A testcase would be great! I suspect a Facelets bug here, though it's kind of a corner case.
-- Adam On 7/9/07, noah <[EMAIL PROTECTED]> wrote:
On 7/6/07, Adam Winer <[EMAIL PROTECTED]> wrote: > On 7/6/07, noah <[EMAIL PROTECTED]> wrote: > > On 7/6/07, Gary VanMatre <[EMAIL PROTECTED]> wrote: > > > > > > > > > >From: noah <[EMAIL PROTECTED]> > > > > > > > > On 7/6/07, Gary VanMatre wrote: > > > > > > > > > > >From: noah > > > > > > > > > > > > Thought I'd post this before creating a JIRA issue, in case I've > > > > > > overlooked something. I just now produced this with 1.2.1 but I think > > > > > > I've seen it in 1.0.x before. > > > > > > > > > > > > In the following page, clicking the link correctly updates 'Foo=' to > > > > > > 'Foo=1' but it renders the paragraph twice. Clicking again updates > > > > > > -both- paragraphs to 'Foo=2' and adds a third. Click again for a 4th, > > > > > > and Foo=3, and so on. > > > > > > > > > > > > Using an h:outputText instead of inline EL keeps the text from being > > > > > > repeated, but additional paragraph elements are added with each click. > > > > > > > > > > > > > > > > > > > > >> ><tr:document xmlns="http://www.w3.org/1999/xhtml"; > > > >> > xmlns:ui="http://java.sun.com/jsf/facelets"; > > > >> > xmlns:f="http://java.sun.com/jsf/core"; > > > >> > xmlns:h="http://java.sun.com/jsf/html"; > > > >> > xmlns:tr="http://myfaces.apache.org/trinidad";> > > > >> > <tr:form> > > > >> > <tr:panelHeader partialTriggers="foo"> > > > >> > <p>Foo=#{pageFlowScope.foo}</p> > > > > ;> > </tr:panelHeader> > > > >> > & nbsp;& nbsp; <tr:commandLink id="foo" partialSubmit="true" text="Do > > > >> it!"> > > > >> > <f:setPropertyActionListener value="#{pageFlowScope.foo + 1}" > > > >> >target="#{pageFlowScope.foo}" /> > > > >> > </tr:commandLink> > > > >> > </tr:form> > > > >> ></tr:document> > > > > > > > > > > > > > > > > > > > > > > > > > What happens if you use the output text with the transient attribute? > > > > > > > > No change, same behavior as transient="false". The paragraph tags > > > > still get repeated of course. I suppose I shouldn't say 'transient' > > > > but all HTML, inline text, and EL. > > > > > > > > > > > > > > > > Are you using client or server side state saving? > > > > Client. > > > > > > > > I'm not familiar with the specifics of Facelets but I believe that it is > > > creating its own jspid to simulate a "JspIdConsumer" and adds it to the > > > component's attributes. This is how Clay currently works after a few tips > > > from Jacob. When the component tree is restored, the components that are > > > marked transient fall out. Transient components have to be recreated and > > > inserted into the component tree. The "jspid" marker is used to map > > > existing components to the meta-data used to build the tree. > > > > > > It sounds like the markup nodes are not being set as transient by facelets > > > or maybe something in the state saving is not discarding the transient > > > components. > > > > > > Can you tell if the outputText components that contain the markup "<p>..." > > > are "transient=true"? > > > > I don't know how to do that. I don't think Facelets uses actual > > h:outputText components though. > > > > > What happens if you switch state saving modes? > > > > No copies. > > > > But, the rest of my application blows up for some reason :) > > So, which mode has the problem with copies getting > created? If it's client-side state saving, what happens > if you set org.apache.myfaces.trinidad.CACHE_VIEW_ROOT > to false in web.xml? Client has the copies. If CACHE_VIEW_ROOT is false, it also keeps it from happening, even with client-side state saving. > > And, does: > > <h:outputText transient="true" escape="false" > value="<p>#{pageFlowScope.foo}</p>"/> > > ... dodge the issue? > Yes it does. Although doing that kind of half defeats the point of using Facelets :) > -- Adam >
