Re: Q on using shell mode remotely

Re: Q on using shell mode remotely

[David Kastrup]

"Drew Adams" <drew.adams@oracle.com> writes:

> I said:
>
>     And, I couldn't cd to a remote directory within the
>     *shell* buffer, even using the syntax
>     /host-address:/home/my-home...

That would be rather silly to expect.  After all, the cd has to be
interpreted by the shell.

Try using
M-x eshell RET
and you'll find things work better.

> Actually, the cd command is accepted, and the new value of
> default-directory reflects it.

Emacs is fooled into believing that the shell is changing directories.

> But if I do pwd in the shell buffer, it still prints the Windows
> Temp directory (and ls shows files in that Temp directory).

Try doing something like

M-!

in a remote file buffer.

-- 
David Kastrup, Kriemhildstr. 15, 44793 Bochum

RE: Q on using shell mode remotely

[Drew Adams]

    >     And, I couldn't cd to a remote directory within the
    >     *shell* buffer, even using the syntax
    >     /host-address:/home/my-home...

    That would be rather silly to expect.  After all, the cd has to be
    interpreted by the shell.

Why? I did `M-x shell' within a Dired of a remote directory. The problem, I
imagine, would be to let the shell executing locally know that the directory
and files are remote.

If Emacs can see the remote files, then Emacs (and ftp) could interface them
to the local shell. I'm not saying it should work that way; I'm saying it
seems like it could. My question was just whether Emacs did anything like
that.

    Try using M-x eshell RET and you'll find things work better.
    Try doing something like M-! in a remote file buffer.

I know about those. I was asking about shell mode (see the Subject line). I
guess the answer is "you can't", so at least I'm no longer wondering.

Re: Q on using shell mode remotely

[Eli Zaretskii]

> From: "Drew Adams" <drew.adams@oracle.com>
> Date: Fri, 28 Jul 2006 15:56:52 -0700
> 
> If Emacs can see the remote files, then Emacs (and ftp) could interface them
> to the local shell.

??? How would Emacs manage to do that?  Could you please describe how
can this work conceptually?

M-x shell simply runs a local shell as a subprocess, making its input
come from what you type in the buffer, and its output redirected to a
pipe that feeds the buffer.  But the program which does the job when
you type "cd" is still the local shell, not Emacs.  So it's the shell
that needs to support some kind of remote file syntax, or else Emacs
will have no way of communicating to the shell that it (the shell)
should work on remote files.

RE: Q on using shell mode remotely

[Drew Adams]

    > If Emacs can see the remote files, then Emacs (and ftp) could
    > interface them to the local shell.

    M-x shell simply runs a local shell as a subprocess, making its input
    come from what you type in the buffer, and its output redirected to a
    pipe that feeds the buffer.  But the program which does the job when
    you type "cd" is still the local shell, not Emacs.  So it's the shell
    that needs to support some kind of remote file syntax, or else Emacs
    will have no way of communicating to the shell that it (the shell)
    should work on remote files.

If the shell is simply plugged in directly, then the shell is simply plugged
in directly. QED.

However, if cd can be interpreted by Emacs in other buffers to support
remote file syntax, then Emacs could do the same in a shell buffer. I'm not
saying it would be easy to do that, or that other problems wouldn't arise if
that were done, but input could be prefiltered and interpreted by Emacs,
before sending it to the shell process. IOW, instead of simply redirecting
input from the buffer, an intermediary could massage the input. No, I'm not
proposing an implementation; maybe someone else will ;-). If Tarantella can
do it, Emacs can do it better, I figure. ;-).

Q on using shell mode remotely

[Drew Adams]

My question is: can you use shell mode remotely?

I use ftp through Emacs (ange ftp or tramp with tramp-default-method = ftp)
to access files on a remote box. I have Emacs running on Windows, and the
remote box is GNU/Linux.

In general, I haven't been using shell-mode, locally or remotely. For remote
use, I've been using VNC or Tarantella, opening a terminal window to use a
shell - I haven't used Emacs remotely, except to edit files and such.

I thought I'd try to use Emacs shell mode on the remote box, so I gave it a
try: I opened Dired on a remote directory (/host-address:/home/my-home). I
then used M-x shell. However, the shell opened with a Windows Temp directory
as the pwd, and the shell that was used was my Windows shell (bash). And, I
couldn't cd to a remote directory within the *shell* buffer, even using the
syntax /host-address:/home/my-home...

So, my question is, can it be done? Can you use Emacs shell mode remotely?
If so, how?

Ideally, shell-mode would somehow figure out where I was (e.g. remote
location) and use the appropriate shell when I do `M-x shell', since I
logged into the remote box for ftp. If that couldn't be done automatically,
I could tell Emacs what shell to use for remote use.

Is anything like this possible? How do others use Emacs remotely?

