convert Failure to exception when using Ask<object>

[ondravondra]

close #7254

[Aaronontheweb]

Thanks for the PR - looks like this broke some tests

[ondravondra]

It seems that the framework relies on this behavior.
For example, test Inbox_Receive_will_timeout_gracefully_if_timeout_is_already_expired uses Akka.Actor.Inbox.Receive that calls Ask<object> and expects the result to be Status.Failure.
A bit of a pickle innit.

[Aaronontheweb]

It seems that the framework relies on this behavior. For example, test Inbox_Receive_will_timeout_gracefully_if_timeout_is_already_expired uses Akka.Actor.Inbox.Receive that calls Ask<object> and expects the result to be Status.Failure. A bit of a pickle innit.

Probably just poor test design - at least that was my gut feeling when I looked at this PR. I'll give it another look when I can

[Arkatufus]

LGTM

[lucavice]

Hi here.

I wanted to report that after updating to 1.5.45+, I noticed some parts of the project I am working on stopped working as intended.
Turns out that some parts of our codebase relied on that behaviour that @ondravondra reported for some of those tests.
In my case, I found:

 var orderQueryResponse = await DomainService<CheckoutDomainService>.Ask<object>(Mediator, getOrderQuery);
 if (orderQueryResponse is Status.Failure sf)
   {
      // does something if you got back a failure

This code was probably bad design in the first place, but I wanted to let you know just in case other people may have used the same pattern and they are experiencing a breaking change after updating. I actually like what the PR is introducing and I think it makes sense.

[Aaronontheweb]

Thanks for weighing in @lucavice - I agree with all of this. Changing the behavior was probably the right call so the PR was a good idea, but yeah it's going to cause breaking behavioral changes for components that relied on the old sub-optimal behavior. Thanks for documenting this in case other users get bit by it too.