RFR(T): 8234272 ProblemList runtime/NMT/HugeArenaTracking.java

[Leonid Mesnik]

Hi

New patch works fine.

Leonid

On 11/17/19 7:31 AM, Zhengyu Gu wrote:
> Hi,
>
> Could you test this patch? 
> http://cr.openjdk.java.net/~zgu/JDK-8234270/webrev.00/index.html
>
>
> I was able to reproduce original bug with 
> compiler/codecache/stress/UnexpectedDeoptimizationTest.java + NMT, 
> newly added assertion caught the problem.
>
> I also fixed the new test failure on Windows. JDK-8204128 patch missed 
> one long -> ssize_t change. long on Windows is 4-bytes vs 8-bytes on 
> other 64-bits platforms, that explains why it only fails on Windows.
>
> This patch passed submit test.
>
> [Mach5] mach5-one-zgu-JDK-8234270-1-20191117-1425-6784823: PASSED
>
> Thanks,
>
> -Zhengyu
>
>
> On 11/16/19 1:42 PM, Leonid Mesnik wrote:
>> Hi
>>
>> Unfortunately, I don't know any publicly available test which could 
>> reproduce this issue right now.
>>
>> We also run some jdk/hotstpot regression tests with NMT enabled, 
>> however not sure if they could be used as reproducers.
>>
>> Leonid
>>
>> On 11/16/19 6:04 AM, Zhengyu Gu wrote:
>>> Hi Leonid,
>>>
>>> On 11/15/19 6:43 PM, Leonid Mesnik wrote:
>>>> It would be better to backout fix.
>>>>
>>>> Other tests executed with NMT triggered this assertion also. So we 
>>>> are going to have a lot of assertions.
>>>
>>> Are any of these "other tests" publicly available? If so, could you 
>>> point me what are they?
>>>
>>> Thanks,
>>>
>>> -Zhengyu
>>>
>>>>
>>>> Leonid
>>>>
>>>> On 11/15/19 3:26 PM, Daniel D. Daugherty wrote:
>>>>> Greetings,
>>>>>
>>>>> runtime/NMT/HugeArenaTracking.java is a new test added by the 
>>>>> following fix:
>>>>>
>>>>>     JDK-8204128 NMT might report incorrect numbers for Compiler area
>>>>>     https://bugs.openjdk.java.net/browse/JDK-8204128
>>>>>
>>>>> The test is failing in the JDK-14 CI on the Win* platforms. That 
>>>>> failure is
>>>>> tracked by:
>>>>>
>>>>>     JDK-8234270 HugeArenaTracking.java failed assert(_size >= sz) 
>>>>> failed: deallocation > allocated
>>>>>     https://bugs.openjdk.java.net/browse/JDK-8234270
>>>>>
>>>>> To keep the noise in the CI down over the weekend, I'm putting the 
>>>>> test
>>>>> on the ProblemList for Win* using this bug:
>>>>>
>>>>>     JDK-8234272 ProblemList runtime/NMT/HugeArenaTracking.java
>>>>>     https://bugs.openjdk.java.net/browse/JDK-8234272
>>>>>
>>>>> Here's the trivial diff:
>>>>>
>>>>> $ hg diff
>>>>> diff -r 8e7f29b1ad4a test/hotspot/jtreg/ProblemList.txt
>>>>> --- a/test/hotspot/jtreg/ProblemList.txt    Fri Nov 15 14:22:24 
>>>>> 2019 -0800
>>>>> +++ b/test/hotspot/jtreg/ProblemList.txt    Fri Nov 15 18:22:00 
>>>>> 2019 -0500
>>>>> @@ -90,6 +90,7 @@
>>>>>  # :hotspot_runtime
>>>>>
>>>>>  runtime/jni/terminatedThread/TestTerminatedThread.java 8219652 
>>>>> aix-ppc64
>>>>> +runtime/nmt/HugeArenaTracking.java 8234270 windows-x64
>>>>>  runtime/ReservedStack/ReservedStackTest.java 8231031 generic-all
>>>>>
>>>>>  ############################################################################# 
>>>>>
>>>>>
>>>>>
>>>>> Thanks, in advance, for any comments, questions or suggestions.
>>>>>
>>>>> Dan
>>>>>
>>>>
>>>
>>
>