[Quickfix-developers] QuickFIX/J mystery....

[Tom D. <to...@di...>]

Hi, folks...

Any insight into this strangeness would be very welcome...

I have a QuickFIX/J application that is working, but only when I put 
some baling wire and chewing gum (a hack) into it.  This application has 
a single initiator session, and I've used ThreadedSocketInitiator.  
Without the hack, on startup the following sequence occurs (via network 
sniffer):
-- a TCP connection is established
-- the TCP connection is taken down (and the debug log shows a 
"responder: disconnect" event).  The takedown is initiated by my 
application's side of the connection.
-- the above two sequences are repeated, exactly
-- a TCP connection is established for the third time, stays 
established, but then nothing else happens.  Even a logon() call has no 
effect.

If I run the same exact code with a debugger (I'm using IntelliJ Idea), 
and if I single-step over the call to initiator.start(), and if I have 
other threads suspended while single-stepping (all three conditions are 
necessary), then everything works just like you'd expect -- the 
connection at the TCP level is established on the first try, and the 
logon proceeds normally. 

While putzing around trying to imagine what the heck this could be, I 
discovered the hack: if I insert a delay (a sleep()) of 350 milliseconds 
or greater in the onCreate(SessionID) method of my Application, then 
everything works great, first time, every time.

Does this ring a bell with anyone?

BTW, I'm a newbie on QuickFIX/J, so don't dismiss any bozo maneuvers by 
saying to yourself "Nah, nobody could be that stupid!"  Because yes I could.

Thanks,
Tom...