18 Feb, 2012
1 commit
17 Feb, 2012
5 commits
-
The SMPP specification specifically lists that a PDU body with a NULL field message_id is required in the deliver_sm_resp message. Using the sendNegativeResponse() will result in a body-less PDU response that some SMPP servers legitimately refuse with an error about invalid command length. Now we correctly send the deliver_sm_resp in case of a ProcessRequestException. For PDUStringException we now send a generick_nack instead of a deliver_sm_resp, just like in other cases were such an exception is thrown.
-
A different commit has added logging of all PDUs at debug level so here we reduce the log messages about activity or inactivity.
-
All PDUs received and sent over SMPP connection are now logged at debug level
16 Feb, 2012
6 commits
-
After the smpp session has been closed with close() method, we shouldn't leave any threads running. Any extra threads should have exited before Session.close() exits, otherwise many webapp containers will complain during webapp shutdown or webapp restart
-
There was duplicate code in SMPPSession and SMPPServerSession that should live in AbstractSession.
15 Feb, 2012
3 commits
-
When an smpp session is bound and there are messages waiting on the smsc side, it is legal for the smsc to send a DeliverSM PDU immediately after it sends a bind response. Because of the PduProcessorDegree, there is a race condition were both the BindResp and the DeliverSM can simultaniously be processed but the DeliverSM will be processed like the session is still in the open state instead of the bound state. This patch changes the size of the executorService pool to remove the race condition. It starts with size 1 in open state and when the session is bound, the pool will be enlarged to the PduProcessorDegree.
-
Change PDUProcessTask so that state processor is retrieved when task is actually executed, not when task is enqueued. This solves a race condition where the bind state changes while a task is enqueued
-
…a session, so the passed object can be a Session instead of a generic Object
14 Feb, 2012
4 commits
-
…er some network conditions, it is possible for the code to receive a partial package over the socket and erronously throw an IOException for the socket. The specs for read() don't guarantee that the full requested length is read in one call. Using readFully() fixes that problem.
-
…eiverListener set, otherwise the smpp server will receive an ok for a message that was not correctly delivered
-
…ession. This allows creating objects that behave like SMPP sessions without subclassing the heavy-weight AbstractSession or SMPPSession classes.
06 Feb, 2012
1 commit
