WeakReferenceTest timeout error with -Xint

Derek White derekw at marvell.com
Mon Mar 4 23:05:43 UTC 2019


FYI, I've seen this too. 

Note that the work increased by 100x but the timeout by 10x.

Thanks,
 - Derek

> -----Original Message-----
> From: jcstress-dev <jcstress-dev-bounces at openjdk.java.net> On Behalf Of
> Nick Gasson (Arm Technology China)
> Sent: Tuesday, December 11, 2018 11:20 PM
> To: jcstress-dev at openjdk.java.net
> Cc: nd <nd at arm.com>
> Subject: [EXT] WeakReferenceTest timeout error with -Xint
> 
> External Email
> 
> ----------------------------------------------------------------------
> Hi,
> 
> I've seen a problem recently where WeakReferenceTest can intermittently
> fail with the following error on some slow Arm systems when run with -Xint.
> It's been failing since changeset
> 420:6510f53ebe89 "Further tune-ups for interrupt.WeakReferenceTest for
> concurrent GCs".
> 
>   org.openjdk.jcstress.tests.interrupt.WeakReferenceTest [-Xint]: Observed
> forbidden state: STALE (Thread had failed to respond.)
> 
> I think it's timing out simply because the busy loop in the work() method
> takes too long (several seconds on the system I measured). You can get a
> similar failure on x86 by increasing the number of iterations.
> 
> Is it possible to make a small change to the test to reduce the number of
> iterations to e.g. 10_000_000 if the JVM is running with -Xint? E.g.
> ManagementFactory.getRuntimeMXBean().getInputArguments().contains("-
> Xint")
> or similar. Or simply increase the timeout for this test?
> 
> Thanks,
> Nick


More information about the jcstress-dev mailing list