dev@glassfish.java.net

Re: sth doubt about the postInvoke of EJB Container

From: Marina Vatkina <marina.vatkina_at_oracle.com>
Date: Tue, 12 Mar 2013 11:25:37 -0700

Hi WuJie,

Did you try to make this change? Do all ejb devtests pass after that?
Look at the 'svn ann' output and then at the svn log for the revisions
that correspond to the changed lines - it will give you some idea on
what the change tried to accomplish.

Best,
-marina

On 3/11/13 9:24 PM, Wu Jie wrote:
> Hi Marina
> CC)Dev team
>
> I found that there is some change about the processing sequence of
> postInvoke of EJB Container between v2.x and v3.x
>
> the processing sequence in v2.x as following.
> ===
> Tx's postInvoke
> ↓
> InvocationManager's postInvoke
> ===
>
> but in v3.x the processing sequence is reversed as following.
> ===
> InvocationManager's postInvoke
> ↓
> Tx's postInvoke
> ===
>
> there is some special consideration about the change?
> #seems that the simple optimization(refactor) for redundant code
>
> I think that v2's processing sequence is more reasonable rather than
> v3 considering the ejb(CMT)'s invocation(the flow figure is attached).
>
> Thanks
> wujie