HotSpot development trees
Erik Trimble
trims at netdemons.com
Sat Oct 29 00:17:25 PDT 2011
On 10/27/2011 11:58 PM, Krystal Mok wrote:
> Hi all,
>
> HS21 introduced quite a few behavioral changes that OpenJDK 6 users
> might not be expecting. One of them would be the progress of PermGen
> removal, which affects some tuning. It might still be pretty much
> compatible, but no one has been testing for JDK6/HS21 compatibility,
> right?
>
> From what we can see from the outside, recent JDK 6 releases from
> Oracle are staying on HotSpot 20.x, with some bug fixes applied.
>
> My question is:
> Are there any plans from Oracle/OpenJDK to release further updates to
> OpenJDK 6, with the bug fixes applied to HotSpot Express 20's master
> repo? Or, could there be a list of recommended list of bug
> fixes/patches that should be backported to HS20?
>
> We're stuck on JDK 6 now, with no plans to move to JDK 7 in the near
> future, our work would have to base on HS20. So backporting bugfixes
> to HS20 is kind of like "scratching my own itch". If there's anything
> outside contributors can do about this, I'm more than willing to help.
>
> Regards,
> Kris Mok
>
> Software Engineer
> Taobao (http://www.taobao.com)
>
I admit I'm not as intimately aware of what's going on in the JVM as I
once was, but I see no real reason why HS21 or better shouldn't work
under OpenJDK 6, *provided* that during the backport process, certain
features are either permanently disabled, or just turned off by default.
I was under the impression that HS20 was being retained as the Oracle
JDK 6 VM so as to lower the workload for testing - that is, Oracle JDK 6
was being moved into a "maintenance mode" process, and that all that was
being applied to that codebase was fixes, rather than the continued
stream of VM feature updates that had been the norm prior to the
official release of JDK 7.
If someone inside Oracle can comment?
-Erik
> On Fri, Oct 28, 2011 at 11:00 AM, David Holmes
> <david.holmes at oracle.com <mailto:david.holmes at oracle.com>> wrote:
>
> On 28/10/2011 10:37 AM, Erik Trimble wrote:
>
> Back to Andrew's question:
>
> The best HSX candidate repository for OpenJDK6 to use is N-1,
> where N is
> the largest one you see. In theory, that repo should be fully
> stabilized. The hsxN repo is current undergoing stabilization,
> and the
> hsx/hotspot-main is undergoing active development.
>
>
> There is no guarantee that hs21+ will work with OpenJDK6
>
> David
>
>
> -Erik
>
>
> On 10/27/2011 8:21 AM, Paul Hohensee wrote:
>
> Good summary, Dave.
>
> The mainline hotspot development train is tagged as hs23
> because back
> in August
> we changed our process so that the hotspot major version
> number bump
> happens
> when we start development on a new major version rather
> than when we
> end it.
> That's so we can deliver the same version into multiple
> under-development
> JDK versions at the same time. Before August, we bumped
> the major version
> number each time we delivered into a new JDK, which can
> cause a nasty
> jam-up.
>
> hs23 is currently being promoted into both the jdk7u train
> and the
> jdk8 train
> on a (semi-)regular basis.
>
> Paul
>
> On 10/26/11 10:35 PM, David Holmes wrote:
>
> On 27/10/2011 8:49 AM, Dr Andrew John Hughes wrote:
>
> I see three HotSpot trees have appeared at
> hg.openjdk.java.net <http://hg.openjdk.java.net> since
> I last imported one into OpenJDK6:
>
> hsx/hsx21/baseline and hsx/hsx21/master
> hsx/hsx22/hotspot
> hsx/hsx23/hotspot
>
> Which of these can be regarded as stable and can
> be integrated into
> OpenJDK6?
>
>
> I think they are all "stable" (hs23 the least) but
> possibly the
> answer for OpenJDK6 is "none of the above".
>
> hs20.X is used in our currentJDK6 update train (and
> hs19.X for
> earlier updates)
> hs21 was jdk7
> hs22 is 7u2
> hs23 is current version for mainline and for next 7u
> after 7u2 (not
> sure why it's already forked off)
>
> The nomenclature also appears to have changed from
> baseline/master
> trees to just one called 'hotspot'. Why is this?
>
>
> The mainline tree is now hsx/hotspot-main, we then
> fork off hsx/hsxNN
> as needed. This started with hs22 IIRC.
>
> David
>
> Thanks,
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.openjdk.java.net/pipermail/hotspot-dev/attachments/20111029/f7cb7546/attachment-0001.html
More information about the hotspot-dev
mailing list