[cvsnt] Re: server method

[Tony Hoyle]

Nicholas Mongeau wrote:

 
> Lol, due to that informative, albeit unhelpful and mean response(I need my
> :server:!!!), I went to take a look at what was wrong, and it seems that
> :the
> only problem is that server_connect() now uses tcp_connect_bind()
> directly, and tcp_connect_bind() doest not set the static tcp_fd
> descriptor. So basically, the socket that has just been connected goes
> into oblivion while we try to use an uninitialized descriptor to send our
> data to cvs...

I tried changing that and it made no difference, although it's difficult to
tell what's wrong since the server is segfaulting all over the place.

> Please do not ditch the :server: method, it's not always a choice at work
> which method we have to use, and some of us are stuck with :server:... We
> need cvsnt, it's so much better than other cvses!!!

At least ssh has some security.  If you're using rsh basically your taking a
risk even having it on the servers (if your network is ever compromised,
even slightly, then your rsh enabled servers will be wide open).

If I can't support or test something it's just going to become a pain.  The
linux rsh server has apparently bitrotted for over a year now, and nobody
is even interested in fixing it.  I did get it working for a few hours when
I wrote the server protocol, but there seems to be no way of getting it to
work any more.

Tony