socket timing problem

[David LeBlanc]

In article <uwv8fq8zu.fsf at ctwd0143.fitlinxx.com>, db3l at fitlinxx.com 
says...
> "Alex Martelli" <aleaxit at yahoo.com> writes:
> 
FWIW, it was reported on comp.lang.tcl that NT4.0sp4 introduced some 
breakage in the tcp/ip stack that was fixed in a subsequent service pack.

Dave LeBlanc

> > I've used it successfully (with my changes as incorporated)
> > under NT4 SP6 (I do have a recent 'Platform SDK' installed,
> > but I don't think it changes anything in this regard).
> 
> Hmm - we're still having difficulties.  For example, the accept()
> wrapper seems to return a native socket and not a TimeoutSocket.  But
> at least for us (also NT4 SP6), the socket that accept() returns when
> the main socket was non-blocking is also non-blocking.  Thus, when you
> later attempt to send/receive over it, you're non-blocking, but
> without the wrapper support of the TimeoutSocket class.  This seems to
> particularly show up when trying to use this with Fredrik's XMLRPC
> library.
> 
> We also seem to get a code 10022 but with why[1] set to "winsock
> error" rather than "Invalid argument" in the connect processing.
> 
> It sounds like I'll just poke around a little more and see if I can
> summarize everything in one swoop.  Thanks.
> 
> --
> -- David
>