[
https://issues.apache.org/jira/browse/MYFACES-4125?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16125986#comment-16125986
]
Eduardo Breijo commented on MYFACES-4125:
-----------------------------------------
I checked the JIRA issue that introduced the flush call and I noticed that the
behavior is gated behind context parameter
"org.apache.myfaces.EARLY_FLUSH_ENABLED". Maybe we can use this same context
parameter in the HtmlResponseWriterImpl.endDocument() to perform the flush only
when this parameter is set to true.
public void endDocument() throws IOException
{
MyfacesConfig myfacesConfig =
MyfacesConfig.getCurrentInstance(FacesContext.getCurrentInstance().getExternalContext());
if (myfacesConfig.isEarlyFlushEnabled())
{
_currentWriter.flush();
}
_facesContext = null;
}
> Response committed too early due to flush from StateWriter
> ----------------------------------------------------------
>
> Key: MYFACES-4125
> URL: https://issues.apache.org/jira/browse/MYFACES-4125
> Project: MyFaces Core
> Issue Type: Bug
> Affects Versions: 2.2.12
> Reporter: Eduardo Breijo
> Assignee: Leonardo Uribe
> Attachments: server.log, StateWriter.war
>
>
> We've found a problem where it seems that MyFaces is flushing output too
> early in the RENDER_RESPONSE PHASE. As a result the response is committed
> before we have a chance to handle an error in that phase.
> This is because the renderView method from FaceletViewDeclarationLanguage
> calls writer.endDocument() which ends up calling the flush() method from
> StateWriter. This commit behavior is different between 2.0 and 2.2. It looks
> like a flush() was empty on 2.0, and now a call was added, which causes the
> issue we are seeing.
> Here's a sample app:
> 1) Drive a request: localhost:9080/StateWriter/index.xhtml
> 2) You should be able to see in the logs that response was committed, so
> redirect to error.xhtml cannot be performed.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
