cygwin_daemon status report.

["Robert Collins" <robert dot collins at itdomain dot com dot au>]

I've moved the sockets code out of the transport base class, so now the
changes to cygwin are even less intrusive.

I've one key change to make, and then I can't think of any core stuff
for the daemon (not the shm code) that needs to happen other than the
directory reorganisation discussion (+ actions).

So please, this is up for review, as I'm requesting inclusion in HEAD
immediately that 1.3.4 goes out the door I'd like to have the directory
reorganisation discussion now, so I can make time to do that.

The key change I referred to above is a new cygwin environment option -
"DAEMON". This will need to be present for any daemon checks to occur.

Why?

On 95, there is a noticeable timeout when process's start if the daemon
hasn't bound the socket. I imagine there is an equivalent, but much
shorter delay under NT to find /not find the pipe. So this will reduce
the potential impact of the daemon.

We could of course do it in reverse, have the option "NODAEMON".

Thoughts?

Rob

daemonstate.patch