RE: Q on using shell mode remotely

[Drew Adams]

I said:

    And, I couldn't cd to a remote directory within the
    *shell* buffer, even using the syntax
    /host-address:/home/my-home...

Actually, the cd command is accepted, and the new value of default-directory
reflects it. But if I do pwd in the shell buffer, it still prints the
Windows Temp directory (and ls shows files in that Temp directory).

Re: Q on using shell mode remotely

[Kevin Rodgers]

Drew Adams wrote:
> My question is: can you use shell mode remotely?
> 
> I use ftp through Emacs (ange ftp or tramp with tramp-default-method = ftp)
> to access files on a remote box. I have Emacs running on Windows, and the
> remote box is GNU/Linux.
> 
> In general, I haven't been using shell-mode, locally or remotely. For remote
> use, I've been using VNC or Tarantella, opening a terminal window to use a
> shell - I haven't used Emacs remotely, except to edit files and such.
> 
> I thought I'd try to use Emacs shell mode on the remote box, so I gave it a
> try: I opened Dired on a remote directory (/host-address:/home/my-home). I
> then used M-x shell. However, the shell opened with a Windows Temp directory
> as the pwd, and the shell that was used was my Windows shell (bash). And, I
> couldn't cd to a remote directory within the *shell* buffer, even using the
> syntax /host-address:/home/my-home...
> 
> So, my question is, can it be done? Can you use Emacs shell mode remotely?
> If so, how?
> 
> Ideally, shell-mode would somehow figure out where I was (e.g. remote
> location) and use the appropriate shell when I do `M-x shell', since I
> logged into the remote box for ftp. If that couldn't be done automatically,
> I could tell Emacs what shell to use for remote use.
> 
> Is anything like this possible? How do others use Emacs remotely?

M-x telnet

-- 
Kevin

RE: Q on using shell mode remotely

[Drew Adams]

    Drew Adams wrote:
    > My question is: can you use shell mode remotely? ...
    > If so, how?

    M-x telnet

Love it! Thanks, Kevin, once again!

This responds perfectly to what I was asking for. (I've used telnet a lot,
of course, but I didn't know that Emacs had a telnet mode.) This will save
me a lot of time futzing around with Tarantella and VNC.

I'm embarassed that I didn't find this on my own, especially since it is
mentioned in the Emacs manual, at least as far back as Emacs 20. RTFM Drew!
Thx.


I do have another question now, for anyone:

When I use M-x telnet, I get a prompt that looks like this:

   > 8:34am

I have some Emacs-Lisp code that recognizes the shell prompt, in order to
let you complete and cycle among past shell inputs. I use
`comint-prompt-regexp' to recognize the prompt; I subtract the prompt from
an input line to get the actual command used.

In plain (local) shell mode on Windows, this works fine; there, the prompt I
see is this: "> ". In telnet mode, however, the time of day interferes. It
is not actually part of the prompt as understood by Emacs (comint mode) -
that is, it doesn't match the default value of `comint-prompt-regexp'. For
both shell and telnet modes, this is the value of `comint-prompt-regexp':

   "^[^#$%>\n]*[#$%>] *"

On the remote machine, I do have a right-side prompt variable ($rprompt) set
to add the time at the *right* side of the screen. I use csh as the
remote-machine shell, and this is my $prompt:

   "%B%.01%b %h:"

This is my $rprompt (it is this that prints the time of day at the right):

   "%@"

Through telnet, the time appears next to the left prompt (i.e. at the
left) - I guess there is no notion of right-side prompt, but it still picks
up $rprompt and concatenates it to $prompt. Naturally, it is not picked up
by `comint-prompt-regexp' (which can't know by itself that $rprompt is being
used and its value is a time-of-day pattern).

Any ideas on how to deal with this? Either 1) how to recognize the presence
of the added time (appearing at the left) and treat it as part of the
prompt, or 2) (not as good, because it won't help others who use my code)
how to tell telnet not to use the time part of the prompt? Is there some
variable similar to `comint-prompt-regexp' that will pick up $rprompt (or
the concatenation of $prompt and $rprompt)?

Note: I'm not interested in hearing about things like
`comint-previous-matching-input' and `comint-previous-prompt' as the
solution here. I'm not looking for how to find previous inputs using vanilla
Emacs. Likewise, I'm not interested here in solutions that tell me how to
use the shell itself to reuse past commands. I want to enable my own
complete-and-cycle Emacs-Lisp code to work with the time-of-day that
$rprompt adds and telnet seems to stick on the left, after the $prompt. FYI,
my code is described here: http://www.emacswiki.org/cgi-bin/wiki/Icicles ...
(the site is down at the moment, so I can't give the exact URL - if you're
interested, look for Other Search Commands on this table-of-contents page).

