On Wed, 31 Mar 2021 08:01:27 GMT, Richard Reingruber <rr...@openjdk.org> wrote:
>> I think this relates to why the PEM was moved from the loop-scope to the >> sync op case only. That said it isn't clear why we need the HM or PEM. > > I guess it should be "... must not execute ~PreserveExceptionMark ..." > ~PreserveExceptionMark calls _thread->pending_exception() which is an oop and > that would be illegal as the vm could be at a safepoint when the async > handshake returns. I don't think we need HM or PEM. PEM was added if the handshake would execute code that did throw an exception. Yes, exactly we don't want to be touching "_thread->pending_exception()" during a safepoint. Updating comment. ------------- PR: https://git.openjdk.java.net/jdk/pull/3191