[abc] RE:

> I meant a pointcut such that the "before" advice runs when you enter a
> locked region, and the "after" advice runs when you exit. Then cflow
of
> this pointcut would be exactly what you're asking, and you could use
> the cflow analysis to do it statically.

I see.
 
> However, such a pointcut probably cannot even be implemented because
in
> bytecode, synchronized regions need not be properly nested (they must
> be nested in source code). I remember Jerome having lots of grief due
> to this fact.

Yes, that's exactly the problem.
 
> If they're not nested, then it's equivalent to static analysis of a
> tracematch.

true, actually. hmm, interesting. So maybe we should think about such a
pointcut then.

Eric
Received on Tue Oct 10 20:11:15 2006