Benjamin Reed commented on ZOOKEEPER-58:

Excellent catch! I think it is better to do this in either primeConnection or 
startConnect, that way we make sure everything is setup correctly at the start 
rather than trying to make sure that we clean up correctly. Of the two my 
preference would probably be startConnect.

> Race condition on ClientCnxn.java
> ---------------------------------
>                 Key: ZOOKEEPER-58
>                 URL: https://issues.apache.org/jira/browse/ZOOKEEPER-58
>             Project: Zookeeper
>          Issue Type: Bug
>          Components: java client
>            Reporter: Flavio Paiva Junqueira
>         Attachments: patch-incoming-race.txt
> There is a race condition involving the ByteByffer incomingBuffer, a field of 
> ClientCnxn.SendThread. SendThread reads a packet in two steps: first it reads 
> the length of the packet, followed by a read of the packet itself. Each of 
> these steps corresponds to a call to doIO() from the main loop of run(). If 
> there is an exception or the session times out, then it may leave 
> incomingBuffer in an inconsistent state. 
> The attached patch adds code to reset incomingBuffer upon a call to 
> SendThread.cleanup(). This method is called upon an exception on run().

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

Reply via email to