[aspectj-dev] CFlow and Threads - implementation changes... testers requ

[Andrew Clement <CLEMAS@xxxxxxxxxx>]

Hi,

Bug 59909 was raised last week highlighting
a problem where some thread related cflow state was persisting longer than
necessary - it was not a memory leak as such because the state would eventually
be garbage collected but in a system where threads are constantly being
created and destroyed, it could cause you an issue in  terms of heap
consumption.  I've just attached a new version of aspectjrt.jar to
bugzilla bug 59909:

https://bugs.eclipse.org/bugs/show_bug.cgi?id=59909

I don't have that many programs that
use cflow and multiple threads  - can I ask anyone out there who has
code that uses cflow and threads (or maybe even just uses cflow) to try
this new version of aspectjrt?  It is a replacement for the one included
in AspectJ 1.2 rc1.

Basically we now have a dual implementation
of the state management logic.  If you are on a 1.1 JVM then you get
the old behavior (for compatibility) but if you are on a newer JVM then
you will get a new thread local based behavior.

I'm looking for anyone who can test
this on either 1.1 JVMs (to check I haven't broken anything) or 1.2+ JVMs
(to check I haven't broken anything and maybe I've made things better).

To verify that you are actually running
with the new version, you can set a debug property when starting the JVM
(this won't be in the final version) and it will tell you which implementation
of the state mgmt logic it is using:

java -Ddebugcflow=yes MyProgram

please append your findings to either
the bug report or aspectj-dev

many thanks,
Andy.
---
Andy Clement
AspectJ/AJDT Development