[Quickfix-developers] bug in Session.cpp, and more.

[Scott K. <sc...@la...>]

See end for my main question here.


I think I found a bug in how the ResendRequest messages are handled. Please
correct me if I'm doing something wrong here as it's hard for me to believe
this version of quickfix is so buggy. I'm using the latest version 1.13.3

This one's a bit more complicated than the extra const's in
MessageCracker.h...

 I have QuickFix 1.13.3 installed on two machines, one as acceptor, one as
initiator. The symptom:Anytime a ResendRequest is received, the QuickFix
that receives it is NOT incrementing  the NextTargetMsgSeqNum for that
ResendRequest message. The ResendRequest is correctly handled, but now the
system that received the original ResendRequest will think the *other*
system is out of sync, as soon as that other system sends another message,
as it's MsgSeqNum will be one too large. This results in the two systems
trading ResendRequests, once each heartbeat.

Looking in the source, it appears this happens only when persistMessages is
NOT on. In that case, the code (line ~400 of Session.cpp) generates a
SequenceReset and returns. The code never reaches the end where (line ~482)

  if( !isTargetTooHigh(msgSeqNum) && !isTargetTooLow(msgSeqNum) )
    m_state.incrNextTargetMsgSeqNum();

is done.

I fixed it by putting all the code that re-sends the messages in an else
clause, and commenting out the line with the return, so that the
incrNextTargetMsgSeqNum still gets called.


In testing this, I found that UseLocalTime=Y appears to have no effect in
the settings, and UTC time was still used for the times I specified in the
settings. I stepped through the code that initializes the settings and it is
correctly picking up the values from the file, and using LocalTimeOnly
container. The problem appears to be in the fact that the TimeRange
container does not do anything differently to accept a LocalTimeOnly as
compared to a UtcTimeOnly object. What's going on here?? I just get around
it now by setting different times...


In testing that, I changed the settings on my server (running the accepter)
for the session time and restarted it. Upon doing that, the first couple
Logon attempts were refused (server sent a Logoff message in response to my
Logon message), and did not increment its sending message number in the
process. Eventually it sorted it's sh*t out and sent a Logon response with
the message number jumped up to the correct value. I have not tried to track
this down.


MY MAIN QUESTION:
Does this sound like I'm doing something wrong? Not setting it up correctly?
Are these problems normal? Or is this version of Quickfix just horribly
buggy? How long has this version been released? I am using it for about a
week and have found what appears to be 4 serious bugs... Either I am the FIX
newbie that I am, or this version of Quickfix is not ready for the
primetime.

I guess that's it. I appreciate any feedback anyone has for me on these
issues.

Thanks,
Scott