[
https://issues.apache.org/jira/browse/ZOOKEEPER-1861?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13898637#comment-13898637
]
Camille Fournier commented on ZOOKEEPER-1861:
---------------------------------------------
Actually, I'm going to go back on this. The whole point of using a concurrent
data structure is so we don't synchronize on this operation. Which means we
have to allocate potentially unnecessarily.
I prefer patch v2. If you really want to try and not worry about allocation,
check for existence, then create, and put if not absent, but meh. I don't know
that I care about an (unlikely) extra allocation.
> ConcurrentHashMap isn't used properly in QuorumCnxManager
> ---------------------------------------------------------
>
> Key: ZOOKEEPER-1861
> URL: https://issues.apache.org/jira/browse/ZOOKEEPER-1861
> Project: ZooKeeper
> Issue Type: Bug
> Reporter: Ted Yu
> Assignee: Ted Yu
> Priority: Minor
> Attachments: zookeeper-1861-v1.txt, zookeeper-1861-v2.txt,
> zookeeper-1861-v3.txt
>
>
> queueSendMap is a ConcurrentHashMap.
> At line 210:
> {code}
> if (!queueSendMap.containsKey(sid)) {
> queueSendMap.put(sid, new ArrayBlockingQueue<ByteBuffer>(
> SEND_CAPACITY));
> {code}
> By the time control enters if block, there may be another concurrent put with
> same sid to the ConcurrentHashMap.
> putIfAbsent() should be used.
> Similar issue occurs at line 307 as well.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
