https://svn.apache.org/repos/asf/incubator/qpid/branches/M2.1.1
........
r598285 | ritchiem | 2007-11-26 14:16:01 +0000 (Mon, 26 Nov 2007) | 3 lines
QPID-92, QPID-564 : Upgraded Mina to 1.0.1 still not good enough but all future versions currently have a bug with the CumulativeProtocolDecoder. It compact()s the buffer which breaks slices. Added MultiThread Support which is some of the feature set of QPID-564
........
r598619 | ritchiem | 2007-11-27 12:51:14 +0000 (Tue, 27 Nov 2007) | 1 line
Renamed POMs to M2.1.1 Removed erroneous equals() in SpecificMethodFrameListenerTest
........
r598721 | ritchiem | 2007-11-27 18:09:33 +0000 (Tue, 27 Nov 2007) | 1 line
QPID-621 : Patch Supplied by Aidan Skinner. Msg Ack after msg consumer is closed.
........
r598834 | ritchiem | 2007-11-28 00:45:32 +0000 (Wed, 28 Nov 2007) | 14 lines
QPID-679 : Patch provided by Aidan Skinner and additional from odd problems during test runs.
AMQChannel - Catch and log AMQException occuring when requeue()-ing. Previously exceptions wouldn't be caught at all. The requeue() is called during closure so there is nothing we can do protocol wise on error other than log the issue and continue with any other shutdown that is needed.
AMQMinaProtocolSession & AMQPFastProtocolHandler . Additions to catch and log AMQExceptions. Changes to AMQMinaProtocolSession were done to ignore all input on a closing session other than the close-ok. Previously only Protocol frames were ignored this resulted in Content*Body-s still being processed. Additional checks were made for the MessageStoreClosedException to log and continue. As said else were we need to seperate protocol exceptoions(AMQException) from internal code exception handling. Further All AMQExceptions occuring in the frameReceived method are now caught and logged. Allowing them to propogate higher will only result in thread death.
AMQPFastProtocolHandler Caught AMQExceptions occuring whilst closing the session. Again allowing these to continue will result in thread death. There is not a lot that can be done other than log the problem as the session is already closed by this point. Prevented the stacktrace associated with a session exception being printed in the exceptionCaught method when the problem was an IO Exception. This doesn't add anything useful and only adds to the log file sizes.
ApplicationRegistry - Added removeAll option which ensures that all ARs are correctly purged so that we can attempt to clean up between Unit Tests.
MemoryMessageStore - This was causing us real problems during the failover testing. Similar checks should probably be made to any other Message Store Impl. The issue was that when shutting down the broker the MS.close() method is called this sets all the storage to null. However, there may still be message processing going on as the close() does not attempt to stop connection processing. Hence we now check to see if the Store is close throwing a MSClosedException if required. This prevents NPEs that have been seen during Unit failover testing. In fact the close() is called as a request to shutdown the ApplicationRegistry, but this only occurs from tests and broker shutdown, no attempt to unbind or prevent further connections during this period is yet done.
CLIENT CHANGES
AMQConnection - Added method to check if failover is in progress.
AMQClient - Upgraded acknowledge() exception to JMSException for errors due to failover. Also , added call to update consumers as a result of failover.
BasicMessageConsumer - Changes to acquireReceiving to take in to consideration blocking for failover to occur. wrt receiveNoWait.. which previously blocked for failover to complete... not exactly noWait. acknowledge will now
TransportConnection - Update to ensure all inVM brokers are correctly killed.
FailoverTest - QPID-679 - Finder of all the above problems.
........
r598835 | ritchiem | 2007-11-28 01:01:05 +0000 (Wed, 28 Nov 2007) | 1 line
CommitRollbackTest - this one just was never right.. now we have something better.
........
r599375 | ritchiem | 2007-11-29 10:58:08 +0000 (Thu, 29 Nov 2007) | 1 line
Update to broker to address fanout python failure.
........
r599531 | ritchiem | 2007-11-29 17:56:12 +0000 (Thu, 29 Nov 2007) | 1 line
QPID-92 QPID-564 Forgot to upgrade mina to 1.0.1
........
r599533 | ritchiem | 2007-11-29 18:25:21 +0000 (Thu, 29 Nov 2007) | 1 line
QPID-564 QPID-92 Tidied up a few points and fixed infinite loop in Read IO Thread
........
r599572 | ritchiem | 2007-11-29 20:56:22 +0000 (Thu, 29 Nov 2007) | 2 lines
Mina Fix: Vm Pipe Starts Connection session before acceptor session. This results in protocol frames arriving before the protocol decoder has been configured on the InVM Broker. Verification of this could be done by adding a client side filter that delays the first message by a few seconds.
........
r599805 | ritchiem | 2007-11-30 12:47:08 +0000 (Fri, 30 Nov 2007) | 1 line
Added new simple Request/Repsonse code as my last commit here seems to have missed the actual code.
........
r602134 | rupertlssmith | 2007-12-07 16:00:14 +0000 (Fri, 07 Dec 2007) | 1 line
Added JDNI config for two broker, failover setup for failover tests. Also passed into FT tests config.
........
r604151 | ritchiem | 2007-12-14 10:40:37 +0000 (Fri, 14 Dec 2007) | 2 lines
QPID-707 : Added new test to check message count on broker as messages are consumed to ensure that an ack is sent at 5000 mgs. Added acks on message consumer closure.
Augmented VMTestCase to have helper methods for accessing broker statistics.
........
r604928 | rupertlssmith | 2007-12-17 17:00:10 +0000 (Mon, 17 Dec 2007) | 1 line
DUPS_OK mode set to be same as AUTO_ACK, fixed broken dups ok test.
........
r605536 | rupertlssmith | 2007-12-19 13:40:05 +0000 (Wed, 19 Dec 2007) | 1 line
Messages were being sent mandatory by default, set to false.
........
r605542 | rupertlssmith | 2007-12-19 13:53:44 +0000 (Wed, 19 Dec 2007) | 1 line
Changed test configs to use colons instead of commas.
........
r606015 | rgodfrey | 2007-12-20 20:08:01 +0000 (Thu, 20 Dec 2007) | 2 lines
QPID-714 : (Patch from Aidan Skinner) Issue with competing, transactional/client-ack consumers
Ack each individual message on commit, not use multiple acks
........
r606016 | rgodfrey | 2007-12-20 20:12:25 +0000 (Thu, 20 Dec 2007) | 2 lines
QPID-714 : (Patch from Aidan Skinner) Issue with competing, transactional/client-ack consumers
Ack each individual message on commit, not use multiple acks
........
git-svn-id: https://svn.apache.org/repos/asf/incubator/qpid/branches/M2.1@614906 13f79535-47bb-0310-9956-ffa450edef68