RFR: 8232231: Use last reported wall clock time when emitting JFR per-thread CPU load events

[Robin Westberg]

Hi Misha,

Thanks for reviewing!

Best regards,
Robin

> On 10 Dec 2019, at 22:34, mikhailo.seledtsov at oracle.com wrote:
> 
> Looks good to me,
> 
> Misha
> 
> On 12/10/19 6:43 AM, Robin Westberg wrote:
>> Hi again,
>> 
>> After some offline discussions with Markus and Erik, the plan is to keep the existing product behavior and instead relax the test. As this particular test only exists to ensure that actual thread CPU usage can be calculated from a series of events (which it can’t), the test case should be removed.
>> 
>> New webrev: https://cr.openjdk.java.net/~rwestberg/8232231/webrev.01/ <https://cr.openjdk.java.net/~rwestberg/8232231/webrev.01/>
>> 
>> Best regards,
>> Robin
>> 
>>> On 27 Nov 2019, at 12:10, Markus Gronlund <markus.gronlund at oracle.com> wrote:
>>> 
>>> Hi Robin,
>>> 
>>> Very nice analysis and fix!
>>> 
>>> Looks good.
>>> 
>>> Thank you
>>> Markus
>>> 
>>> -----Original Message-----
>>> From: Robin Westberg
>>> Sent: den 27 november 2019 09:21
>>> To: hotspot-jfr-dev at openjdk.java.net
>>> Subject: RFR: 8232231: Use last reported wall clock time when emitting JFR per-thread CPU load events
>>> 
>>> Hi all,
>>> 
>>> Please review this change that ensures that the last reported wall clock time is used when emitting JFR per-thread CPU load events. Previously, it was possible that per-thread CPU usage could appear to exceed the actual usage if an intermediate event was dropped (see issue for more details).
>>> 
>>> Issue: https://bugs.openjdk.java.net/browse/JDK-8232231
>>> Webrev: https://cr.openjdk.java.net/~rwestberg/8232231/webrev.00/
>>> Testing: tier1-2+5
>>> 
>>> Best regards,
>>> Robin