To: stik-beta-testers@list.zetnet.co.uk
From: dlanor@oden.se (Ronald Andersson)
Organization: Atari
Subject: Re: New TCP.STX Beta UDP.STX
X-Mailer: NEWSie Version 0.94 (Atari)


Hi Olivier,

>I confirm that TCP_close do not kill CIB struct ( for LOOPBACK address and/or
>more ) so I suppose STING's memory decrease.

This is probably not the case.  I think it is only an effect of the internal
delay while the interrupt driven routines wait for a correct termination of
the TCP connection.  This will time out 1000 seconds after the TCP_close call
was made, and then the connection will be forcibly shut and all its RAM will
be released.  Your later follow-up on this report seems to confirm this.

Now the question is WHY the connection could not be closed in normal fashion,
but had to be handled by this internal timeout routine.  That probably is
tied to the kind of test you were making at the time though...


>If you want, I have a graphic tool for watch/close all TCP-UDP handles 
>present, is a beta but very usefull for see the problem.
>
Yes please.  I want all useful STinG tools that exist (of course!).
Other STinG users probably feel the same, so please post it here.


>I do not kown if you may fix a old problem about CNbyte_count on UDP 
>listen handle, always return E_NORMAL and not E_LISTEN as TCP :(

I can, but it will not be quite that easy.  I'll also have to add the other
stuff Peter has added since I got the UDP source I have (from August 1997).
That is mainly the new open modes for having listening connections with fully
specified socket data (only listens to a specific remote IP:port combination).
This is work I'll have to do anyway, sooner or later, unless Peter returns.


Btw:
I have sent an enquiry to Laser Zentrum Hannover (Peter's employer) asking
them for any information at all concerning his present whereabouts, but as
yet I have not received any answer.

