RFR: 8143847: Remove REF_CLEANER reference category
kim.barrett at oracle.com
Mon Jan 11 05:51:49 UTC 2016
On Jan 8, 2016, at 3:04 PM, Jesper Wilhelmsson <jesper.wilhelmsson at oracle.com> wrote:
> Looks good!
> Since you change vmStructs.cpp it's a good idea to run some SA test just to make sure there is no use of the deleted field in there. There's a bunch of jtreg tests in hotspot/test/serviceability/sa/. Last time I checked these were not run by jprt.
Oh, you are right, jprt doesn't run the SA tests. I ran all the
servicability tests locally with no problems.
> Slightly unrelated: Is it worth to make number_of_subclasses_of_ref() slightly more robust by adding a "REF_LAST" or similar to use in there? I guess we don't intend to add more reference types, but it is easy to forget to update this method if a new type is added last in the enum.
I was going to say "sure", but then looked at the code more carefully,
and think I'd rather not, at least not as part of this change.
number_of_subclasses_of_ref not only knows about REF_PHANTOM (was
REF_CLEANER) being the last, but it also knows about REF_OTHER. I'm
inclined to think REF_OTHER is a mistake, and should be removed. I'd
be happy to add a "last reference type" constant as part of such a
change. But my preference would be to not add any of that to the
current change, which is about backing out REF_CLEANER.
More information about the hotspot-gc-dev