Confirmed on master. Does not have to be two ZeroDivisionError either. We're just not setting cause when raising a new ZeroDivisionError, probably because we don't call through Kernel.raise.

Hey, can I ask how critical this is? The paths along which we construct exceptions to be raised from within JRuby share a lot with the paths for raising an exception from Ruby, so fixing this in the short term will require a fair bit of untangling. We need to fix it obviously, but if it's something we can punt until after RubyConf/9.1.6.0 it would be better.

Pretty minor. It was used in our test suite for sentry-raven, but we can work around it by just constructing Exceptions via raise.

In the meantime, anyone who uses an exception reporting service that reports cause, like ours, will just receive slightly inaccurate reports.