request for review (M): 4965777 GC changes to support use of discovered field for pending references
y.s.ramakrishna at oracle.com
Wed Sep 7 09:53:18 PDT 2011
On 9/7/2011 9:32 AM, Alan Bateman wrote:
> Mandy Chung wrote:
>> It'd be good to write a regression test for this bug to be included
>> in the jdk fix. I can help writing it if you want.
> I looked at the changes in the jdk repository and they look good to me
Thanks for the review, Alan.
> Adding the bit to the jdk_version_info structure is a neat way to deal
> with the mis-match until hotspot and the JDK meet up in a promoted
> build. Do you think this bit will be temporary or do you plan to
> remove it once the changes have got into all releases?
Yes, it was Mandy's idea (thanks!) to use that structure. It's temporary
only in a long-term sense. It'll be there for at least the next couple of
HSX releases, and possibly beyond. There was no immediate plan for
removing it, although as you say once all the JDK's have the
changes, the JVM need not test for it and the bit can perhaps be
reclaime, and the adaptivity of the JVM along that dimension dropped
entirely. The idea, though, is also to allow the newer JVM's to be able
to interoperate for a while yet (may be a year or more) with older JDK's,
which assists with traiging and debugging of issues that involve both
JDK and JVM (sometimes performance issues) -- a capability that
some folks felt was definitely worthwhile.
> On the test, I agree with Mandy that it would be good to have a test
> in the jdk repository. I assume this will need to wait until HotSpot
> is in a promoted build so it would need to be a follow-on changeset
I'll work with Mandy on the test, and on the integration into the JDK
repos. Yes, the JDK changes will need to wait until the HotSpot
changes have been promoted and are used by JPRT.
More information about the hotspot-gc-dev