RFR: 8161208: Unable to run jtreg tests with MinimalVM

[Michail Chernov]

Hi Dima,

I took into account your comments.

http://cr.openjdk.java.net/~mchernov/8161208/webrev.03/

JFR is enabled only with FlightRecorder or StartFlightRecording options. 
Please look at closed src (jfrActivator.cpp - JfrActivator::initialize()).

I would keep checking of UnlockCommercialFeatures because it makes this 
checking more clear and this is similar to code from closed part.


Thanks,
Michail

On 07/20/2016 11:12 AM, Dmitry Fazunenko wrote:
> Michail,
>
> The new version looks much better, thanks for finding existing WB API 
> to access VM flags.
> So, the fix looks good, but some rather cosmetic comments:
>
> 1) 45 private WhiteBox whiteBox = WhiteBox.getWhiteBox(); --> 45 
> private *static final* WhiteBox WB = WhiteBox.getWhiteBox();
>
> 2) startFR is not boolean: 137 String isStartFR = --> 137 String startFR =
>
> 3) For the sake of code simplicity you can skip check for 
> UnlockedCommercialFatures But it's up to you. 4) another way to turn 
> JFR on: |-XX:FlightRecorderOptions| (https://docs.oracle.com/javacomponents/jmc-5-5/jfr-runtime-guide/comline.htm#JFRRT184)
> I'm not sure if this option has an effect standalone, but it's better to check
> for its setting as well.
>
> Thanks,
> Dima
>
>
> On 19.07.2016 16:35, Michail Chernov wrote:
>> Hi David,
>> Thanks for your answer. I revised code that checks FlightRecorder 
>> status. At my point of view better way is to use WhiteBox to get VM 
>> flag value. This allows not to use additional dependencies for VMProps.
>>
>> Updated webrev:
>> http://cr.openjdk.java.net/~mchernov/8161208/webrev.01/
>>
>> Tested locally on Open JDK build, and Oracle JDK (Linux x64 , Linux 
>> x86 - server and minimal VM).
>>
>> Thanks,
>> Michail
>>
>> On 15.07.2016 08:31, David Holmes wrote:
>>> Hi Michail,
>>>
>>> On 15/07/2016 2:21 AM, Michail Chernov wrote:
>>>> Hi,
>>>>
>>>> Could I have a reviews for this change, please?
>>>>
>>>> https://bugs.openjdk.java.net/browse/JDK-8161208
>>>> http://cr.openjdk.java.net/~mchernov/8161208/webrev.00/
>>>> http://cr.openjdk.java.net/~mchernov/8161208/webrev.hotspot.00/
>>>>
>>>> This issue is started since this fix:
>>>> http://mail.openjdk.java.net/pipermail/hotspot-gc-dev/2016-June/018319.html 
>>>>
>>>
>>> I think I would have objected to the form of that fix - VMProps was 
>>> only looking at information available via system properties.
>>>
>>>> This change fixes two issues:
>>>> 1. Avoiding of java.management usage with minimal VM (JTreg was 
>>>> started
>>>> with '-vmoptions:"-minimal"' and worked as expected).
>>>
>>> Okay.
>>>
>>>> 2. Adding appropriate modules for proper VMProps execution on non-full
>>>> JDK images (JTreg was tested with '-vmoptions:"-limitmods java.base"',
>>>> java.compact1, java.compact2 and so on - JTreg is started and works as
>>>> expected).
>>>
>>> Was that testing done with a full VM or the Minimal VM? On a full VM 
>>> limited to compact1, compact2 or java.base, java.lang.management 
>>> will not be present, but we will execute the query code in VMProps 
>>> as it isn't the minimal VM.
>>>
>>> Thanks,
>>> David
>>>
>>>> Tested locally and via RBT/JPRT.
>>>>
>>>> Thanks
>>>> Michail
>>
>