Unsafe: removing the monitorEnter/monitorExit/tryMonitorEnter methods

Peter Levart peter.levart at gmail.com
Thu Feb 27 07:29:01 UTC 2014


On 02/26/2014 09:54 PM, Martin Buchholz wrote:
> I don't recall having added this particular wart
> to test/java/lang/ProcessBuilder/Basic.java, and history suggests that
> others did that.
>
> It does seem that being able to tell whether a java object monitor is
> currently locked is useful for debugging and monitoring - there should be a
> way to do that.

Thread.holdsLock(Object) ?

Regards, Peter

>
>
> On Wed, Feb 26, 2014 at 7:12 AM, Paul Sandoz <paul.sandoz at oracle.com> wrote:
>
>> Hi,
>>
>> Out of all the methods on Unsafe i think the
>> monitorEnter/monitorExit/tryMonitorEnter are the least used and are very
>> strong candidates for removal.
>>
>> 99% of use-cases are supported by classes in the
>> java.util.concurrent.locks package.
>>
>>
>> Within the JDK itself it is only used in one JDK test file
>> test/java/lang/ProcessBuilder/Basic.java:
>>
>>                      while (unsafe.tryMonitorEnter(s)) {
>>                          unsafe.monitorExit(s);
>>                          Thread.sleep(1);
>>                      }
>>
>> for a test verifying an EOF is received on pending reads and it is polling
>> to check when the process builder acquires the lock before destroying the
>> process, presumably to avoid some sort of race condition that occasionally
>> causes the test to fail.
>>
>> I believe this test as been through a number of rounds, i stared at things
>> for a bit, but cannot quickly work out a replacement; i lack the knowledge
>> on this area.
>>
>>
>> Outside of the JDK i can only find one usage of monitorExit/Enter
>> (according to grep code) in JBoss modules, and i believe this is only used
>> on Java 1.6 to work around some limitations in class loading that were
>> fixed in 1.7.
>>
>>
>> Given such very limited use i propose to remove these methods after having
>> worked out a fix for ProcessBuilder/Basic.java test.
>>
>> Paul.
>>




More information about the core-libs-dev mailing list