OutputText Not Rendering w/ Verbatim Tag (Client Mode)

[eclipse-ee4j]

OutputText Not Rendering w/ Verbatim Tag (Client Mode)

[ren-zhijun-oracle]

• Issue Imported From: https://github.com/javaserverfaces/mojarra/issues/16
• Original Issue Raised By:@javaserverfaces
• Original Issue Assigned To: @glassfishrobot
• Original Issue Closed By:@javaserverfaces

[ren-zhijun-oracle]

@javaserverfaces Commented
Reported by rogerk

[ren-zhijun-oracle]

@javaserverfaces Commented
rogerk said:
Created an attachment (id=7)
Patch For Issue 12

[ren-zhijun-oracle]

@javaserverfaces Commented
rogerk said:
The bug only surfaces in "client side" state saving.

M src/javax/faces/component/UIComponentBase.java

• ony store a child's state if state is not null

M systest/src/com/sun/faces/systest/NavigationTestCase.java

• new test case with verbatim tag
  A systest/web/jsp/verbatim-one-test.jsp

Index: UIComponentBase.java

RCS file:
/cvs/javaserverfaces-sources/jsf-api/src/javax/faces/component/UIComponentBase.java,v
retrieving revision 1.101
diff -u -r1.101 UIComponentBase.java
— UIComponentBase.java 16 Jun 2004 00:06:55 -0000 1.101
+++ UIComponentBase.java 19 Jul 2004 16:05:38 -0000
@@ -943,7 +943,6 @@

• @exception NullPointerException {@inheritdoc}

  */
  public Object processSaveState(FacesContext context) {

• if (context == null)

  { throw new NullPointerException(); }

  @@ -964,7 +963,10 @@
  Iterator kids = getChildren().iterator();
  while (kids.hasNext()) {
  UIComponent kid = (UIComponent) kids.next();

• childState[i++] = kid.processSaveState(context);

  • Object savedState = kid.processSaveState(context);
  • if (savedState != null) { + childState[i++] = savedState; + }

  }

Iterator myFacets = getFacets().keySet().iterator();

Index: NavigationTestCase.java

RCS file:
/cvs/javaserverfaces-sources/jsf-ri/systest/src/com/sun/faces/systest/NavigationTestCase.java,v
retrieving revision 1.5
diff -u -r1.5 NavigationTestCase.java
— NavigationTestCase.java 1 May 2004 00:48:43 -0000 1.5
+++ NavigationTestCase.java 19 Jul 2004 16:06:19 -0000
@@ -120,4 +120,28 @@
assertTrue(false);
}
}
+

• public void testNavigateWithVerbatim_One() throws Exception {
• HtmlForm form;

[ren-zhijun-oracle]

@javaserverfaces Commented
@edburns said:

On Mon, 19 Jul 2004 12:40:51 -0400, Roger Kitain [email protected] said:

I appreciate your thorough testing. Something this core has to be
thoroughly tested.

I don't understand how this change fixes the bug however. Can you
please explain it?

RK> Index: UIComponentBase.java
RK> ===================================================================
RK> RCS file:
/cvs/javaserverfaces-sources/jsf-api/src/javax/faces/component/UIComponentBase.java,v
RK> retrieving revision 1.101
RK> diff -u -r1.101 UIComponentBase.java
RK> — UIComponentBase.java 16 Jun 2004 00:06:55 -0000 1.101
RK> +++ UIComponentBase.java 19 Jul 2004 16:05:38 -0000
RK> @@ -943,7 +943,6 @@
RK> * @exception NullPointerException

{@inheritdoc}

RK> */
RK> public Object processSaveState(FacesContext context) {
RK> -
RK> if (context == null)

{ RK> throw new NullPointerException(); RK> }

RK> @@ -964,7 +963,10 @@
RK> Iterator kids = getChildren().iterator();
RK> while (kids.hasNext()) {
RK> UIComponent kid = (UIComponent) kids.next();
RK> - childState[i++] = kid.processSaveState(context);
RK> + Object savedState = kid.processSaveState(context);
RK> + if (savedState != null)

{ RK> + childState[i++] = savedState; RK> + }

RK> }

I would think that you'd want to increment "i" each time around the
while loop, regardless of savedState being null. Otherwise, the
restoreState may not synch up, right? Perhaps I'm missing something.

Also, you're going to have a null in that slot anyway, even if you skip
filling it in with the null return from processSaveState(). We
initialize childState to be new Object[len].

Can you please explain why this works?

