scp with openssh on the server side and $PATH.

Philip Hands phil at
Thu Dec 30 22:24:18 EST 1999

Marc.Haber-lists at (Marc Haber) writes:

> On Tue, 28 Dec 1999 13:43:10 -0500 (EST), you wrote:
> >Don't feel stupid.  The configure script accepts the --with-default-path
> >option, but doesn't do anything with it. 
> I see. This needs to be documented or other people will fall into that
> trap.
> >I haven't figured out autoconf that well yet, so maybe someone who knows
> >it can hack this into
> Just to solve it _now_, can I safely put
> # if defined (_PATH_STDPATH)
> #  undef _PATH_STDPATH
> # endif
> # define _PATH_STDPATH "_my_path_"
> #endif
> in config.h without breaking anything? Installing a broken sshd on the
> machine in question will result in a 300 km drive :-(

If you kill only the server process, you should be left with the ssh
session(s) on which you have already logged in (This is how the Debian
ssh packages ensure that you can do an upgrade via an ssh session
without cutting your own throat).  The server process should have
written its PID into a pid file, the name of which will probably be
revealed by:

  strings /usr/sbin/sshd | grep '\.pid'

Once you've done that, you can run a new server, and try logging in
again before you log out on the original session.  If that fails, you
can back out by running the old server.  Once you've proved it's all
working, you can move the sshd into place in /usr/sbin (or whatever).

If you're paranoid, you can always schedule a reboot in ten minutes
time before you start, so that if you kill your own session at least
things will recover.  You'll need to remember to cancel the reboot if
things go well though.

Cheers, Phil.
Boycott Amazon! ---

More information about the openssh-unix-dev mailing list