[ESS] tcsh, bash, and the inferior R process

Martin Maechler maechler at stat.math.ethz.ch
Tue Oct 16 08:32:36 CEST 2007

>>>>> "EI" == Erik Iverson <iverson at biostat.wisc.edu>
>>>>>     on Mon, 15 Oct 2007 17:12:05 -0500 writes:

    EI> I now see from the ESS manual (Section 4.4.3) that
    EI> 4.4.3 Re-submitting commands from the transcript

    EI> ...

    EI> When the cursor is not after the current prompt, the <RETURN> key has a 
    EI> slightly different behavior than usual. Pressing RET on any line 
    EI> containing a command that you entered (i.e. a line beginning with a 
    EI> prompt) sends that command to the ESS process once again. If you wish to 
    EI> edit the command before executing it, use C-c RET instead; it copies the 
    EI> command to the current prompt but does not execute it, allowing you to 
    EI> edit it before submitting it.

    EI> ...

    EI> If I interpret the manual correctly, it appears that invoking Xemacs 
    EI> under tcsh is actually doing the wrong thing.  Regular Emacs under tcsh 
    EI> appears to do the right thing, that is, send the command that point is 
    EI> on *and* submit it. Xemacs under tcsh will copy the command to the 
    EI> command prompt and leave point at the *beginning* of the command upon 
    EI> hitting <RETURN> on a previous command.  C-c <RETURN> copies the command 
    EI> to the prompt and puts point at the *end* of the command.

    EI> C-c Ret works in Xemacs lauched from bash for me as documented.  I will 
    EI> now use C-c <RETURN> to do what I was used to doing, or remap the 
    EI> <RETURN> key since I know the elisp function names now.

    EI> So the upshot is that I (believe) XEmacs lauched from a tcsh shell is 
    EI> actually doing the *wrong* thing, and I had just gotten used to it.

Yes, that's correct.

As to *why* you observe this,
Sebastian Luque gave you a very good answer!

[ The following is a bit technical and only very vaguely related
  to ESS :
  I've seen similar behavior (both with KDE and GNOME) and with
  recent releases of OS / desktop actually have been quite
  frustrated by the fact that login via gdm/xdm (the "graphical login")
  seems to do less and less ``shell & login initialization''
  (~/.bashrc ~/.profile, etc) as compared with simple 'ssh' login
  or starting a shell in a terminal.

  I vaguely remember I've been able to work around these
  infelicities by doing things like  ". ~/.profile" inside 
  my ~/.Xclients ...
  but then I've also found that ~/.Xclients had been completely
  replaced just by clicking some desktop GUIs like "Desktop switcher"
  ...  {aarrgh ... feels like using Micro$oft products ..}


More information about the ESS-help mailing list