RFR(XS): 8139900: [TESTBUG] Remove G1UpdateBufferSize and InitialBootClassLoaderMetaspaceSize from TestOptionsWithRanges

[sangheon.kim]

On 10/20/2015 10:33 AM, Dmitry Dmitriev wrote:
> Hi Sangheon,
>
> Thank you for the review!
>
> About 'timeout' value and compiler flags: probably this will be not 
> needed, because I am planning to submit review for another RFR which 
> improve processing of numeric flags in HotSpot VM which will also 
> affect TestOptionsWithRanges in a good way, i.e. execution time of 
> this test will be less.
Okay, sounds good!

Thanks,
Sangheon


>
> Thanks,
> Dmitry
>
> On 20.10.2015 20:29, sangheon.kim wrote:
>> Hi Dmitry,
>>
>> I was also wondered about addTypeDependency() but I'm clear now.
>> Changes look good, reviewed.
>>
>> When we merge with 'compiler' part, the 'timeout' value would be 
>> increased again.
>>
>> Thanks,
>> Sangheon
>>
>>
>> On 10/20/2015 10:23 AM, Dmitry Dmitriev wrote:
>>> Gerard, thank you for the review!
>>>
>>> Dmitry
>>>
>>> On 20.10.2015 20:21, gerard ziemski wrote:
>>>>
>>>>
>>>> On 10/20/2015 12:14 PM, Dmitry Dmitriev wrote:
>>>>> Hi Gerard,
>>>>>
>>>>> Thank you for looking into that!
>>>>>
>>>>> On 20.10.2015 19:38, gerard ziemski wrote:
>>>>>> hi Dmitry,
>>>>>>
>>>>>> Thank you very much for looking into this and making the test 
>>>>>> more robust.
>>>>>>
>>>>>> #1 Did you intend to incluse the change in
>>>>>> "test/runtime/CommandLine/OptionsValidation/common/optionsvalidation/JVMOptionsUtils.java" 
>>>>>> removing
>>>>>> addTypeDependency()? Can you please explain that code removal?
>>>>> Yes, sorry that I miss that in an e-mail. This was added for C1&C2 
>>>>> flags. But -Xcomp can dramatically increase execution
>>>>> time and since soon we will get compiler flags in hs-rt repo I 
>>>>> decided to remove that. I will think about more precise
>>>>> approach for that.
>>>>>
>>>>>>
>>>>>> #2 How sure are we that 780 sec is enough if the arm's tests are 
>>>>>> still in progress?
>>>>>
>>>>> I think yes, because "G1UpdateBufferSize" and 
>>>>> "InitialBootClassLoaderMetaspaceSize" are removed and test time
>>>>> significantly goes down. Result for linux-arm-vfp-sflt are looks 
>>>>> good. Several other arms are stuck in submitted state,
>>>>> but I hope they will be ran until tomorrow.
>>>>>
>>>>
>>>> I would be more comfortable with more than a single test run that 
>>>> doesn't time out on ARM, but we can always come back and revisit I 
>>>> suppose - this test is not ON by default, so we can take time to 
>>>> zero in.
>>>>
>>>> Reviewed (with small "r")
>>>>
>>>>
>>>> cheers
>>>>
>>>>
>>>>> Thanks,
>>>>> Dmitry
>>>>>>
>>>>>>
>>>>>> cheers
>>>>>>
>>>>>> On 10/20/2015 11:06 AM, Dmitry Dmitriev wrote:
>>>>>>> Hello,
>>>>>>>
>>>>>>> Please review fix for TestOptionsWithRanges.java test. In this 
>>>>>>> fix I remove two flags("G1UpdateBufferSize",
>>>>>>> "InitialBootClassLoaderMetaspaceSize") from testing because 
>>>>>>> their maximum value would consume too much memory and take a
>>>>>>> lot of time. Also I increase timeout value for the test, since 
>>>>>>> number of flags with ranges were increased recently.
>>>>>>>
>>>>>>> JBS: https://bugs.openjdk.java.net/browse/JDK-8139900
>>>>>>> webrev.00: http://cr.openjdk.java.net/~ddmitriev/8139900/webrev.00/
>>>>>>> <http://cr.openjdk.java.net/%7Eddmitriev/8139900/webrev.00/>
>>>>>>> Testing: ran on all platforms including arm's(test for few arm's 
>>>>>>> are still in progress).
>>>>>>>
>>>>>>> Thanks,
>>>>>>> Dmitry
>>>>>>>
>>>>>
>>>>>
>>>
>>
>