[jira] Commented: (AMQ-1235) Scheduler.cancel uses incorrect argument to shutdown threads

["Alex Burgel (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/activemq/browse/AMQ-1235?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#action_39041 ] 

Alex Burgel commented on AMQ-1235:
----------------------------------

a quick clarification....

you can't call ScheduledThreadPoolExecutor.remove with a ScheduledFuture, because it expects a Runnable.

two alternatives are

1) don't call ScheduledThreadPoolExecutor.remove at all... which is equivalent to the current behavior, tho this wil leave the executor with lots of cancelled tasks (see AMQ-1205)

2) cast ticket to RunnableScheduledFuture, and call remove on that. this will work because in this case ticket is a RunnableScheduledFuture, tho its probably worthwhile adding an instanceof check just to be sure.

> Scheduler.cancel uses incorrect argument to shutdown threads
> ------------------------------------------------------------
>
>                 Key: AMQ-1235
>                 URL: https://issues.apache.org/activemq/browse/AMQ-1235
>             Project: ActiveMQ
>          Issue Type: Bug
>    Affects Versions: 4.1.1
>            Reporter: Alex Burgel
>            Priority: Critical
>
> looking at the code from 4.1.1 in org.apache.activemq.thread.Scheduler, in the cancel method:
> the Runnable task argument is passed to clockDaemon.remove(). i think this is incorrect. ScheduledFuture ticket should be passed to clockDaemon.remove().
> the javadocs of ScheduledThreadPoolExecutor.remove discuss the possibility that Runnables might be stored in some other form internally, so calling remove with a plain Runnable might not do anything. I think the solution is to call remove with a ScheduledFuture, which is how they are stored internally in ScheduledThreadPoolExecutor.
> i came across this bug after upgrading to the java 5 version of backport-util-concurrent 3.0. that version makes more assumptions about the types that are passed into ScheduledThreadPoolExecutor.remove, so when you pass in a regular Runnable you'll get a ClassCastException.
> this is trivial to fix, so i don't think a patch is necessary. also i think this might address the memory leak mentioned in AMQ-1205

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.