[prev in list] [next in list] [prev in thread] [next in thread] 

List:       openjdk-serviceability-dev
Subject:    Re: RFR 8064441: java/lang/management/ThreadMXBean/Locks.java fails intermittently, blocked on wrong
From:       Jaroslav Bachorik <jaroslav.bachorik () oracle ! com>
Date:       2014-11-27 8:33:26
Message-ID: 5476E1D6.8010601 () oracle ! com
[Download RAW message or body]

On 11/27/2014 01:43 AM, David Holmes wrote:
> On 26/11/2014 10:57 PM, Jaroslav Bachorik wrote:
>> Please, review the following test change
>>
>> Issue : https://bugs.openjdk.java.net/browse/JDK-8064441
>> Webrev: http://cr.openjdk.java.net/~jbachorik/8064441/webrev.00
>>
>> The test fails because of System.out.println() may induce unexpected
>> locking. The solution is to use an already existing LockFreeLogManager
>> library class to collect the logs in the lock-free manner and print the
>> logs only after all the test code has been run.
>
> I seem to remember spending a lot of time on this test in the past,
> where the phaser usage was intended to ensure that we could not be seen
> to be blocking on the internal locks associated with the println. Where
> has this gone wrong?

The remaining problem is the usage of System.out.print[ln]() inside the 
test threads. System.out.print[ln]() might involve blocking the calling 
thread and eg. create false positives when the test is waiting for a 
thread to enter the blocked state.

>
> What are the semantics of the lockfreelogger? When will we see its output?

It is using thread local lists of the logging records. Each logging 
record has an id which is unique and monotonous over all the threads -> 
an AtomicInteger instance is used to achieve this.

You can retrieve the log output (ordered by the log record id) by 
calling LockFreeLogManager.toString().

Review for the LockFreeLogManager is available here 
[http://mail.openjdk.java.net/pipermail/serviceability-dev/2014-November/015848.html]

-JB-

>
> Thanks,
> David
>
>> Thanks,
>>
>> -JB-

[prev in list] [next in list] [prev in thread] [next in thread]