RFR: JDK-8277092: TestMetaspaceAllocationMT2.java#ndebug-default fails with "RuntimeException: Committed seems high: NNNN expected at most MMMM"

[Thomas Stuefe]

On Thu, 18 Nov 2021 10:50:14 GMT, Thomas Stuefe <stuefe at openjdk.org> wrote:

> Test error after [JDK-8276731](https://bugs.openjdk.java.net/browse/JDK-8276731). That patch changed the way metaspace chunks are uncommitted: Before, chunks were uncommitted when being returned to the system; now they are uncommitted only if `Metaspace::purge()` is explicitly called. 
> 
> In real VM life, that makes no difference since these two things happen back to back in `CLDG::purge()`. 
> 
> But these metaspace tests isolate metaspace arenas and test them as individual units, without involving GC or CLDG, and therefore we need to purge manually. Otherwise, the commit numbers are unpredictable and may be larger than expected.
> 
> Tests: manual x64+x86, GHAs (one unrelated error in hs compiler tests on linux x64), SAP nightlies

GHA display is off; the only real test error is linux x64 hotspot compiler, but that has nothing to do with this change.

-------------

PR: https://git.openjdk.java.net/jdk/pull/6451