Thanks,

Ed

[ren-zhijun-oracle]

@javaserverfaces Commented
rogerk said:
Additional Background:
The current client side state saving processing is broken in some cases
(primarily when "verbatim" components are used). A cause of this is
a "misallignment" when processing state save/restore. Consider the
following example:

JSP
—
<f:view>
<h:form id="form">

<h:panelGrid id="panel" columns="2">
<f:facet name="header" >
<h:outputText id="outputheader" value="this is the header" />
</f:facet>

<h:commandLink id="commandLink">
<h:outputText id="redisplay" value="redisplay"/>
</h:commandLink>

<f:verbatim>
verbatim text
</f:verbatim>
</h:panelGrid>

</h:form>
</f:view>

Let's focus on the verbatim tag and the components surrounding the
verbatim tag. At the completion of save state processing, we end up
with the following:

In state saving terms, the panel component ends up with the following
structure:

The "Panel" Component has 3 children:
childState[0] Component:commandLink State:[Ljava.lang.Object;@1e090ee
childState[1] Component:_id0 (This is verbatim tag) State: null
childState[2] Component:outputHeader State:[Ljava.lang.Object;@1123eb0

When the restore state process kicks in, we end up with the following:
Component:commandLink is restored with state from "childState[0]" (correct)
Component:outputHeader is restored with state from "childState[1]

• • which is null ** hence, this component will not appear on postback.
    this component should have been restored with "childState[2]"...

As explained in the issue comments, there are two possible fixes
for this state saving bug:

• Option 1: api code change (deal with ignoring transient children
  and facets during state saving process in UIComponentBase).
  The tree state that will be saved will not include
  "verbatim" component slots.
• Option 2: ri code change (remove transient/facets from the physical tree
  in StateManagerImpl for client side state saving. This is
  already done for server side state saving). This would be
  done before the API (UIComponentBase.processSaveState) is
  executed, hence, end result would be the same as
  Option 1 - the tree state that will be saved will not
  include "verbatim" components slots.

Keep in mind the following:
The spec (javadoc comments in the api - UIComponent.processSaveState)
mention that as each component is examined, the "transient" property should
be consulted, and if true, return "null";
The spec (javadoc comments in the api - StateManager) say:

"Components may opt out of being included in the serialized view
by setting their transient property to true.
This must cause the component itself, as well as all of that component's
children and facets, to be omitted from the saved tree structure
and component state information."

To me, this implies that we can also handle transient components in
the StateManager implementation.

Note: If Option 2 is used (the RI approach), then I can't think of a reason,
that there would be transient components in the tree during
API (UIComponentBase.processSaveState) processing - I kept the
checks in because the spec mentions it...

**Option 1 Change Bundle **

M src/javax/faces/component/UIComponentBase.java

M systest/src/com/sun/faces/systest/NavigationTestCase.java
A systest/web/jsp/verbatim-one-test.jsp
A systest/web/jsp/verbatim-two-test.jsp

Index: UIComponentBase.java

RCS file:
/cvs/javaserverfaces-sources/jsf-api/src/javax/faces/component/UIComponentBase.java,v
retrieving revision 1.101
diff -u -r1.101 UIComponentBase.java
— UIComponentBase.java 16 Jun 2004 00:06:55 -0000 1.101
+++ UIComponentBase.java 22 Jul 2004 15:35:27 -0000
@@ -951,20 +951,20 @@
return null;
}
Object [] stateStruct = new Object[2];

• Object [] childState = null;

// Process this component itself
stateStruct[MY_STATE] = saveState(context);

// Process all the children of this component

• int i = 0, len = getChildren().size() + getFacets().keySet().size();

• childState = new Object[len];

• stateStruct[CHILD_STATE] = childState;

  • List nonTransientState = new ArrayList();
    Iterator kids = getChildren().iterator();
    while (kids.hasNext()) {
    UIComponent kid = (UIComponent) kids.next();

• childState[i++] = kid.processSaveState(context);

  • if (kid.isTransient()) { + continue; + }

  • nonTransientState.add(kid.processSaveState(context));
    }

Iterator myFacets = getFacets().keySet().iterator();
@@ -975,18 +975,16 @@
while (myFacets.hasNext()) {
facetName = (String) myFacets.next();
facet = (UIComponent) getFacets().get(facetName);

• if (!facet.isTransient()) { - facetState = facet.processSaveState(context); - facetSaveState = new Object[1][2]; - facetSaveState[0][0] = facetName; - facetSaveState[0][1] = facetState; - childState[i] = facetSaveState; - }

• else {

• childState[i] = null;

  • if (facet.isTransient()) { + continue; }

• i++;

  • facetState = facet.processSaveState(context);
  • facetSaveState = new Object[1][2];
  • facetSaveState[0][0] = facetName;
  • facetSaveState[0][1] = facetState;
  • nonTransientState.add(facetSaveState);
    }
  • stateStruct[CHILD_STATE] = nonTransientState.toArray();
    return stateStruct;
    }
• • Option 2 Change Bundle **

M src/com/sun/faces/application/StateManagerImpl.java

M systest/src/com/sun/faces/systest/NavigationTestCase.java
A systest/web/jsp/verbatim-one-test.jsp
A systest/web/jsp/verbatim-two-test.jsp

Index: StateManagerImpl.java

RCS file:
/cvs/javaserverfaces-sources/jsf-ri/src/com/sun/faces/application/StateManagerImpl.java,v
retrieving revision 1.22
diff -u -r1.22 StateManagerImpl.java
— StateManagerImpl.java 1 Jun 2004 17:06:26 -0000 1.22
+++ StateManagerImpl.java 22 Jul 2004 15:37:16 -0000
@@ -101,6 +101,7 @@
sessionMap.put(FACES_VIEW_LIST, viewList);
}
} else {

