ant elder wrote:
On Sat, Jul 26, 2008 at 12:08 AM, Simon Nash <[EMAIL PROTECTED] <mailto:[EMAIL PROTECTED]>> wrote:Raymond Feng wrote: The system id was set to <namespace> + <a file name generated by JAXB, such as schema3.xsd>. In the failing test case, it was "http://other.ws.binding.sca.tuscany.apache.org/schema3.xsd"; and "other.ws.binding.sca.tuscany.apache.org <http://other.ws.binding.sca.tuscany.apache.org>" was taken as the host name. The system ID has always been set to this value. The recent change I made for TUSCANY-2479 did not change the system ID, but made sure that its value is propagated to the XmlSchema object that is created by Interface2WSDLGenerator.loadXSD(). This allows XmlSchema imports to resolve correctly. I built 1.3 RC2 with an empty maven repo. The modules databinding-jaxb, binding-ws-wsdlgen and binding-ws-axis2 all built OK. We are doing the same thing, but our results are different. One possibility is such cases is a difference between the Sun and IBM JDKs. I'm on Sun JDK 1.5.0_13-b03, so I tried changing to the IBM JDK and this produced the build failure you and Sebastien are seeing when the system ID is set to a non-null string. I applied your change to set the system ID to "" and it seems this works OK for me. I'd like to know whether this also works for Scott in his environment. If it does, it seems we should go with this approach for 1.3 as this appears to be the only combination that works on both the Sun and IBM JDKs.Seems fine to me too if it fixes it but it does work fine for me with both Sun JDK and IBMs "1.5.0" Java(TM) 2 Runtime Environment, Standard Edition (build pwi32dev-20080315 (SR7))
Do you mean RC2 works fine with both of these, without the recent change from Raymond? The IBM JDK that I used for testing was "J2RE 1.5.0 IBM Windows 32 build pwi32dev-20070201 (SR4)". Maybe there is an IBM JDK bug that was fixed in the newer level. Simon
...ant
