Creeping along with ESS ... more details

Rich Heiberger rmh at
Thu Apr 25 03:25:25 CEST 2002

ssh is probably the problem.  On NTemacs (I don't know about Xemacs)
the bash running in the *shell* is not a tty.  I don't know if it is
an emacs issue or a cygwin issue.  In either case, when you type
you can't reply to the login/passwd sequence because they are handled by
a subprocess that depends on your being in a tty.

Check this by typing
to the bash prompt in the *shell* buffer and again to the bash prompt
in the stand-alone cygwin window.

My first attempt at a partial workaround isn't working very well.
I used ssh-agent to setup an identity that is recognized by the
remote machine.  Then 
   ssh remote.machine
gets me into a shell that thinks it is not interactive, hence no prompt.
Two choices here, run another interactive shell with
   sh -i
or don't bother with it since it doesn't pass that information to R.

Start R with
   R --vanilla
I am now in R with no prompts.  I can type and get something back.
The behavior isn't good enough to be useful.
I can type
and it's not much better.

C-c C-n works from the tmp.s buffer if the command is valid.  If the
R/S command you send over generates an error, then R halts and you are
back at the unix shell with no prompt.

Net result:
either ssh or cygwin or emacs needs to be changed.  I don't immediately
see how a fix to ESS would get past this behavior.  Maybe a change to R
would help, but I don't have any sense of what that would be.

I want, for this and other reasons, for bash running in a *shell* buffer
to be recognized as a tty.  That is either emacs or cygwin's responsibility.


ps. telnet version,
I use the telnet based on the one that I got from the NTemacs site.
           with a bug fix by Larry Smith in

ess-help mailing list -- To (un)subscribe, send
subscribe	or	unsubscribe
(in the "body", not the subject !)  To: ess-help-request at

More information about the ESS-help mailing list