RFR [10]: Allocation path should not touch GC barriers for metadata

[Roman Kennke]

OK. Thanks!

Am 8. September 2018 11:38:25 MESZ schrieb Aleksey Shipilev <shade at redhat.com>:
>http://cr.openjdk.java.net/~shade/shenandoah/10-alloc-barriers/webrev.01/
>
>Found this during backports. The allocation path touches Shenandoah WB,
>which is bad enough in
>itself. It fails with new asserts that verify class fields, because
>allocation path calls those
>barriers on incomplete objects, for example *before* the klass is
>initialized.
>
>sh/jdk11 onward do this correctly with raw accesses. We need to do the
>same in lower JDKs.
>
>Testing: tier3_gc_shenandoah for sh/jdk10 and sh/jdk8u
>
>Thanks,
>-Aleksey