RFR(L): 8225797: OldObjectSample event creates unexpected amount of checkpoint data

[Erik Gahlin]

Looks good.

Erik
> Friendly reminder :)
>
> Thank you
> Markus
>
> -----Original Message-----
> From: Markus Gronlund
> Sent: den 4 september 2019 15:07
> To: hotspot-jfr-dev at openjdk.java.net
> Subject: RFR(L): 8225797: OldObjectSample event creates unexpected amount of checkpoint data
>
> Greetings,
>
> Please review the following change set:
>
> Bug: https://bugs.openjdk.java.net/browse/JDK-8225797
> Webrev: http://cr.openjdk.java.net/~mgronlun/8225797/webrev03/
> Testing: jdk_jfr
>
> Summary:
> Old object sampling will currently do special tagging of artifacts (klasses, methods etc) at the sample site but it does not include another subset selection pass to reduce the size of the sample set surviving the chunk / epoch.
> For heavily used systems, this can result in a very large set of constants being carried over into subsequent chunks (the constant set being associated with surviving sample candidates).
> We change this to only do "normal" tagging at the sample site and introduce a subset selection pass (hooking into the rotation logic) to reduce the amount of constants to be collected, now only for survivors.
> Some auxiliary changes are also included in order to better accommodate a simplified tagging scheme.
>
> Thanks
> Markus