I just did some more investigation and what it ends up as is a CannedFormattableValue. Is this what would be expected from an externally received fault message?
-----Original Message----- From: Grant McDonald Sent: Wednesday, 9 August 2006 2:28 AM To: '[email protected]' Subject: RE: Commented code in StartCatchAction I think it's a combination of things: 1) (In ServiceMix JbiInvokeAction) Initially the faulted external action is transformed into a Document, which is then set on an XMLInteraction object and the interaction is added to a BPRuntimeException object which is then thrown (this is simplified: read the fault name/part is determined etc etc): if (me.getFault() != null) { Document fault; try { fault = (Document) transformer.toDOMNode(me.getFault()); me.setStatus(ExchangeStatus.DONE); } catch (Exception e) { me.setError(e); throw new ActionSystemException(e); } finally { channel.send(me); } Element e = fault.getDocumentElement(); // Try to determine fault name String faultName = e.getLocalName(); String partName = BPEComponent.PART_PAYLOAD; QName elemName = new QName(e.getNamespaceURI(), e.getLocalName()); if (wsdlOperation != null) { for (Iterator itFault = wsdlOperation.getFaults().values().iterator(); itFault.hasNext();) { Fault f = (Fault) itFault.next(); Part p = (Part) f.getMessage().getParts().values().iterator().next(); if (elemName.equals(p.getTypeName())) { faultName = f.getName(); partName = p.getName(); } } } BPRuntimeException bpre = new BPRuntimeException(faultName, ""); bpre.setNameSpace(e.getNamespaceURI()); XMLInteractionObject interaction = new XMLInteractionObject(); interaction.setDocument(fault); bpre.addPartMessage(partName, interaction); throw bpre; } else { try { nm = me.getMessage("out"); if (nm != null) { XMLInteractionObject result = new XMLInteractionObject(); result.setDocument((Document) transformer.toDOMNode(nm)); output.put(outputPartName, result); } me.setStatus(ExchangeStatus.DONE); } catch (Exception e) { me.setError(e); throw new ActionSystemException(e); } finally { channel.send(me); } } } else if (me.getStatus() == ExchangeStatus.ERROR) { // Extract error Exception error = me.getError(); throw new BPRuntimeException("Unknown", error); } 2) Coming back out of BPE again we attempt to load the payload in the following manner: protected Source getSourceFromPayload(Object payload) { Source inputSource; if (payload instanceof IFormattableValue) { IFormattableValue value = (IFormattableValue) payload; if (value.supportsGetValueAs(Document.class)) { Document doc = (Document) value.getValueAs(Document.class); inputSource = new DOMSource(doc); } else if (value.supportsGetValueAs(byte[].class)) { byte[] data = (byte[]) value.getValueAs(byte[].class); inputSource = new BytesSource(data); } else if (value.supportsGetValueAs(String.class)) { String data = (String) value.getValueAs(String.class); inputSource = new StringSource(data); } else { throw new UnsupportedOperationException("Unable to retrieve value"); } } else if (payload instanceof Document) { inputSource = new DOMSource((Document) payload); } else if (payload instanceof byte[]) { inputSource = new BytesSource((byte[]) payload); } else if (payload instanceof String) { inputSource = new StringSource((String) payload); } else { throw new UnsupportedOperationException("Unable to retrieve value"); } return inputSource; } In this case the payload (which is the payload of the faultHandlers variable) implements IFormattableValue and in actual fact is an instance FormattableValueAdapter (I added logging to check this). What's strange is that although the value is initially created as a Document object it doesn't return true to the supportedGetValueAs method. The only way I could fudge it was for the else case was to use the value.toString() to generate a string response out of the object (which was in actual fact an xml message). I don't know if any of this sheds some light on the situation or not. Can you tell me how these Interaction objects fit into the picture? -Grant -----Original Message----- From: cory [mailto:[EMAIL PROTECTED] Sent: Wednesday, 9 August 2006 2:11 AM To: [email protected] Subject: Re: Commented code in StartCatchAction It's a bug that the catch by fault type was never fully implemented. I think the commented out code in the StartCatchAction is a remnant from an attempt to implement the catch by fault type. Do you think the exception your are seeing is related to the catch by fault type bug? Is your invoke in the fault handler trying to access a variable that is not initialized or out of scope? -cory On 8/8/06, Grant McDonald <[EMAIL PROTECTED]> wrote: > Hi, > > My colleagues and I have been using the BPE subproject from Apache ODE > integrated with ServiceMix for some time now. During the development of our > BPEL flows it became obvious that the contents of fault messages is not > properly available to faultHandlers and results in the following stacktrace: > > java.lang.UnsupportedOperationException: Unable to retrieve value > at > org.apache.servicemix.bpe.external.JbiInvokeAction.getSourceFromPayload(JbiI > nvokeAction.java:276) > at > org.apache.servicemix.bpe.external.JbiInvokeAction.execute(JbiInvokeAction.j > ava:179) > at > org.apache.ode.bpe.action.bpel.ExternalServiceAction.execute(ExternalService > Action.java:230) > at > org.apache.servicemix.bpe.external.JbiExternalAction.execute(JbiExternalActi > on.java:129) > at > org.apache.ode.bpe.engine.ProcessInstance.executeActions(ProcessInstance.jav > a:359) > at > org.apache.ode.bpe.engine.ProcessInstance.evaluate(ProcessInstance.java:325) > at > org.apache.ode.bpe.engine.ProcessInstance.evaluate(ProcessInstance.java:414) > at > org.apache.ode.bpe.engine.ProcessInstance.processInternalEvent(ProcessInstan > ce.java:254) > at > org.apache.ode.bpe.engine.ProcessInstance.processEvent(ProcessInstance.java: > 209) > at > org.apache.ode.bpe.scope.service.impl.FCScopeInstanceImpl.executeProcess(FCS > copeInstanceImpl.java:392) > at > org.apache.ode.bpe.scope.service.impl.FCScopeInstanceImpl.handleFault(FCScop > eInstanceImpl.java:531) > at > org.apache.ode.bpe.engine.ProcessInstance.processInternalEvent(ProcessInstan > ce.java:287) > at > org.apache.ode.bpe.engine.ProcessInstance.processEvent(ProcessInstance.java: > 209) > at > org.apache.ode.bpe.correlation.CorrelationService.createInstanceAndRoute(Cor > relationService.java:231) > at > org.apache.ode.bpe.correlation.CorrelationService.correlateEvent(Correlation > Service.java:338) > at > org.apache.ode.bpe.bped.unmanaged.EventDirectorSLImpl.sendEvent(EventDirecto > rSLImpl.java:116) > at > org.apache.servicemix.bpe.BPEEndpoint.process(BPEEndpoint.java:146) > > <snip> > > After do some code diving I came across areas of the code I'd like to ask > about: > > 1) FScopeInstanceImpl - the HashMap which contains the faultHandlers keyed > by faultType is never actually populated by fault type since the > addFaultHandler method is only called from StartScopeAction which populates > the faultType as "" (and hence the ServiceMix integration layer does not > populate the faultType on the BPRuntimeException). Was this a planned > implementation that never got finished? > > 2) StartCatchAction - this class has commented out code that deals with > faultTypes created on the faultHandler and the creation of an Interaction to > hold the faultMessage contents instead of using the straight data object. > Was this also an area that was yet to be fully implemented? What was the > intention of the commented out code and was there some caveat with the > approach that was previously being taken? > > I am aware that the majority of work is being done in trunk now as PXE and > BPE have been subsumed into ODE now. Any help would be appreciated in this > area. > > Kind regards, > > Grant McDonald > >
