A couple of weeks ago (I think it was), I posted about a strange problem
I had in my QuickFIX/J application, wherein TCP connections were not
being correctly made (they would shut down immediately after the
three-way handshake was complete) under certain conditions until I
introduced a delay in my Application implementation's onCreate()
method. This is on the Initiator side of the FIX conversation. A
one-second delay seemed to do the trick, 100% of the time while running
in a debugger, and perhaps 95% of the time while not in a debugger. I
wrote this off, basically, as an unexplained problem solved by an
unexplained hack that worked enough of the time that I could just ignore
it for now.
Well, today I discovered something else completely by accident; this may
provide a clue for someone knowledgeable about QuickFIX/J's internals to
track it down.
My discovery: if I call <Session>.logout() in the
<Application>.onCreate() method, the problem seems to completely
disappear, even with my hack one-second delay removed. I have, at this
point, started successfully over 100 times without the debugger --
something I've never been able to do before. Quite pleasant, actually
<smile>.
I only found this because I was implementing the logic to let my
application manage the logons and logouts (desirable for many reasons in
my particular application), and I just happened to notice that the
occasional hang at startup disappeared.
I hope someone can gnaw at this clue and come up with an answer...
Tom...
|
