionContext at this
> >>> point:
> >>> > >
> >>> > > public POMSession getSession() {
> >>> > > SessionContext context = configurator.getContext();
> >>> > > return context != null ? (POMSession)
> >>> > > context.getAttachment("mopsession") : null;
> >>> > > }
> >>> > >
> >>> > > This is needed to PortalNamesCache.java to perform a query to
> >>> > > POMDataStorage component:
> >>> > >
> >>> > > @Override
> >>> > > public <V> V execute(POMSession session, POMTask<V> task) throws
> >>> > > Exception {
> >>> > > if (!session.isModified()) {
> >>> > > if (task instanceof SearchTask.FindSiteKey) {
> >>> > > SearchTask.FindSiteKey find =
> >>> (SearchTask.FindSiteKey)
> >>> > > task;
> >>> > > List<PortalKey> data = (List<PortalKey>)
> >>> > > session.getFromCache(find.getKey());
> >>> > > if (data == null) {
> >>> > > V result = super.execute(session, task);
> >>> > > LazyPageList<PortalKey> list =
> >>> > > (LazyPageList<PortalKey>) result;
> >>> > > session.putInCache(find.getKey(),
> >>> > > Collections.unmodifiableList(new
> >>> ArrayList<PortalKey>(list.getAll())));
> >>> > > return result;
> >>> > > } else {
> >>> > > return (V) new LazyPageList<PortalKey>(new
> >>> > > ListAccessImpl<PortalKey>(PortalKey.class, data), 10);
> >>> > > }
> >>> > > } else if (task instanceof PortalConfigTask.Save || task
> >>> > > instanceof PortalConfigTask.Remove) {
> >>> > > V result = super.execute(session, task);
> >>> > >
> >>> > > session.scheduleForEviction(SearchTask.FindSiteKey.PORTAL_KEY);
> >>> > >
> >>> > > session.scheduleForEviction(SearchTask.FindSiteKey.GROUP_KEY);
> >>> > > return result;
> >>> > > }
> >>> > > }
> >>> > >
> >>> > > //
> >>> > > return super.execute(session, task);
> >>> > > }
> >>> > >
> >>> > > With the current change of version in exo.core, the getSession() in
> >>> > > POMSessionManager is always null, so a NPE is thrown breaking the POM
> >>> > > service.
> >>> > >
> >>> > > I wonder how I can integrate these new changes to have same
> >>> behaviour that
> >>> > > previous version.
> >>> > >
> >>> > > Thanks,
> >>> > > Lucas
> >>> > >
> >>> > >
> >>> > > ----- Mensaje original -----
> >>> > > > De: "Nicolas Filotto" <nicolas.filotto@exoplatform.com>
> >>> > > > Para: "Lucas Ponce" <lponce@redhat.com>
> >>> > > > CC: "gatein-dev" <gatein-dev@lists.jboss.org>
> >>> > > > Enviados: Jueves, 28 de Noviembre 2013 17:29:27
> >>> > > > Asunto: Re: [gatein-dev] SessionManagerContainer in
> >>> exo.kernel.container
> >>> > > 2.6.0-Alpha3
> >>> > > >
> >>> > > > What is the need exactly ? Beeing able to scope a component to a
> >>> Session?
