Instead of giving control to the interpreter so that it rethrows the original exception at the original bci, or alternatively, throws the new exception at the exception handler bci, the fix for JDK-8367002 will throw the new exception at the original bci. This might be a reasonable interpretation of the spec, but it could give different results from historical HotSpot behavior. For example, if the test from 8367002 was modified so that the idiv has an exception handler for IllegalAccessError that matches before the exception handler for AgentConfigurationError is tried.