-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Hi Jonathan,
@31--2002, 08:56 -0500 (14:56 UK time) Jonathan Angliss said: > It was actually Marck that suggested it may be an internal > component that is not recording the disconnect. It was - > ... I'd still be curious to see if they are using a global boolean > to record a connection though. I frowned on that thought (although it's a possibility) because of TB's multi-threading on the connection side. It would be very difficult to monitor which task would be the "last to leave the building" and "turn out the light". I would say that as an object-oriented piece of software I would make it an incremental count mechanism within the socket class - self maintaining. > Hrm... then Windows is sending a valid connection terminate to the > client, and it is acknowledging it safely, and terminating the > connection. So it is clearly not that. There must be some other > factor. The only factor that counts is that TB believes a connection is live that isn't. Something internal to TB is causing it to think that. We don't seem to be reliably able to duplicate the circumstances that would induce it to happen at will. RITlabs haven't a hope in hell of fixing it until we can find one. - -- Cheers -- .\\arck D Pearlstone -- List moderator TB! v1.61 on Windows 2000 5.0.2195 Service Pack 2 ' -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.1.91-nr1 (Windows 2000) iD8DBQE9cNMNOeQkq5KdzaARAnB2AKCfHBhZbGSekpI+pLiTwfqFeTD6uQCgyfC/ 3cbGnK7NJf9XwZt3gA4q0FQ= =yIOU -----END PGP SIGNATURE----- ________________________________________________ Current version is 1.61 | "Using TBUDL" information: http://www.silverstones.com/thebat/TBUDLInfo.html