RFR (trivial) 8214356: Verification of class metadata unloading takes a long time

[coleen.phillimore at oracle.com]

On 11/28/18 5:28 PM, Aleksey Shipilev wrote:
> Forgot to mention: this does indeed look trivial to me.

Thanks!  The one line assert is nicer.

http://cr.openjdk.java.net/~coleenp/8214356.02/webrev/src/hotspot/share/memory/binaryTreeDictionary.inline.hpp.udiff.html

Coleen
>
> -Aleksey
>
> On 11/28/18 11:24 PM, coleen.phillimore at oracle.com wrote:
>> Thanks for the quick review Aleksey. I'll make the suggested change.  It's a lot better.
>> Coleen
>>
>> On 11/28/18 5:22 PM, Aleksey Shipilev wrote:
>>> On 11/28/18 11:11 PM, coleen.phillimore at oracle.com wrote:
>>>> Summary: conditionalize verify_chunk_in_freelist call.
>>>>
>>>> Tested with stress test that exercises this code:
>>>> test/hotspot/jtreg/vmTestbase/metaspace/stressDictionary
>>>>
>>>> open webrev at http://cr.openjdk.java.net/~coleenp/8214356.01/webrev
>>> Looks okay.
>>>
>>> Flag description suggests this is a gated option for stress tests. Which means we can just say:
>>>
>>>     assert(!FLSVerifyDictionary || !this->verify_chunk_in_free_list(chunk), "Double entry");
>>>
>>> ...like we already do in the same file:
>>>
>>>     assert(!FLSVerifyDictionary || total_size_in_tree(root()) == total_size(), "_total_size
>>> inconsistency");
>>>
>>> Thanks,
>>> -Aleksey
>>>
>