w32 ssh trouble

w32 ssh trouble

[Bruce Ingalls]

I am trying to get TRAMP to work under w32.
Note that Linux works fine for me.

I tried executing M-x shell, then ssh.
This fails, with the error
Pseudo-terminal will not be allocated because stdin is not a terminal.

Why am I getting this error, and what can I do about it?

I also tried running ftp under the shell, but Emacs hangs, no error.

I do not have these shell problems with w32 XEmacs.
However, I have to figure out how the tramp syntax works.

I am running the latest cygwin zsh as my default shell, latest tramp 
v2.0.33 and Emacs v21.3.1

Re: w32 ssh trouble

[Jason Rumney]

Bruce Ingalls <bingalls.NO_SPAM@fit-zones.com> writes:

> I am trying to get TRAMP to work under w32.
> Note that Linux works fine for me.
> 
> I tried executing M-x shell, then ssh.
> This fails, with the error
> Pseudo-terminal will not be allocated because stdin is not a terminal.
> 
> Why am I getting this error, and what can I do about it?

You are getting the error because the version of ssh you are using
expects a certain method of faking Unix pttys (I could take a wild
guess here and say that you're using a Cygwin port of ssh).

You could try using an alternative ssh client. plink (the putty
command-line client) seems like it could work if you can get comint
to recognize the passphrase prompt.

> I also tried running ftp under the shell, but Emacs hangs, no error.

Cygwin ftp? Same problem perhaps.

> I do not have these shell problems with w32 XEmacs.

Cygwin version? You would probably get the ptty emulation that Cygwin
expects if you used the Cygwin port of Emacs as well.

Re: w32 ssh trouble

[Bruce Ingalls]

OK.
Thanks to Kai's help on XEmacs notation, I can noew try with XEmacs,
and discover that it works almost the same as Emacs (in this regard).

I tried plink v0.53b (should be latest binary release), and I was
able to get the prompt for the password, when I run it inside M-x shell.
However, it does not recognize my password, even though I know that I
am entering it correctly.
I was able to get the following to work inside M-x shell. I don't think
it matters, whether I use Cygwin or Cmd.exe for the child shell:

plink -ssh -batch -pw mypassword user@domain.com

This is less secure, as  users can spy on `ps`, but that's not an issue 
on my single user machine. I'm not sure if this is a security problem on 
Windows in general, as I don't know of a way to look at command line 
args to running processes, even if I were to connect my machine in a 
public area, leave it logged in, and walk away, or allow Administrators 
to snoop (via e.g. RDP), for that matter.

As far as Emacs versions:
It looks that I was using native Emacs. This is the only binary widely
distributed. I could not find a version of Emacs, which runs on the w32
desktop, yet supports Cygwin.

I did discover that running the Cygwin Emacs package inside the Cygwin
shell does work.
There is still an annoying bug, where Cygwin Emacs does not recognize 
C-x C-c to quit Emacs, and I must use the F10/M-' menu.
This is a bug in the port of Emacs to Cygwin.

The version of XEmacs I work with, runs on the w32 desktop, but is
compiled with Cygwin.

Finally, I tried plink under both Cygwin & cmd shells, and it recognized 
my password the first time. Thus, M-x shell in Windows is making some 
kind of change to my password.

Jason Rumney wrote:
> Bruce Ingalls <bingalls.NO_SPAM@fit-zones.com> writes:
> 
>>I am trying to get TRAMP to work under w32.
>>Note that Linux works fine for me.
>>
>>I tried executing M-x shell, then ssh.
>>This fails, with the error
>>Pseudo-terminal will not be allocated because stdin is not a terminal.
>>
>>Why am I getting this error, and what can I do about it?
> 
> You are getting the error because the version of ssh you are using
> expects a certain method of faking Unix pttys (I could take a wild
> guess here and say that you're using a Cygwin port of ssh).
> 
> You could try using an alternative ssh client. plink (the putty
> command-line client) seems like it could work if you can get comint
> to recognize the passphrase prompt.
> 
>>I do not have these shell problems with w32 XEmacs.
> 
> Cygwin version?

1.3.22 - Everything is latest binary release

Re: w32 ssh trouble

[Jason Rumney]

Bruce Ingalls <bingalls.NO_SPAM@fit-zones.com> writes:

> I tried plink v0.53b (should be latest binary release), and I was
> able to get the prompt for the password, when I run it inside M-x shell.
> However, it does not recognize my password, even though I know that I
> am entering it correctly.

Right. I got the same when I tried, that's why I mentioned making
comint recognize that plink is prompting for a password/passphrase so
it treats it differently. Perhaps that would not solve the problem
though. Plink works for me with certificate authentication and
pageant running (so it doesn't prompt for anything).

> As far as Emacs versions:
> It looks that I was using native Emacs. This is the only binary widely
> distributed. I could not find a version of Emacs, which runs on the w32
> desktop, yet supports Cygwin.

Right, the Cygwin version only supports tty and X. I know of no plans
to make it support the native Windows GUI.

> Finally, I tried plink under both Cygwin & cmd shells, and it
> recognized my password the first time. Thus, M-x shell in Windows is
> making some kind of change to my password.

Maybe (line end characters perhaps). Or maybe the way that plink
reads its password is not compatible with the indirection of running
a shell within Emacs.

Re: w32 ssh trouble

[Kai Großjohann]

Bruce Ingalls <bingalls.NO_SPAM@fit-zones.com> writes:

> I tried plink v0.53b (should be latest binary release), and I was
> able to get the prompt for the password, when I run it inside M-x shell.
> However, it does not recognize my password, even though I know that I
> am entering it correctly.

