readv/writev patch for sockets

[Conrad Scott]

"Corinna Vinschen" <cygwin-patches@cygwin.com> wrote:
>
> On Tue, Sep 03, 2002 at 02:49:14PM +0100, Conrad Scott wrote:
>
[ChangeLog entry snipped]
>
> Applied.
>
> Thanks!

And thanks to you (and Chris) too for looking through all these socket
patches and committing them for me.

I started on the socket patches because of issues generated by my
cygserver work.  Since I've done all I need for that (and a little
more), I'll get back to the cygserver stuff now and leave the sockets
alone.

Having said that there is still the race condition in the UNIX domain
socket implementation that needs to be fixed sometime.  I've not got a
good idea currently on how to achieve this; my idea of using the
"connect data" won't work (AFAIK) since it's not supported on TCP/IP
(obviously).

The only other loose end I came across while reading SUSv3 (a
fascinating read for insomniacs) is that connect(2) can be called with a
null address to "disconnect" a datagram socket.  Currently the argument
checking in net.cc disallows this; presumably this is a simple one to
fix but I've not found the energy for it (and the associated testing)
this week etc.

So, for the moment, thanks for the support and I'll get back to the
cygserver code.

Cheers,

// Conrad