• removeTransientChildrenAndFacets(context, viewRoot, new HashSet());
  if (log.isDebugEnabled())

{ log.debug("Begin creating serialized view for " + viewRoot.getViewId()); Test Case Modifications/Additions Common To Both Options: Index: NavigationTestCase.java =================================================================== RCS file: /cvs/javaserverfaces-sources/jsf-ri/systest/src/com/sun/faces/systest/NavigationTestCase.java,v retrieving revision 1.5 diff -u -r1.5 NavigationTestCase.java --- NavigationTestCase.java 1 May 2004 00:48:43 -0000 1.5 +++ NavigationTestCase.java 22 Jul 2004 15:40:46 -0000 @@ -120,4 +120,58 @@ assertTrue(false); }

}
+

• public void testNavigateWithVerbatim_One() throws Exception {
• HtmlForm form;
• HtmlSubmitInput submit;
• HtmlPage page, page1;
• page = getPage("/faces/jsp/verbatim-one-test.jsp");
• form = getFormById(page, "form");
• assertNotNull("form exists", form);
• submit = (HtmlSubmitInput)
• form.getInputByName("form" + NamingContainer.SEPARATOR_CHAR +
• "submit");
• // press the link, return to the same page, and check that
• // output text (header) is still present...
• try

{ + page1 = (HtmlPage) submit.click(); + assertTrue(-1 != page1.asText().indexOf("this is the header")); + }

catch (Exception e)

{ + e.printStackTrace(); + assertTrue(false); + }

• }

• public void testNavigateWithVerbatim_Two() throws Exception {

• HtmlForm form;

• HtmlSubmitInput submit;

• HtmlPage page, page1;

• page = getPage("/faces/jsp/verbatim-two-test.jsp");

• form = getFormById(page, "form");

• assertNotNull("form exists", form);

• submit = (HtmlSubmitInput)

• form.getInputByName("form" + NamingContainer.SEPARATOR_CHAR +

• "submit");

• // submit the form, return to the same page, and check that

• // output text (header) is still present...

• // and verbatim text is still present...

• try { + page1 = (HtmlPage) submit.click(); + assertTrue(-1 != page1.asText().indexOf("verbatim one text here")); + assertTrue(-1 != page1.asText().indexOf("this is the header")); + assertTrue(-1 != page1.asText().indexOf("verbatim two text here")); + } catch (Exception e) { + e.printStackTrace(); + assertTrue(false); + }

• }

verbatim-one-test.jsp

<%@ taglib uri="http://java.sun.com/jsf/core" prefix="f" %> <%@ taglib uri="http://java.sun.com/jsf/html" prefix="h" %>

<f:view>

<title></title>

<f:facet name="header" >
<h:panelGrid id="panel2" columns="1" >
<h:outputText id="outputheader" value="this is the header" />
<f:verbatim>

---

</f:verbatim>
</h:panelGrid>
</f:facet>

<h:commandButton id="submit" value="submit"/>

<f:verbatim >
verbatim text here
</f:verbatim>

</h:panelGrid>
</h:form>

verbatim-two-test.jsp

<%@ taglib uri="http://java.sun.com/jsf/core" prefix="f" %> <%@ taglib uri="http://java.sun.com/jsf/html" prefix="h" %>

<f:view>

<title></title>

<f:verbatim >
verbatim one text here
</f:verbatim>

<h:panelGrid id="panel2" columns="1" >
<h:outputText id="outputheader" value="this is the header" />
<f:verbatim>

---

</f:verbatim>
</h:panelGrid>

<h:commandButton id="submit" value="submit"/>

<f:verbatim >
verbatim two text here
</f:verbatim>

</h:panelGrid>
</h:form>

Comments are welcome at this point....

[ren-zhijun-oracle]

@javaserverfaces Commented
@edburns said:
Option 2 is certainly simpler, but couldn't you move the
removeTransientChildrenAndFacets() call above the if()else{} block, since in
your new world it would execute in either case?

If you assert that option 2 works, and I can indeed see how it would at a top
level, you can go with option 2.

r=edburns

[ren-zhijun-oracle]

@javaserverfaces Commented
rogerk said:
Went with Option 2.

[ren-zhijun-oracle]

@javaserverfaces Commented
@manfredriem said:
Closing issue out

[ren-zhijun-oracle]

@javaserverfaces Commented
File: patch.tar.gz
Attached By: rogerk

[ren-zhijun-oracle]

@javaserverfaces Commented
Was assigned to rogerk

[ren-zhijun-oracle]

@javaserverfaces Commented
This issue was imported from java.net JIRA JAVASERVERFACES-12

[ren-zhijun-oracle]

@javaserverfaces Commented
Marked as fixed on Wednesday, April 19th 2006, 5:44:32 am