Re: [aspectj-dev] change in runtime execution order

["Wes Isberg" <wes@xxxxxxxxxxxxxx>]

The party line: The rules programmers can rely on are
stated in the programming guide appendix, language
semantics (which are different from what you said).
 Engineering regularities of a particular implementation
are not reliable.  To understand the engineering, you can
read the code in open source.  

The party line protects the ability of other
implementations of AspectJ (e.g., the abc compiler) to use
different advice orderings where precedence is not
determinable.  It also encourages programmers to "state"
any inter-advice dependencies, rather than relying on
implicit orderings.  To avoid confusing users, we're
hesitant to state "rules" that can/will change.

Paternalism aside, er, what was your question again?  

Wes

On Wed,  5 Oct 2005 17:15:06 +0200
 pfefferm@xxxxxxxxxxxxxxxxx wrote:
> Hallo!
> 
> I am a graduate student and I am working on a project to
> determine the actual
> runtime execution order of advice at a given joinpoint.
> Up to now I used AspectJ 1.5.0M2 for my examination. Now
> I updated to AspectJ
> 1.5.0M3 and it seems to me that in this release there are
> major changes in the
> way in which the order of pieces of advice are excuted if
> more than one aspect
> applies to a joinpoint.
> 
> I know that officially precedence between two concrete
> aspects is undefined if
> there is no explicit order defined.
> But in M2, if precedence between two concrete aspects
> could not be determined by
> a 'declare precedence' statement, precedence was resolved
> by the names of
> aspects, e.g. aspects A and B, so all advice in aspect B
> had precedence over
> advice in A.
> This seemed to be a default order if there was no
> explicit order defined by a
> 'declare precedence' statement..
> In M3 this default order for aspects seems to be replaced
> by another method. The
> names of aspects do not play a role in determining
> precedence between them
> anymore. Instead it is the position of  the first piece
> of advice in the source
> file that seems to have influence on aspect precedence.
> 
> Could someone please explain these new rules to me!
> Thank you!
> 
> _______________________________________________
> aspectj-dev mailing list
> aspectj-dev@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/aspectj-dev