Csla.DataPortalException.BusinessException in 5.3.0

[ajohnstone-ks]

We have some tests which are failing now, and I tracked it down to the fact that Csla.DataPortalException.BusinessException is returning a Csla.Reflection.CallMethodException, which has its inner exception as the one actually thrown from within the Child_Update exception.

Is this change intentional? Previously it was the exception directly throw from within the Child_Update method.

[rockfordlhotka]

Interesting. Not intentional.

The way the data portal invokes methods is entirely different in version 5, because we added the ability to inject parameter values and support for the data portal operation attributes (Create, Fetch, Update, etc.).

The BusinessException method tries to strip out DataPortalException and CallMethodException exceptions to find the underlying exception. If you look at the code it is pretty simplistic, and perhaps there's some intermediate exception in the way?

Can you provide the entire stack trace (exception tree) from your test?

[ajohnstone-ks]

It almost looks like an off by 1 error. Is this what you're looking for? The exception highlighted in blue is our actual exception which were were expecting in BusinessException.

[rockfordlhotka]

This is what I was looking for, thank you.

The implementation of BusinessException has never been robust, and that is causing the issue.

Though I am also wondering why there are 3 CallMethodException instances in the stack, that seems odd to me. It makes me wonder if there's something not working right in the way the methods are being invoked, or the way exceptions are being thrown as part of the invocation process.

[rockfordlhotka]

I've recorded this as a bug: #1764

[ajohnstone-ks]

I doubt it matters, but this is via the WCF data portal.

I think I've seen multiple nested exceptions like this before, I think because the object throwing the exception is actually a grand-child object, so it seems like each level is adding its own CallMethodException.

[rockfordlhotka]

@ajohnstone-ks does this look like a workable solution to you?

c1e744d