If the FIX engine tries to connect to a remote engine under little or no network connectivity, it tends to be stuck in a unrecoverable state and cannot reconnect automatically when the connectivity restores. This appears to be caused by a NPE in the quickfix.mina.initiator.IoSessionInitiator$ConnectTask.connect() line: 136

quickfixSession.getLog().onEvent(e.getMessage());

where e.getMessage() is a null pointer, which causes quickfix.FileLog.writeMessage(java.io.FileOutputStream, java.lang.String, boolean) line: 104

stream.write(message.getBytes(CharsetSupport.getCharset()));

to throw a null pointer exception, which in turn puts the engine into an unrecoverable state. The event log looks something like this:

20070820-08:59:02: Sent test request TEST
20070820-08:59:06: Timed out waiting for heartbeat
20070820-08:59:06: Disconnecting
20070820-09:00:08: <------- stuck here

The implication is that if the network connectivity is poor enough to force the engine to disconnect, and if the condition lasts during the first reconnect attempt, the engine will be stuck and never recover even if the network condition gets better.

All line numbers are from 1.2.1 version of the source code. Java 1.5 was used.