RFR: 8209345: Merge SATBMarkQueueFilter into SATBMarkQueueSet

[Thomas Schatzl]

Hi,

On Sun, 2018-08-12 at 22:08 -0400, Kim Barrett wrote:
> Please review this change that eliminates SATBMarkQueueFilter in
> favor
> of an extension to SATBMarkQueueSet.  The QSet's _filter member and
> filter function that invokes it are replaced by a virtual function on
> the QSet.  G1SATBMarkQueueSet provides an implementation copied from
> the former G1SATBMarkQueueFilter.
> 
> Changed SATBMarkQueueSet's constructor and destructor from public to
> protected, since it's a base class.  Also made initialize protected,
> as it should only be called from derived classs.
> 
> This changeset is based on the in-review change for JDK-8209396.
> http://mail.openjdk.java.net/pipermail/hotspot-gc-dev/2018-August/022
> 868.html
> 
> CR:
> https://bugs.openjdk.java.net/browse/JDK-8209345
> 
> Webrev:
> http://cr.openjdk.java.net/~kbarrett/8209345/open.00/
> 
> Testing:
> Mach5 tier1-3, hs-tier4-5, in conjunction with JDK-8209346 and JDK-
> 8209347.
> Locally tier1 and hotspot_gc without those other changes.
> 

  - I think the SATBMarkQueueSet's destructor should be virtual, given
that there will be a second implementor with unknown functionality (and
in general I prefer to have virtual destructors as soon there is any
virtual method). Also the SATBBufferClosure's one.

Looks good otherwise.

Thanks,
  Thomas