[
https://issues.apache.org/jira/browse/POOL-162?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12846729#action_12846729
]
Phil Steitz commented on POOL-162:
----------------------------------
Thanks, Mark!
Gave me a headache, but the test and fix look good, modulo one comment. With
your patch, if the latch has been served when the thread is interrupted, we do
not propagate the interrupt, but let the thread continue and get served. This
seems reasonable for most use cases. The alternative would be to handle
mayCreate and pair != null separately, in the second case destroying the object
and in both cases removing the latch from the queue and setting thread
interrupt status. If the interrupt is to shorten or end the wait, the patch
impl is probably best. If for another reason, the client might rather propagate
the exception and not use pool capacity. I can't t think of realistic use cases
where the second would be the case, so unless someone else can, I am +1 on
resolving this based on the fix in r924479
> When waiting threads are interrupted, pool can leak capacity
> ------------------------------------------------------------
>
> Key: POOL-162
> URL: https://issues.apache.org/jira/browse/POOL-162
> Project: Commons Pool
> Issue Type: Bug
> Affects Versions: 1.5, 1.5.1, 1.5.2, 1.5.3, 1.5.4
> Reporter: Phil Steitz
>
> As reported on commons-dev (http://markmail.org/message/aqb23nnzyy2ar3vs),
> when waiting threads are interrupted, GOP, GKOP may leak capacity. I do not
> yet have a test case to confirm this, but I suspect that the problem reported
> by the user is caused by a missing _allocationQueue.remove(latch) before
> rethrowing InterruptedException in borrowObject.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