Re: Q on using shell mode remotely

[Kevin Rodgers]

Drew Adams wrote:
> When I use M-x telnet, I get a prompt that looks like this:
> 
>    > 8:34am
> 
> I have some Emacs-Lisp code that recognizes the shell prompt, in order to
> let you complete and cycle among past shell inputs. I use
> `comint-prompt-regexp' to recognize the prompt; I subtract the prompt from
> an input line to get the actual command used.
> 
> In plain (local) shell mode on Windows, this works fine; there, the prompt I
> see is this: "> ". In telnet mode, however, the time of day interferes. It
> is not actually part of the prompt as understood by Emacs (comint mode) -
> that is, it doesn't match the default value of `comint-prompt-regexp'. For
> both shell and telnet modes, this is the value of `comint-prompt-regexp':
> 
>    "^[^#$%>\n]*[#$%>] *"
> 
> On the remote machine, I do have a right-side prompt variable ($rprompt) set
> to add the time at the *right* side of the screen. I use csh as the
> remote-machine shell, and this is my $prompt:
> 
>    "%B%.01%b %h:"
> 
> This is my $rprompt (it is this that prints the time of day at the right):
> 
>    "%@"
> 
> Through telnet, the time appears next to the left prompt (i.e. at the
> left) - I guess there is no notion of right-side prompt, but it still picks
> up $rprompt and concatenates it to $prompt. Naturally, it is not picked up
> by `comint-prompt-regexp' (which can't know by itself that $rprompt is being
> used and its value is a time-of-day pattern).
> 
> Any ideas on how to deal with this? Either 1) how to recognize the presence
> of the added time (appearing at the left) and treat it as part of the
> prompt, or 2) (not as good, because it won't help others who use my code)
> how to tell telnet not to use the time part of the prompt? Is there some
> variable similar to `comint-prompt-regexp' that will pick up $rprompt (or
> the concatenation of $prompt and $rprompt)?

Why can't you just add a time pattern to telnet-prompt-pattern, which is
used to set the buffer-local value of comint-prompt-pattern:

(eval-after-load "telnet"
   '(setq telnet-prompt-pattern
       (concat telnet-prompt-pattern "\\([12]?[0-9]:[0-5][0-9][ap]m\\)?")))

-- 
Kevin

RE: Q on using shell mode remotely

[Drew Adams]

    Drew Adams wrote:
    > When I use M-x telnet, I get a prompt that looks like this:
    >
    >    > 8:34am
    ...
    > Any ideas on how to deal with this? Either 1) how to
    > recognize the presence of the added time (appearing at the left)
    > and treat it as part of the prompt, or 2) (not as good, because
    > it won't help others who use my code) how to tell telnet not to
    > use the time part of the prompt? Is there some
    > variable similar to `comint-prompt-regexp' that will pick up
    > $rprompt (or the concatenation of $prompt and $rprompt)?

    Why can't you just add a time pattern to telnet-prompt-pattern, which is
    used to set the buffer-local value of comint-prompt-pattern:

    (eval-after-load "telnet"
       '(setq telnet-prompt-pattern
           (concat telnet-prompt-pattern
    "\\([12]?[0-9]:[0-5][0-9][ap]m\\)?")))

Thanks again. I didn't know about `telnet-prompt-pattern'. I used `C-h m'
and searched for "prompt", but the only occurences there are
`comint-next-prompt' and `comint-previous-prompt'. I found nothing in the
manual either. (I should have just used completion or apropos to look for a
"telnet-" prompt variable.)

I'll use `telnet-prompt-pattern', myself, to hack this problem, but it would
be good if there were a variable that dealt with $rprompt somehow.

Re: Q on using shell mode remotely

[Kevin Rodgers]

Drew Adams wrote:
> Kevin Rodgers wrote:
>     Why can't you just add a time pattern to telnet-prompt-pattern, which is
>     used to set the buffer-local value of comint-prompt-pattern:
> 
>     (eval-after-load "telnet"
>        '(setq telnet-prompt-pattern
>            (concat telnet-prompt-pattern
>     "\\([12]?[0-9]:[0-5][0-9][ap]m\\)?")))
> 
> Thanks again. I didn't know about `telnet-prompt-pattern'. I used `C-h m'
> and searched for "prompt", but the only occurences there are
> `comint-next-prompt' and `comint-previous-prompt'. I found nothing in the
> manual either. (I should have just used completion or apropos to look for a
> "telnet-" prompt variable.)
> 
> I'll use `telnet-prompt-pattern', myself, to hack this problem,

Please let us know if it works.  Looking at telnet.el:

1. M-x telnet runs telnet-program via make-comint, with
    telnet-initial-filter as the process filter, then calls telnet-mode.

2. telnet-mode sets the buffer-local value of comint-prompt-regexp to
    telnet-prompt-pattern.

