Re: tramp with shell mode and absolute paths

[Matthew Curry]

>
> That's what's possible as of today. I understand that you would like to
> have it more comfortable, but I don't see how this could be reached by
> simple means. How shall Tramp determine, whether "cd / <TAB>" shall
> complete on the remote host, or on the local host? Both would be acceptable.

I guess I'm confused, if you *start* a shell on a remote host via
tramp (and I'm not talking about ssh'ing in from a shell buffer you've
already started directly, that's totally different and that's obvious
why it wouldn't work), how is it *ever* useful to refer to files on
the local host?  [e]shell-mode should *always* use the remote host's
filesystem for all [e]shell-mode tasks.

Obviously for things like the M-x find-file function, it makes sense
to be able to refer to both local and remote hosts, and that works
really well.

Just to make sure we're talking about the same thing here, if I do a
M-x find-file:
/ssh:somemachine:/etc/hosts

I'll open the /etc/hosts file on the host somemachine.  While viewing
that file's buffer, if I do a M-x shell, it will start a shell *on the
remote machine*, in this case, somemachine.  It seems to me that once
I do that, it should never, ever use the local host for the life of
the shell-mode buffer.

Now maybe there is a reason why it acts like it does, but I'm not sure
why yet.  Do you know how [e]shell-mode keeps track of the current
directory of the shell it runs?

-Matt