[
https://issues.apache.org/jira/browse/CASSANDRA-5691?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13692707#comment-13692707
]
Mikhail Mazursky commented on CASSANDRA-5691:
---------------------------------------------
I see following reasons to make this change:
1. It removes code that have a bug;
2. It removes code that is not tested;
3. Why invent a wheel? CountDownLatch(1) is a common way to do this kind of
signaling and is widely used and known. So it's easier to understand code.
If the patch is not applied then at least singnal() invocations should be
replaced with signalAll() and signal() should throw
UnsupportedOperationException.
p.s. I don't think CDL offers any better or worse performance.
> Remove SimpleCondition
> ----------------------
>
> Key: CASSANDRA-5691
> URL: https://issues.apache.org/jira/browse/CASSANDRA-5691
> Project: Cassandra
> Issue Type: Bug
> Components: Core
> Reporter: Mikhail Mazursky
> Priority: Minor
> Attachments: trunk-5691.patch
>
>
> Problematic scenario:
> 1. two threads get blocked in SimpleCondition.await();
> 2. some thread calls SimpleCondition.signal();
> 3. one of blocked threads wakes up and runs;
> 4. spurious wakeup happens in the second thread and it wakes up too and runs
> even though nobody signaled it.
> Thus this is a broken implementation of Condition interface.
> Anyway, looking at how code uses it, SimpleCondition can just be replaced
> with CountDownLatch.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
