RFR 8207964: [TESTBUG] Change stressTime to default to 30 for nsk tests

[David Holmes]

On 10/01/2019 7:50 am, David Holmes wrote:
> On 10/01/2019 7:46 am, coleen.phillimore at oracle.com wrote:
>> On 1/9/19 4:28 PM, David Holmes wrote:
>>> Hi Harold,
>>>
>>> cc'd serviceability as a lot of nsk tests are in that area.
>>>
>>> On 10/01/2019 4:56 am, Harold David Seigel wrote:
>>>> Hi,
>>>>
>>>> Please review this fix to change the default stress time for hotspot 
>>>> vmTestbase tests from 60 seconds to 30 seconds.
>>>
>>> Which tests actually use this default value?
>>
>> All tests that don't pass -stressTime <some other number> eg: 
>> vmTestbase/metaspace/stressHierarchy tests.
>> The closed tests also don't pass -stressTime except a couple.
>>
>> Some of the GC unloading tests pass -stressTime 180.
> 
> Let me rephrase my question :) What nsk tests actually use stressTime to 
> control their execution? I can't tell if we affecting 10 tests or 10,000 
> with this change. :)

Poking around a bit this is a hard question to answer. The stressTime is 
used by the Stresser (AFAICS) and I see 48 uses of the Stresser in the 
various test support files, but those files can in turn be used by 
multiple tests.

Anyway this wasn't a blocking query. 30 seconds should be long enough in 
general, and a number of tests explicitly bump the time up. This may 
help reduce overall test execution time.

It will be interesting to see if this impacts timeouts.

Thanks,
David
-----

> 
> Thanks,
> David
> 
>>>
>>>> Open Webrev: 
>>>> http://cr.openjdk.java.net/~hseigel/bug_8207964/webrev/index.html
>>>
>>> The actual change to the default appears correct.
>>>
>>> I just don't know what impact this is going to have on any actual tests.
>>>
>>
>> I ran some of these manually setting -stressTime 30 like 
>> vmTestbase/metaspace/stressDictionary/StressDictionary.java and it ran 
>> enough iterations to do the work needed in the test.  That's why I 
>> suggested making the default lower.
>>
>> This change looks good to me.
>> Thanks!
>> Coleen
>>> Thanks,
>>> David
>>>
>>>> JBS Bug: https://bugs.openjdk.java.net/browse/JDK-8207964
>>>>
>>>> The fix was tested by running Mach5 hotspot tiers 1-5 on Linux-x64, 
>>>> Windows, Solaris, and Mac OS X.
>>>>
>>>> Thanks, Harold
>>>>
>>