[
https://issues.apache.org/jira/browse/MYFACES-3840?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13901401#comment-13901401
]
Xavier Cho edited comment on MYFACES-3840 at 2/14/14 1:07 PM:
--------------------------------------------------------------
Sorry for the late response. It seems that it's the famous 'chicken-egg'
problem with using view scope with component binding, as described (and fixed)
by this Mojarra issue :
- https://java.net/jira/browse/JAVASERVERFACES-1492
When PSS is enabled and a managed bean with a component binding is itself view
scoped or depends on one, new instance of the view scoped bean is created as
existing instance cannot be referenced, because the view attributes is not yet
restored when _vdl.buildView (context, view);_ is invoked in
_DefaultFaceletsStateManagementStrategy:222_ (line numbers are from 2.2.0).
However, I wonder exactly when _view.restoreViewScopeState(context,
viewRootState);_ (325 line) is triggered, as if this line is executed, the
above problem can be solved as well.
I verified this behavior, by temporarily override UIViewRoot, so that it's
clientId to be the same with its view ID, which would make the said line to be
executed.
I'm wondering if it could really be the solution to the problem, or for now,
it's just impossible to use component binding with PSS enabled for MyFaces.
was (Author: mysticfall):
Sorry for the late response. It seems that it's the famous 'chicken-egg'
problem with using view scope with component binding, as described (and fixed)
by this Mojarra issue :
- https://java.net/jira/browse/JAVASERVERFACES-1492
When PSS is enabled and a managed bean with a component binding is itself view
scoped or depends on one, new instance of the view scoped bean is created as
existing instance cannot be referenced, as the view attributes is not yet
restored when _vdl.buildView (context, view);_ is invoked in
_DefaultFaceletsStateManagementStrategy:222_ (line numbers are from 2.2.0).
However, I wonder exactly when _view.restoreViewScopeState(context,
viewRootState);_ (325 line) is triggered, as if this line is executed, the
above problem can be solved as well.
I verified this behavior, by temporarily override UIViewRoot, so that it's
clientId to be the same with its view ID, which would make the said line to be
executed.
I'm wondering if it could really be the solution to the problem, or for now,
it's just impossible to use component binding with PSS enabled for MyFaces.
> UIViewRoot uses different id while saving and restoring states.
> ---------------------------------------------------------------
>
> Key: MYFACES-3840
> URL: https://issues.apache.org/jira/browse/MYFACES-3840
> Project: MyFaces Core
> Issue Type: Bug
> Affects Versions: 2.2.0-beta
> Reporter: Xavier Cho
> Assignee: Leonardo Uribe
>
> After I upgraded to 2.2.0-beta, every postback requests which requires
> @ViewScoped managed beans fails as they lose states after the initial request.
> I couldn't spend sufficient time to investigate so not perfectly sure if it's
> not caused by some misconfiguration on my end.
> Though, after a quick debugging, I found that in the
> DefaultFaceletsStateManagementStrategy class, state of an UIViewRoot instance
> is saved using its client ID in saveStateOnMapVisitTree:976, but it tries to
> restore it using its view ID in restoreView:301, thus failing to restore the
> state.
> Is this behavior normal? If so, what possible configuration could cause it to
> use different IDs between saving and restoring state?
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
