QuickFIX/J

• Type: Bug
• Status: Closed
• Priority: Critical
• Resolution: Duplicate
• Affects Version/s: 1.4.0
• Fix Version/s: 1.5.0
• Component/s: Engine
• Labels:

  None
• Environment:

  All

QuickFIX/J has an issue with the file store and message sequence numbers when simultaneously answering a ResendRequest on a connection, and processing brand new messages routed to that connection which are also to be sent to the client. The file pointers were not handled correctly for reading stored messages which are to be sent as PossDups, and for new messages to be written to the outgoing file. This caused outgoing message store file corruption, and repeated / out of sequence seqnums being sent.

Fixed: synchronization QFJ patch, where the thread is blocked from sending new messages (i.e. buffering) until a ResendRequest is fully answered.

    private void nextResendRequest(Message resendRequest) throws IOException, RejectLogon,
     FieldNotFound, IncorrectDataFormat, IncorrectTagValue, UnsupportedMessageType,
            InvalidMessage {
     if (!verify(resendRequest, false, false)) {
               return;
            }

// ******* ADDED *******
state.lockSenderMsgSeqNum();
     try {

int beginSeqNo = resendRequest.getInt(BeginSeqNo.FIELD);
int endSeqNo = resendRequest.getInt(EndSeqNo.FIELD);

getLog().onEvent(
"Received ResendRequest FROM: " + beginSeqNo + " TO: " + formatEndSeqNum(endSeqNo));
....

// ******* ADDED *******
     } finally{
     state.unlockSenderMsgSeqNum();
     }
    }

(Internal reference: IC_57785)

This issue duplicates:
QFJ-429	Concurrency reading / writting file store

• Assignee:

  Unassigned

  Reporter:

  SSE

• Created:

  28/Jun/10 2:47 PM

  Updated:

  12/Apr/11 3:42 PM

  Resolved:

  12/Apr/11 3:42 PM