3. telnet-initial-filter calls invokes
    telnet-check-software-type-initialize, then resets the process filter
    to telnet-filter.

    a. If telnet-check-software-type-initialize sees "unix" in its STRING
    argument, it resets telnet-prompt-pattern to comint-prompt-regexp;
    then regardless of STRING, it resets comint-prompt-regexp to
    telnet-prompt-pattern.

So it seems that (3) could prevent (2) from accomplishing its intended
effect, if (3) ran before (2).

Then, comint.el ignores comint-prompt-regexp anyway unless
comint-use-prompt-regexp is non-nil -- but it is nil by default, and
telnet doesn't set it.  So comint is assigning `field' text properties
to distinguish program (csh) output such as prompts from user input.

> but it would
> be good if there were a variable that dealt with $rprompt somehow.

I don't see how it could.  The remote csh presumably displays $rprompt
on the right side of the screen via terminal escape codes or simple
ASCII control characters; but the telnet subprocess is discarding that
data, either explicitly via telnet-filter or implicitly via process-
connection-type.  So you end up with $rprompt displayed immediately
adjacent to $prompt in your *telnet-HOST* buffer, and you've got to
account for that via telnet-prompt-regexp.

-- 
Kevin

RE: Q on using shell mode remotely

[Drew Adams]

    >     (eval-after-load "telnet"
    >        '(setq telnet-prompt-pattern
    >            (concat telnet-prompt-pattern
    >     "\\([12]?[0-9]:[0-5][0-9][ap]m\\)?")))
    >
    > I'll use `telnet-prompt-pattern', myself, to hack this problem,

    Please let us know if it works.

It works perfectly. Thanks, Kevin.

    Looking at telnet.el:

I didn't have time to track down why it works, but it does, so I'm happy.

    > but it would
    > be good if there were a variable that dealt with $rprompt somehow.

    I don't see how it could.  The remote csh presumably displays $rprompt
    on the right side of the screen via terminal escape codes or simple
    ASCII control characters; but the telnet subprocess is discarding that
    data, either explicitly via telnet-filter or implicitly via process-
    connection-type.  So you end up with $rprompt displayed immediately
    adjacent to $prompt in your *telnet-HOST* buffer, and you've got to
    account for that via telnet-prompt-regexp.

Yes, but if the telnet subprocess can discard the control characters or
whatever, then it could also, say, discard the $rprompt stuff. I'm not
saying it should do that systematically, but it might be good to have an
option to do so.

Anyway, thanks again.

BTW, I filed a bug, to see if we can't 1) make `telnet-prompt-pattern' into
a user option, like `shell-prompt-pattern', and 2) get more visibility for
some of the telnet-specific stuff like `telnet-prompt-pattern' in `C-h m'.
Dired mode describes options and commands, for instance, and that's quite
helpful.

RE: Q on using shell mode remotely

[Drew Adams]

    It works perfectly. Thanks, Kevin.

BTW, if anyone's interested, here is a brief description of my code that
lets you use substring and other regexp matches to search for and reuse
previous shell inputs (possibly modifying them, of course):
http://www.emacswiki.org/cgi-bin/wiki/Icicles_-_Other_Search_Commands#Icicle
sShell. And here is the code itself:
http://www.emacswiki.org/cgi-bin/wiki/Icicles_-_Libraries.

The same thing works for other interactive interpreters, besides shells; it
should work for any interpreter that uses `comint-mode'.

RE: Q on using shell mode remotely

[Drew Adams]

        Please let us know if it works.

    It works perfectly. Thanks, Kevin.

        Looking at telnet.el:

    I didn't have time to track down why it works, but it does, so
    I'm happy.

Actually, FYI, I realized belatedly that I was not using the vanilla Emacs
telnet.el - it doesn't seem work for me on Windows. I'm using some
Emacs-Lisp code by Zoltan Kemenczy, which works with a native Windows telnet
executable by Igor Milavec and Naftali Ramati. (I've had it for a few years,
but I don't remember trying it before.)

In hopes that it might help others (unless there is something better), I put
the Lisp code here: http://www.emacswiki.org/cgi-bin/wiki/telnet-contrib.el.
This also includes an Emacs telnet command by Ngai Kim Hoong that works with
the Cygwin telnet executable for Windows.

The Hoong code is from here: http://www.khngai.com/emacs/cygwin.php. Links
to the Milavec and Naftali-Ramati executable (sources) and the Kemenczy Lisp
code are here: http://www.gnu.org/software/emacs/windows/faq8.html#telnet.

I don't know if there is something better or more recent, and I don't know
why GNU Emacs doesn't also supply Windows executables and Lisp code that
work (as part of the distribution), but it doesn't, to my knowledge. If
someone knows better, please let me know - post an update at Emacs Wiki:
http://www.emacswiki.org/cgi-bin/wiki/TelnetMode.

HTH.