mailing-list for TeXmacs Users

[Michael Lachmann <address@hidden>]

On 16 Nov 2009, at 16:37, Joris van der Hoeven wrote:

> In principle, when launching ssh with the right flags plug-ins  
> launched
> over a pipe behave *exactly* the same way as on your local machine.
> One important thing is that you *must* set up the necessary keys
> in your personal .ssh so that log-in does *not* require a password.


I see.. so you are saying that one should figure the exact flags/ 
configurations one needs to have ssh allocate a pseudo-terminal when  
you call something like
ssh host /usr/bin/R
instead of doing
ssh host
which launches a shell, in which you have to then run R.

And then, if you want a series of remote hosts, you need to do
ssh host1 ssh host2 R
again with some flags, if possible.

Wow! It works! And I spent all day on the train getting tm_r to work  
"well" without allocating a pseudoterminal. At least I cleaned up the  
code a bit in the process...

The flags for ssh are -t -t

So, to log in remotely, you have to use

ssh -t -t host -l user path_to_R

or
ssh -t -t host1 -l user1 ssh -t -t host2 -l user2 path_to_R
(which is what I just tried, and it worked!)

And you think the best is to require the user to set up everything so  
that no username/password is needed to log to the hosts, and the  
command is always R, or should the user just set things up so that no  
password is needed, but she can still specify host/username  
combinations and command to run?

Michael



On 16 Nov 2009, at 16:37, Joris van der Hoeven wrote:

> Hi Michael,
>
> I did some experiments with remote sessions in the past.
>
> If you have the courage, please refigure out how to do the above
> (I forgot in the meantime), and give a more precise procedure.
> We might then consider a Remote session menu in which a remote host,
> or even a sequence of remote hosts could be specified,
> which could be used as a wrapper for any plug-in.
>
> Best wishes, --Joris
>
>
> On Fri, Nov 06, 2009 at 04:25:21PM +0100, Michael Lachmann wrote:
> > Hi,
> >
> > There is a hidden feature of the R session that allows a user to log
> > to a remote host to execute the R session there. I use it quite a  
> > lot,
> > because my local machine is sometimes not strong enough, so I can ssh
> > to another machine, and run R there.
> >
> > The way it works, though in the current version it seems a bit  
> > broken,
> > is that you set two environment variables:
> > TEXMACS_CMD and TEXMACS_SEND
> >
> > TEXMACS_CMD you set to a command to execute instead of R. For  
> > example /bin/tcsh
> > TEXMACS_SEND you set to a string to send to that command, instead of
> > the load library command, for example "".
> > Then you have a shell prompt, you ssh to the remote machine, and then
> > run R. You still have to load the TeXmacs library on the remote
> > machine, so you should install it there. And, it is important that  
> > you
> > can log in without a password prompt, because that is not handled.
> >
> > Quite a complicated procedure...
> > It has to be a bit flexible, because sometimes you'll have to log  
> > to a
> > gateway and then from there log to the computer you want, or some
> > such.
> >
> > Anyway, it would probably be good if one could tell TeXmacs from
> > within the program which type of session you want - R or shell.
> > Possibly other session types also benefit from accessing a remote
> > server.
> >
> > The question is how to do that best? One could easily have two
> > different session types - R and R/shell, one starts R, the other
> > starts a shell, but is ready to interpret R... But I don't think that
> > is a really good way to handle things.
> > A fairly good way would be to be able to switch the type of the
> > session from within a session. So you start a shell session, log to
> > the remote machine, and then launch R, and switch to an R session.
> > (But currently it seems that the shell session does not handle ssh
> > very wel...)
> >
> > A related question is this: when I run an R session in a buffer, and
> > then another R session in another buffer, the buffers share the R
> > seesion - only one instance of R is launched. Sometimes it is useful
> > to have a second instance. Currently the only way to achieve that is
> > to launch TeXmacs again, I think.
> >
> > Michael