SIGSEGV in Method::set_on_stack while running JMH benchmark

[Aleksey Shipilev]

On 1/21/21 4:22 PM, Eirik Bjørsnøs wrote:
> Debugging this is above my pay grade, but I found this in
> Method::set_on_stack:
> 
> 2301: constants()->set_on_stack(value);
>
> Could it be that constants() returns something unexpected?

si_addr = 0x0 kinda hints that we tried to dereference NULL.

> Another observation is that class redefinition seems to be involved,.
> 
> This is no showstopper for me (In benchmark, only happened once), but
> perhaps a Hotspot wiz wants to take a look. I can send the entire hs_err
> file on request.

Might be one of these things:
   https://bugs.openjdk.java.net/browse/JDK-8226690
   https://bugs.openjdk.java.net/browse/JDK-8222004

Run with fastdebug JDK build, if you can.

-- 
Thanks,
-Aleksey