> >>> > > > If so it is now possible to define a scope to a component thanks to
> >>> > > > https://jira.exoplatform.org/browse/KER-258. Out of the box, the
> >>> kernel
> >>> > > > supports the scope Singleton, ApplicationScoped and Dependent. By
> >>> adding
> >>> > > > some configuration you can extend the list to SessionScoped and
> >>> > > > RequestScoped. Here is the configuration to add if you want both:
> >>> > > >
> >>> > > > <component>
> >>> > > > <key>org.exoplatform.container.context.ContextManager</key>
> >>> > > > <type>org.exoplatform.container.context.ContextManagerImpl</type>
> >>> > > > <component-plugins>
> >>> > > > <component-plugin>
> >>> > > > <name>main-scopes</name>
> >>> > > > <set-method>addContexts</set-method>
> >>> > > > <type>org.exoplatform.container.context.ContextPlugin</type>
> >>> > > > <init-params>
> >>> > > > <object-param>
> >>> > > > <name>request-scope</name>
> >>> > > > <object type="org.exoplatform.container.context.RequestContext"/>
> >>> > > > </object-param>
> >>> > > > <object-param>
> >>> > > > <name>session-scope</name>
> >>> > > > <object type="org.exoplatform.container.context.SessionContext"/>
> >>> > > > </object-param>
> >>> > > > </init-params>
> >>> > > > </component-plugin>
> >>> > > > </component-plugins>
> >>> > > > </component>
> >>> > > >
> >>> > > > Please note that the components that you want to scope to
> >>> SessionScoped
> >>> > > or
> >>> > > > RequestScoped must be proxyable as they are normal scopes (which
> >>> means
> >>> > > that
> >>> > > > you will access to the methods through a proxy and you must not
> >>> access to
> >>> > > > the fields directly), moreover components that you want to scope to
> >>> > > > SessionScoped must be Serializable as SessionScoped is a
> >>> passivating
> >>> > > scope.
> >>> > > >
> >>> > > > Please also note that thanks to KER-252 you can use the annotations
> >>> > > defined
> >>> > > > in JSR 330 such as Inject
> >>> > > >
> >>> > > >
> >>> > > > On Thu, Nov 28, 2013 at 3:36 PM, Lucas Ponce <lponce@redhat.com>
> >>> wrote:
> >>> > > >
> >>> > > > > Hello,
> >>> > > > >
> >>> > > > > There is a issue in current master that is impacting to REST
> >>> management
> >>> > > > > services in GateIn.
> >>> > > > >
> >>> > > > > https://issues.jboss.org/browse/GTNPORTAL-3308
> >>> > > > >
> >>> > > > > Investigating it looks like that issue appears since migration to
> >>> > > > > 2.6.0-Alpha3 core.
> >>> > > > >
> >>> > > > > Main symptom is a NPE in POMSessionManager class:
> >>> > > > >
> >>> > > > >
> >>> > > > >
> >>> > >
> >>> https://github.com/gatein/gatein-portal/blob/428874e25ee9e6dada3fd1a54a4aa9f02f42d641/component/portal/src/main/java/org/exoplatform/portal/pom/config/POMSessionManager.java#L224-L229
> >>> > > > >
> >>> > > > > One of the changes of 2.6.0-Alpha3 is to drop the
> >>> > > SessionManagerContainer
> >>> > > > > interface from PortalContainer.
> >>> > > > >
> >>> > > > > Is there a workaround to access to same functionality in order
> >>> to fix
> >>> > > this
> >>> > > > > issue ?
> >>> > > > >
> >>> > > > > Thanks in advance,
> >>> > > > > Lucas
> >>> > > > > _______________________________________________
> >>> > > > > gatein-dev mailing list
> >>> > > > > gatein-dev@lists.jboss.org
> >>> > > > > https://lists.jboss.org/mailman/listinfo/gatein-dev
> >>> > > > >
> >>> > > >
> >>> > >
> >>> >
> >>>
> >>
> >>
> >> _______________________________________________
> >> gatein-dev mailing list
> >> gatein-dev@lists.jboss.org
> >> https://lists.jboss.org/mailman/listinfo/gatein-dev
> >>
> >
> >
> >
> > --
> > *Trong Tran*
> > *(+84) 983841909 <%28%2B84%29%20983841909> | *trongtt@gmail.com
> > Twitter: http://twitter.com/trongtt
> >
>
_______________________________________________ gatein-dev mailing list gatein-dev@lists.jboss.org https://lists.jboss.org/mailman/listinfo/gatein-dev