RFR[13]: 8224531: SEGV while collecting Klass statistics

[Kim Barrett]

> On Jul 3, 2019, at 8:48 AM, Erik Österlund <erik.osterlund at oracle.com> wrote:
>> Did you run the jvmti tests?  There used to be tests that failed if dead objects weren't found, but the tests may have been fixed.
> 
> I will take it for a spin. Note though that behaviour relying on always getting dead objects will fail; the caller of the API will race with concurrent sweeping and either get or not get the dead objects depending on whether the heap iteration happened to kick in before or after sweeping. There is just no way you can rely on that. So if there is a test failure because of that, the test is wrong. Nevertheless, I will try to hunt down such tests.

Good.  If there are still such tests, they need to be fixed.  That shouldn’t hold back this change.