Loom crash

Alan Bateman Alan.Bateman at oracle.com
Mon Jun 8 07:28:05 UTC 2020


On 07/06/2020 21:41, Mark Raynsford wrote:
> Hello!
>
> I've been playing around with Loom and have managed to find a
> reproducible VM crash.
>
> The easiest way is to run src/main/java/com/io7m/loom_20200607/Main.java
> in an IDE and then run the siege [0] tool to trigger a crash:
>
> https://github.com/io7m/loom_20200607
>
> $ siege -c 512 -r 100 http://localhost:9090/tasks
>
> This will cause a lot of logging to the output of the java program,
> followed by a VM crash. I've added the various logs to the repository
> above.
I see you've captured a few error logs where it's a SEGV walking a 
thread stack during GC. I tried it to duplicate it with your test but it 
didn't duplicate for me (I used vegeta to  generate the load but it 
shouldn't make a difference). There is a stability issue with the new 
algorithm for freezing/thawing stacks that may be the issue. Can you run 
with -XX:-UseContinuationChunks and see if it goes away?

-Alan


More information about the loom-dev mailing list