Please ignore the 2nd part of my previous note -- It was part of an earlier composition that I meant to delete.
Kelvin. On 09/10/2007, kelvin goodson <[EMAIL PROTECTED]> wrote: > Hi Ron, > > Ron, sorry not to get to this earlier. I just took a look at this, > expecting to have to dig quite deep into an area I'm not very familiar > with, but debugging through a test case I spotted an anomaly [1] > which strikes a chord with your description. I'll commit a fix. Can > you let me know if it addresses this issue please? > > [1] https://issues.apache.org/jira/browse/TUSCANY-1838 > > > It seems clear the instanceof test and the cast should be using > SDOObjectOutputStream. As it stands the operation will be using the > helpers belonging to the HelperProviderBase instance, rather than that > attached to the output stream. I feel sure this is the issue. Can > you check this over and confirm please and I'll make the update. > > Regards, Kelvin. > > > On 25/09/2007, Ron Gavlin <[EMAIL PROTECTED]> wrote: > > Greetings, > > > > I am passing SDO's as parameters between two tiers of my application using > > RMI-IIOP. The server code is running within an application server that uses > > a complex classloading schema which the Tuscany SDO > > DefaultHelperContextImpl.SDOPackageRegistryDelegator class does not handle > > well. Specifically, I end up 7 classloaders with different registeries in > > different classloaders which totally breaks my application. In pre-SDO 1.0 > > days, I got around this problem by setting the following EMF system > > property > > org.eclipse.emf.ecore.EPackage.Registry.INSTANCE='org.eclipse.emf.ecore.impl.EPackageRegistryImpl'. > > This no longer appears to work. I would like to use > > SDOUtil.createHelperContext() and manage references to the HelperContext > > myself. However, the HelperContext created using this technique does not > > seem to be available/referenced when CORBA attempts to marshall my > > DataGraphs between tiers. > > > > 1. Any thoughts for a quick work-around to get me up and running in short > > order? > > > > 2. Any ideas on a longer-term solution to the problem which doesn't rely on > > single shared registry for the entire VM but might work better than the > > existing classloader-based scheme? > > > > Thanks in advance for your assistance. > > > > - Ron > > > > > > > > --------------------------------------------------------------------- > > To unsubscribe, e-mail: [EMAIL PROTECTED] > > For additional commands, e-mail: [EMAIL PROTECTED] > > > > > --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
