Hi Jeremy,
On 3/6/13 3:31 PM, Jeremy Bauer wrote:
> Hi Marina,
>
> 1. I think any changes to make the behavior consistent wherever
> setRollbackOnly() is called would cause significant compatibility
> problems. For better or worse, the intent of the spec seems to be to
> give users the flexibility to suppress an exception when rolling back,
> so there are likely applications that are taking advantage of it.
> Changing the spec to require the container to ignore system
> exceptions that occur after rollback may also break applications,
> specifically those applications that were coded to look for those
> exceptions.
>
> 2. One complaint we've heard from uses is that, in some cases, they
> don't get an exception when a transaction rolls back.
This is exactly the complain logged in the EJB_SPEC-90. I agree with the
user, that silently rolling back a transaction *is* confusing.
> If a change is made, I think a useful change would to be to always
> throw some exception when a transaction is rolled back, and it would
> need to be optionally enabled to preserve backward compatibility.
I think it's too late to introduce a new option to do it right. Let's
leave it until the next version.
thanks,
-marina
>
> -Jeremy
>
>
>
>
> From: Marina Vatkina <marina.vatkina_at_oracle.com>
> To: jsr345-experts_at_ejb-spec.java.net,
> Date: 03/06/2013 02:46 PM
> Subject: [jsr345-experts] EJB_SPEC-90
> ------------------------------------------------------------------------
>
>
>
> Experts,
>
> I'd like to know your opinion on http://java.net/jira/browse/EJB_SPEC-90.
>
> 1. Will it be a backward incompatible change for the EJB container to
> suppress all system exceptions after the transaction has been marked for
> rollback?
>
> 2. If the answer to #1 is yes, do you think the spec should have more
> details on what happens when an exception is thrown after the
> transaction has been marked for rollback?
>
> thanks,
> -marina
>
>