Description:
------------
After a recent PHP upgrade (to 5.2.11-r1), some existing code on a few of my sites suddenly "broke."

In both instances, it's XML-related PHP code that silently and completely drops html entities from XML code.

In one instance, it's an RSS feed. "<content:encoded>&lt;p&gt;Lorem..." becomes "<content:encoded>pLorem..."

The (newly) offending code contains the xml_parse_into_struct function.


In the other, it's a CDATA section of an XML-RPC ping. Same problem. The entity-escaped tags are preserved, but without the surrounding lt and gt entities, rendering the payload useless.

This code uses DOMDocument::LoadXML and schemaValidate

Searching a bit turned up the desiccated carcass of bug #35271, but nothing recent that I could find.

Downgraded to PHP 5.2.9-r2. Same problem

Reproduce code:
---------------
    libxml_use_internal_errors(true);
    $xdoc= new DomDocument;
    $xml=$params[1];
    if (!$xml) {
        xmlrpc_error(10, "No payload detected.");
    }
    
    $xmlschema='payload2.xsd';
    $xdoc->LoadXML($xml);
    
    if ($xdoc->schemaValidate($xmlschema)) {

Expected result:
----------------
$xml (payload from incoming XML-RPC ping) is successfully validated against the schema doc), schemaValidate if statement is true, & code inside is executed.

Actual result:
--------------
Schema validation fails with "The document has no document element." A dump of the payload reveals that lt and gt entities have been stripped from the payload: tag attr="true"tag attr="10046"tag /tagtagTag Contents/tagtag  Tag Contents/tag/tag/tag //tag

schemaValidate if statement is false, else code (omitted) is executed, returning aforementioned error to RPC client.