[
https://issues.apache.org/struts/browse/STR-2976?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Paul Benedict updated STR-2976:
-------------------------------
Fix Version/s: 1.4.0
> NestedRootTag and NestedPropertyTag's doStartTag() could return
> EVAL_BODY_INCLUDE
> ---------------------------------------------------------------------------------
>
> Key: STR-2976
> URL: https://issues.apache.org/struts/browse/STR-2976
> Project: Struts 1
> Issue Type: Improvement
> Components: Taglibs
> Affects Versions: 1.3.5
> Reporter: Hayden Foot
> Fix For: 1.4.0
>
>
> The nested:nest (NestedRootTag) and nested:root (NestedRootTag) only set
> context and scope for other tags within its body, they do not output any
> content themselves or manipulate the content with in their body, and yet they
> return EVAL_BODY_TAG (equivalent to EVAL_BODY_BUFFERED) rather than
> EVAL_BODY_INCLUDE. This means a new BodyContent object is created for the
> body rather than using the existing one (passed through to the current out).
> This can be an issue if the body of either a nested:nest or a nested:root tag
> generates a large amount of data because the BodyContent object of the tag's
> body will not flush to the underlying stream until the tag's body is
> complete. When running under Tomcat the BodyContent implementation is
> org.apache.jasper.runtime.BodyContentImpl, and because these objects are
> pooled and their underlying buffers (char[]) will grow (but not shrink), when
> rendering large pages (i.e. large content inside nested:nest or nested:root
> tags) the BodyContent object used for the tag's body can grow very large
> (because it can not flush) and will also be retained after the page has
> rendered (because the object is pooled - hence not released).
> I have experimented by extending the NestedPropertyTag and NestedRootTag
> classes and overriding the doStartTag() method in the following way:
> public int doStartTag() throws JspException {
> super.doStartTag();
> return EVAL_BODY_INCLUDE;
> }
> In a test case, using this implementation rather than returning
> EVAL_BODY_TAG, meant the body of the tag could stream/flush and hence did not
> need to grow the BodyContent buffer. The result was that the retained memory
> (of BodyContentImpl) was significantly reduced (approx 35k as opposed to >
> 100M).
> I proposed that you change the implementation of both NestedRootTag and
> NestedPropertyTag's doStartTag() method to return EVAL_BODY_INCLUDE as there
> is no obvious reason why it needs to buffer the body.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
