[Quickfix-users] Possible Checksum bug originating in FieldBase::setString()

[Brendan B. B. <br...@ka...>]

Hello,

In Field.h::void FieldBase::setString( const std::string& string ),
shouldn't m_total be inited to 0 before accumulating it? e.g.

  void 
  { 
    m_string = string;
    m_data = IntConvertor::convert(m_field) + "=" + string + "\001";
    m_length = m_data.length();
    const char* iter = m_data.c_str();
    const char* end = iter + m_data.length();

    // tbd1234: brendan - bug?
    m_total = 0;
    while( iter != end )
      m_total += *iter++;
  }


W/o this, something like:

FIX::ExecType execType;
FIX::OrdStatus ordStatus;
execType.setValue(FIX::ExecType_NEW);
ordStatus.setValue(FIX::OrdStatus_NEW);

...send ExecReport... 

would cause a checksum mismatch between executor and a client due to
these two fields.

In addition to not getting the checksum errors, this also appears to
have fixed another problem I saw where executor would repeatedly ask
my client (and possible the client would ask the executor - depending
on the state at that moment) for resends.  While debugging the
checksum problem I found cases where the client, in attempting to
honor the resend would throw an InvalidMessage due to a message gap
in the FileStore e.g. the FileStore would have messages 5, 6, 7, 9,
10 as a "" string was being pushed onto the queue of messages to
resend.

Since applying the m_total = 0 fix, I've been able to down executor,
bring it up and have it resync w/the client with none of the problems
I've previously seen.

Regards,
Brendan