Major Description
From mailing list discussion with Gary VanMatre:
>From: "Richard Eggert" <reggert@proteus-technologies.com>
>
> I'm running into a problem using "Xml views" using HTML templates. Loading a
> page initially works fine, but if I reload the page either by clicking on a link
> or hitting Refresh, I get the following error.
>
>
> java.lang.IllegalStateException: Client-id : _id0 is duplicated in the faces
> tree.
> at
> org.apache.myfaces.application.jsp.JspStateManagerImpl.checkForDuplicateIds(JspS
> tateManagerImpl.java:241)
> at
> org.apache.myfaces.application.jsp.JspStateManagerImpl.checkForDuplicateIds(JspS
> tateManagerImpl.java:255)
> (the above line is repeated several times)
> at
> org.apache.myfaces.application.jsp.JspStateManagerImpl.savedSerializedView(JspSt
> ateManagerImpl.java:204)
> at
> org.apache.shale.clay.faces.ClayViewHandler.renderView(ClayViewHandler.java:418)
> ...
>
> The error goes away if I load the page again, and then comes back if I load the
> page again after that.
>
> I created a very simple test page to reproduce the problem:
>
> bugTest.html
> ---------------
><!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Strict//EN">
><html>
> <head>
> <title>Testing</title>
> <meta http-equiv="Content-Type" content="text/html;charset=UTF-8" />
> <meta http-equiv="pragma" content="no-cache">
> <meta http-equiv="cache-control" content="no-cache">
> <meta http-equiv="expires" content="0">
> </head>
> <body>
> foo
> </bod y>
></html>
>
>
>
>
> clay-views.xml (excerpt)
> ----------------------------
>
><component jsfid="/bugTest.clay" extends="clay">
> <attributes>
> <set name="clayJsfid" value="/path/to/bugTest.html" />
> </attributes>
></component>
>
> (web.xml maps *.clay to the Faces servlet and the Shale application filter)
>
> I don't know if this is a bug in MyFaces (I'm using version 1.1.1), a bug in
> Clay, or if I'm doing something wrong, but I've never seen this problem before
> with my other (non-Clay-based) pages. Any ideas?
>
Darn, I thought we had this one covered for all the releases but I must not have tested again with myfaces 1.1. There is some history with this issue.
This has to do with how the view root generates component ids. For components that you don't explicitly assign a component id, there is a utility function on the view root to generate a unique Id. In the first versions of Myfaces and the RI, the internal sequence that generates the ids was always reset even after the view was restored.
This is a significant because of components marked transient. Transient components do not save state in the view. The "verbatim" component is actual an outputText with the transient property tunned on. When the view is restored, we have to recreate transient components and these components will need generated component id's since this is just text. In the early version of clay, after the view was restored, we would generate, in sequence, component ids for all components including components that were restored because this counter was always reset. A clay html template contains allot of transient components because the entire template is defined using components (Thanks again to Manfred Klug who helped figure this all out). Interesting enough, making all the view represented by components is the same approach they took in JSF 1.2.
The EG team determined that reseting the counter was a "bug". The view root should restore the internal counter. Starting with version 1.1.2, myfaces changed the view roots to restore the counter.
Clay Release 1.0.3
http://issues.apache.org/struts/browse/SHALE-195
The counter being reset made PPR or any changes to a restored component tree problematic. So, in JSF 1.2 they changed how component id's are generated. They are now generated by the the container for JSP tags implementing JspidConsumer (JSP 2.1). The JspidConsumer will always return the same id for a component.
Most recently, changes were made in this area to allow Clay to work with JSF 1.2.
Clay Release 1.0.4
http://issues.apache.org/struts/browse/SHALE-67
I thought that I had this covered for all releases but it sounds like it's an issue. Trying to play with all versions is challenge. I'd like to move the trunk to full JSF 1.2 support and leave the 1_0_X branch for JSF 1.1 flavors. There are features in JSF 1.2 that we can not support until we commit to the new EL API.