Martin Clough wrote:
We have a Tuscany 1.5 installation under Tomcat on one server (the application server), which is accessed using Apache httpd and the Apache/Tomcat connector on a separate web server.

We are exposing SOAP web services in Tuscany, but the WSDL for these is incorrect because the SOAP address location has the application server address instead of the web server name:

This is what we get:

<wsdl:service name="Service">

<wsdl:port name="Port" binding="tns:Binding">

<SOAP11:address location="http://*[ip address of app server]*/serviceurl"/>

</wsdl:port>

</wsdl:service>

This is what we need:

<wsdl:service name="Service">

<wsdl:port name="Port" binding="tns:Binding">

<SOAP11:address location="http://*webserver*/serviceurl"/>

</wsdl:port>

</wsdl:service>

Without this change, clients using Java 6 JAXB cannot connect to the web service.

I have corrected this with an updated tuscany-binding-ws-axis2-1.5.jar which has fix in org.apache.tuscany.sca.binding.ws.axis2.TuscanyListingAgent to the setIPAddress method. The fix is:

private static String setIPAddress(String wsdlURI, String requestURI) {

try {

URI wsdlURIObj = new URI(wsdlURI);

String wsdlHost = wsdlURIObj.getHost();

int wsdlPort = wsdlURIObj.getPort();

String wsdlAddr = wsdlHost + (wsdlPort != -1 ? ":" +

Integer.toString(wsdlPort) : "");

URI requestURIObj = new URI(requestURI);

-              String ipAddr = HttpUtils.getIpAddress();

+              String ipAddr = requestURIObj.getHost();

int requestPort = requestURIObj.getPort();

String newAddr = ipAddr + (requestPort != -1 ? ":" +

Integer.toString(requestPort) : "");

return wsdlURI.replace(wsdlAddr, newAddr);

} catch (Exception e) {

// URI string not in expected format, so return the WSDL URI unmodified

return wsdlURI;

}

}

I'm happy to create a JIRA and submit a patch to this effect. However, I suspect that this undermines the purpose of this method. So my questions are: what is the correct way to remedy this? Would this patch work for you? Is there another way of fixing this?

Martin,
Thanks for reporting this.  I think the inconsistency between the
treatment of ipAddr and requestPort is accidental rather than deliberate.
So please go ahead and create a JIRA with your patch, and I will commit
the patch unless someone else comes up with a reason for not making this
change.

  Simon

Many thanks

Martin




Reply via email to