RFR: Testbug: VerifyJCStressTest leaks memory

Aleksey Shipilev shade at redhat.com
Fri Apr 27 14:07:17 UTC 2018


http://cr.openjdk.java.net/~shade/shenandoah/tests-verify-jcstress-leak/webrev.01/

This is the reason for intermittent timeouts in our testing. Futures retain lots of Test instances,
which retain lots of StampedLock instances, and the whole thing runs very close to max heap. Minute
differences in JDK implementation may prolong the test time significantly in these conditions, and
this is why it fails more frequently in some backports, but not the others.

Testing: hotspot_gc_shenandoah, adhoc test runs

Thanks,
-Aleksey



More information about the shenandoah-dev mailing list