On Wed, 27 Sep 2023 00:44:30 GMT, Chris Plummer <cjplum...@openjdk.org> wrote:

>> I made a pass over the deadlock detection code involved. Nothing stuck out. 
>> There's not much in the way of SA-isms in this code. I think anyone with an 
>> understanding of how hotspot locking works should be able to navigate it 
>> pretty easily and maybe conjure a hypothesis as to why this happens as 
>> frequently as it does.
>
> I wonder if we are in the middle of a GC, the object has been moved, and the 
> lock stack object reference has been updated but the monitor reference has 
> not (or vice versa).

Possible I guess. I added some instrumentation to try and see what threads we 
have and what their lockstacks show when the failure happens. Of course the 
test didn't fail after I did that, so having to re-run.

-------------

PR Review Comment: https://git.openjdk.org/jdk/pull/15907#discussion_r1337987283

Reply via email to