| Steve Ebersole, it might be best to keep this for 5.2.1 or 5.3.0, giving Scott Marlow time to test Wildfly and me to verify our various OSGi demo projects are improved. Tim Ward, for now, we ended up keeping your fix as is. I investigated using the ClassLoaderService and "set once" approach, but ran into issues due to the static nature of Javassist. Instead, we'll use your on-demand approach until we're able to enhance Javassist itself or switch to something else. Additionally, I fixed the OSGi test (working well, thanks – was just missing a few packages) and am working through failures in hibernate-core. You may have exposed a few broader Javassist-related issues...way to go  |