Re: Problem with Pluto Portlets and the Sun JSF Portlets

If you are using the Bridge I pointed you at it also has the benefit of
being Faces impl independent. It will run with the Faces 1.2 RI/Sun's
Faces impl. In fact I do most of my testing with a version of the RI.

-Mike-

Domazlicky, Eric wrote:

Well
the Apache Bridge seems to work. Unfortunately though
NetBean’s Visual Designer seems to be tied to Sun JSF’s so that
means you can’t use the Designer to design a Portlet. I can get a pure
MyFaces example portlet to work using the Apache JSF Bridge however so
at least
I can confirm that Faces can work under uPortal/PLUTO.

From:
Michael Freedman
[ mailto:[email protected] ]

Sent: Wednesday, August 06, 2008 2:07 PM

To: [email protected]

Subject: Re: Problem with Pluto Portlets and the Sun JSF
Portlets

Definitely let me know if you run into any
problems running
your portlet with this bridge -- the JSR is supposed to define standard
behavior for all bridges -- but obviously there is/was alot of existing
prior
art so I am interested in seeing if/what complications occur when users
migrate
so I can determine if its an issue related to specific implementation
dependent
function in the legacy bridge or the standard's definition needs
expanding.

-Mike-

Domazlicky, Eric wrote:

Thanks
I’ll give the Apache portlet-bridge a try and see
what I can find.

From:
Michael Freedman [ mailto:[email protected] ]

Sent: Wednesday, August 06, 2008 1:47 PM

To: [email protected]

Subject: Re: Problem with Pluto Portlets and the Sun JSF
Portlets

Request and Response wrapping is not supported
in the
Portlet 1.0 (JSR 168) standard hence the use of a request wrapper by
the Sun
JSF bridge ( it is the Sun bridge, right?) is implementation dependent
and can
only be assumed to work in a Sun portlet container. Instead, try the
JSR
301 Bridge whose (early) Reference Implementation is on Apache ( http://myfaces.apache.org/portlet-bridge/index.html ).
I would suggest getting the sources and building yourself as the last
release
was posted coincident with the last early draft earlier this year and
things have progressed a bit since then. (Though that binary release
is
also stable and fully functional for most uses). Note: this impl is
intended to be portlet container independent. It doesn't rely on
wrapping.

When you look at this Apache project you will notice there are two code
lines. The JSR is producing a spec and impl for both portlet 1.0 and
portlet 2.0 with JSF 1.2. Portlet 2.0 (JSR 286) does define wrapping,
hence its the portlet 2.0 bridge version that wanted to use wrapping
and hit the
bug referenced by Craig.

-Mike-

[email protected]
wrote:

PLUTO-478
covers this issue. It was reported by Mike Freedman, spec lead of
JSR-301, the
JSF-Portlet Bridge.

Any
fix to this issue should be done on the 1.1.x and 2.0-refactoring
branch
and the trunk.

/Craig

[email protected]
wrote:
-----

To: [email protected]

From: [email protected]

Date: 08/05/2008 07:29PM

Subject: Re: Problem with Pluto Portlets and the Sun JSF Portlets

>
I had
directed Eric over here from the uPortal lists.

>

> It appears that the JSF-Portlet library adds a response wrapper
before

> calling the request dispatcher. Does anyone here know if that is
allowed

> under JSR-168? There is no mention I can find in the spec or
errata. I

> do know in 286 explicit wrapper classes were added that mimic the

> servlet API wrapper classes but that is all I've been able to find.

> Should this be viewed as a Pluto bug or a JSF-Portlet bug?

>

> -Eric

I would expect the PortletRequestDispatcher.include() should accept any

implementation of PortletRequest/Response, but that is probably not

guaranteed by the spec. Even if this is not a Pluto bug, I wonder if a

simple fix might be to use a ThreadLocal to store the internal

request/response objects when o.a.p.core.PortletServlet passes the
request

off to the portlet, rather than using wrapper classes.

Actually, as I look at the code, I wonder if

PortletRequestDispatcherImpl.include should wrap the

PortletRequest/Response objects in adapters that implement

HttpServletRequest/Response rather than unwrap them to pull out the

internal implementation. That way, the client's wrapper classes could

still function as expected. I could put together a patch for this if
you

think it is a good idea.

-- Ben

P.S. - I'm replying to pluto-user, but I wonder if this is more

appropriate for pluto-dev?