[aarch64-port-dev ] Fwd: Compact strings enablement - JTreg runs

[Stuart Monteith]

So, I got the machines back. I ran JCStress - with a VM error with
compact strings off and off.

With compact strings off, I get one error:
   http://people.linaro.org/~stuart.monteith/CompactStrings/JCStress/org.openjdk.jcstress.tests.accessAtomic.varHandles.fields.opaque.DoubleTest.html

With compact strings on, I get two:
   http://people.linaro.org/~stuart.monteith/CompactStrings/JCStress/org.openjdk.jcstress.tests.atomicity.buffers.DirectByteBufferViewsAtomicityTests.DoubleViewTest.html
   http://people.linaro.org/~stuart.monteith/CompactStrings/JCStress/org.openjdk.jcstress.tests.locks.stamped.StampedLockPairwiseTests.tRLt_tUR.aRWLw_U.html

These errors of the form
   Warning: 'NoSuchMethodError' on register of sun.hotspot.WhiteBox:......
         :              :            :

#
#  Internal Error
(/home/stuart/repos/jdk9-compact/hotspot/src/share/vm/gc/g1/g1YoungRemSetSamplingThread.cpp:103),
pid=3090, tid=3180
#  assert(hr == g1cs->inc_tail() || hr->next_in_collection_set() !=
__null) failed: next should only be null at tail of icset

for the first two reports. The second has the following error:

#  Internal Error
(/home/stuart/repos/jdk9-compact/hotspot/src/share/vm/gc/g1/heapRegion.inline.hpp:234),
pid=39487, tid=39565
#  assert(in_collection_set()) failed: should only invoke on member of CS.

I ran it again - it failed on the 5622nd iteration:
   http://people.linaro.org/~stuart.monteith/CompactStrings/JCStress/hs_err_pid17549.log

Repeat attempts failed - it seems I was "lucky".

Have G1GC problems been seen before on aarch64 recently? As this
happens with and without CompactStrings, it looks like it is an
unrelated problem.

Would it be possible to raise a bug? I don't see any duplicate issues
in bugs.openjdk.java.net

I don't see any changes since adc8c84b7cf8 "8160565: AIX port: cleanup
of libo4 wrapper stub" on my build that would resolve the issue.


Regards,
   Stuart

On 8 July 2016 at 17:36, Andrew Haley <aph at redhat.com> wrote:
> On 07/07/16 16:58, Stuart Monteith wrote:
>> The JSR166TestCase.java fails with:
>> JavaTest Message: JUnit Failure: testInvokeAll1(ForkJoinTaskTest):
>> ExecutorService java.util.concurrent.ForkJoinPool at 234bf9ce[Shutting
>> down, parallelism = 48, size = 20, active = 1, running = 0, steals =
>> 23, tasks = 0, submissions = 0] did not terminate in a timely manner
>>
>> On another run JSR166 simply times out completely, and in another is
>> passed fine. An intermittent failure of a concurrency test - worrying?
>
> Yes.  Have you tried jcstress with the same build?
>
>> The failures I find are not /obviously/ affected by CompactStrings.
>
> I agree.  The test failures I've seen are not anything to do with
> CompactStrings IMO.
>
> These are the failures I'm seeing on AArch64 (vs x86):
>
> FAILED: compiler/cpuflags/TestAESIntrinsicsOnUnsupportedConfig.java
> FAILED: compiler/interpreter/DisableOSRTest.java
> FAILED: compiler/jvmci/compilerToVM/MaterializeVirtualObjectTest.java
> FAILED: compiler/uncommontrap/TestDeoptOOM.java
> FAILED: gc/g1/TestLargePageUseForAuxMemory.java
> FAILED: runtime/CommandLine/OptionsValidation/TestOptionsWithRanges.java
> FAILED: runtime/NMT/PrintNMTStatistics.java
>
> I just fixed TestDeoptOOM.
>
> Andrew.
>