Insufficient optimisation in JBossXSModel traversing XML schema

The JBossXSModel implementation requires further optimisation to reduce deployment overhead for complex schemas.

Take the following type: -

<xs:complexType abstract='true' name='cmisBaseType'>
<xs:sequence>
<xs:choice maxOccurs='unbounded' minOccurs='0'>
<xs:element ref='tns:base'/>
<xs:element ref='tns:baseDateTime'/>
<xs:element ref='tns:baseBoolean'/>
<xs:element ref='tns:baseHtml'/>
<xs:element ref='tns:baseInteger'/>
<xs:element ref='tns:baseDecimal'/>
<xs:element ref='tns:baseString'/>
<xs:element ref='tns:baseId'/>
<xs:element ref='tns:baseUri'/>
<xs:element ref='tns:baseXml'/>
</xs:choice>
</xs:sequence>
</xs:complexType>

If each of the element types referenced here extend this type then there are many routes to reach each type. The class JBossXSModel does detect cyclic references but it does not detect if a type has already been processed if re-visit in a non cyclic way.

This Jira issue it to modify the current optimisation to detect second visits to a type by any route.