> Are the memory leaks completely gone in RC2_UNSTABLE as they > are in RC1_REFACTOR? If so then then that's a big cross > section I won't look at.
I think there is still one leak in the project loading (caused by betwixt?) but is very minor - a few kb per project I think. I can rebuild all of maven-plugins in 16m. > In RC1_REFACTOR I basically eliminated the use of Jelly/Ant > until as late as possible which is in the PluginManager which > did in a big chunk of the memory problem. That's the bit I did - one ant project per project and fewer contexts by loading plugins only once. All of the existing problems in RC2 that cause build failures relate to contexts and jelly. One thing I've currently identified is that context.project != project.context in all cases - I need to review the allocation and find out why a project has more than one context assigned. This applies to plugin projects. I'll spend some time in the next couple of days too - I'll get online before starting anything and see where you are at. Perhaps we can both update the JIRA entry with what we are attacking? I'd be looking at the thing I mentioned above that causes touchstone to fail inside the reactor. Cheers, Brett
