Am 27.07.2017 um 08:02 schrieb Andrew Hughes:
On 24 July 2017 at 16:12, Roman Kennke <rkennke@redhat.com> wrote:
This is a fairly huge backport from Shenandoah's dev line (then jdk9) to shenandoah/jdk8u and from there to aarch64-port/jdk8u-shenandoah. It brings:
- bunch of new features (e.g. a separate update-refs phase, parallel and concurrent code cache scanning, etc)
- much improved performance (both macro- and micro-performance)
- much improved stability (we did test this fairly extensively, and added many regression tests)
- much improved usability (e.g. better option handling, much more useful output)
- many bugfixes and tests and code improvements
- lesser diff compared to upstream
All of this has baked considerably in shenandoah/jdk8u for a while, gone through testing and testing and testing (also, testing by interested early adopters) and fixing. I simply transplanted all relevant changesets from shenandoah/jdk8u to aarch64-port/jdk8u-shenandoah with no fuzz or any problems.
It's then been tested again by running the hotspot_gc_shenandoah jtreg test group (i.e. the whole Shenandoah jtreg test suite), again without issues.
A list of all changesets can be found here:
http://cr.openjdk.java.net/~rkennke/aarch64-import-2017-07-24/out.txt <http://cr.openjdk.java.net/%7Erkennke/aarch64-import-2017-07-24/out.txt>
The full webrev:
http://cr.openjdk.java.net/~rkennke/aarch64-import-2017-07-24/webrev.00/ <http://cr.openjdk.java.net/%7Erkennke/aarch64-import-2017-07-24/webrev.00/>
Ok to push?
The update breaks building without precompiled headers. See attached patch.
Thanks for the heads-up and patch. I pushed the fix to all Shenandoah trees and aarch64-port/jdk8u-shenandoah and added a new tag: aarch64-shenandoah-jdk8u141-b16-shenandoah-merge-2017-07-27 I'll also see to it that we (the Shenandoah team) set up a regular build without precompiled headers. This seems to keep biting us... Thanks, Roman