james-server-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Noel J. Bergman" <n...@devtech.com>
Subject RE: svn commit: r471243 - /james/server/trunk/src/java/org/apache/james/util/connection/ServerConnection.java
Date Mon, 06 Nov 2006 23:45:14 GMT
Stefano Bagnara wrote:

> The second connection made (using a connection limit of 1) expect 
> to received a connection exception, instead it seems to get
> stalled waiting to connect.

If I understand you (and the code in testConnectionLimitExceeded) correctly, it seems that
the defect is in the test's assumption not the ServerConnection code.

A connection limit of 1 means that the first connection will be handled, and the second connection
will be ESTABLISHED in the TCP/IP connection queue.  That is the expected and desired behavior.
 The old behavior was wrong: it constantly purged the backlog queue and discarded connections.
 The correct behavior processes the queued connection when a worker is available.

Kudos to Berd's code for detecting the change, but what needs to be changed is the test, not
the revised ServerConnection.  :-)

	--- Noel

To unsubscribe, e-mail: server-dev-unsubscribe@james.apache.org
For additional commands, e-mail: server-dev-help@james.apache.org

View raw message