What happens when you use M-: (process-send-string nil "your-pw\n")
RET to send the password?  What happens when you use \r instead of
\n?  What happens when you use \r\n instead of \n?
-- 
This line is not blank.

Re: w32 ssh trouble

[Bruce Ingalls]

Kai Großjohann wrote:
> Bruce Ingalls <bingalls.NO_SPAM@fit-zones.com> writes:
> 
> 
>>I tried plink v0.53b (should be latest binary release), and I was
>>able to get the prompt for the password, when I run it inside M-x shell.
>>However, it does not recognize my password, even though I know that I
>>am entering it correctly.
> 
> 
> What happens when you use M-: (process-send-string nil "your-pw\n")
> RET to send the password?  What happens when you use \r instead of
> \n?  What happens when you use \r\n instead of \n?

Well, It's a bit hard to do this, in the middle of a password prompt.
It tries to read in M-: (whatever control char that generates) as my
password.
I did try to do a C-g (a second time, after I got a "Non echoed text" 
prompt) and I noticed this phenomenon:
The shell seems to chop off the last 2 characters of my input.

Thus, "your-pw\r\n" would work best, as the \r and \n will get chopped 
out. In other words, this seems to work, but not for the reasons you 
expected.

Re: w32 ssh trouble

[Kai Großjohann]

Bruce Ingalls <bingalls.NO_SPAM@fit-zones.com> writes:

> Kai Großjohann wrote:
>> Bruce Ingalls <bingalls.NO_SPAM@fit-zones.com> writes:
>> 
>>>I tried plink v0.53b (should be latest binary release), and I was
>>>able to get the prompt for the password, when I run it inside M-x shell.
>>>However, it does not recognize my password, even though I know that I
>>>am entering it correctly.
>> What happens when you use M-: (process-send-string nil "your-pw\n")
>> RET to send the password?  What happens when you use \r instead of
>> \n?  What happens when you use \r\n instead of \n?
>
> Well, It's a bit hard to do this, in the middle of a password prompt.
> It tries to read in M-: (whatever control char that generates) as my
> password.

Hm.  Well, you can remove comint-watch-for-password-prompt from
comint-output-filter-functions.  That ought to remove the minibuffer
prompt and thus you can do M-: normally.

> I did try to do a C-g (a second time, after I got a "Non echoed text"
> prompt) and I noticed this phenomenon:
> The shell seems to chop off the last 2 characters of my input.

Huh?  That's weird.

I could put stuff in Tramp that adds junk to passwords.  (Are we
talking about Tramp?)  Surely I would do this if it makes users
happy.  But it makes me feel REALLY REALLY uneasy.  *shudder*
-- 
This line is not blank.

Re: w32 ssh trouble

[Bruce Ingalls]

Kai Großjohann wrote:
> Bruce Ingalls <bingalls.NO_SPAM@fit-zones.com> writes:
>
>>>>I tried plink v0.53b (should be latest binary release), and I was
>>>>able to get the prompt for the password, when I run it inside M-x shell.
>>>>However, it does not recognize my password, even though I know that I
>>>>am entering it correctly.
>>>
>>>What happens when you use M-: (process-send-string nil "your-pw\n")
>>>RET to send the password?  What happens when you use \r instead of
>>>\n?  What happens when you use \r\n instead of \n?
>>
>>Well, It's a bit hard to do this, in the middle of a password prompt.
>>It tries to read in M-: (whatever control char that generates) as my
>>password.
> 
> Hm.  Well, you can remove comint-watch-for-password-prompt from
> comint-output-filter-functions.  That ought to remove the minibuffer
> prompt and thus you can do M-: normally.
> 
>>I did try to do a C-g (a second time, after I got a "Non echoed text"
>>prompt) and I noticed this phenomenon:
>>The shell seems to chop off the last 2 characters of my input.
> 
> Huh?  That's weird.
> 
> I could put stuff in Tramp that adds junk to passwords.  (Are we
> talking about Tramp?)  Surely I would do this if it makes users
> happy.  But it makes me feel REALLY REALLY uneasy.  *shudder*

I tried the above, and yes, it works.
I tried appending any 2 chars to the end, and this worked.

We are not talking about Tramp, yet. I figure that I have to get plink 
(ssh) working in M-x shell, before we can talk about Tramp.

Now that I can get plink working under M-x shell, I tried adding garbage 
to the end of my password in Tramp, but that did not work.
Unfortunately, this is not a good indicator, as Plink only works for me 
via protocol 1, and I'll have to fool with Tramp settings for this.

I understand your unease. The other solution is to pass in passwords, 
using Plink's -pw command line option.

Re: w32 ssh trouble

[Kai Großjohann]

Bruce Ingalls <bingalls@SpAmBlOcK.fit-zones.com> writes:

> I tried the above, and yes, it works.
> I tried appending any 2 chars to the end, and this worked.

Amazing.

> We are not talking about Tramp, yet. I figure that I have to get plink
> (ssh) working in M-x shell, before we can talk about Tramp.

Good idea.

> Now that I can get plink working under M-x shell, I tried adding
> garbage to the end of my password in Tramp, but that did not work.
> Unfortunately, this is not a good indicator, as Plink only works for
> me via protocol 1, and I'll have to fool with Tramp settings for this.

What's the command line you use to tell plink to use protocol 1?
I'll add an entry plink1 to tramp-methods for you.

> I understand your unease. The other solution is to pass in passwords,
> using Plink's -pw command line option.

Whee.  That doesn't seem better.
-- 
This line is not blank.

Re: w32 ssh trouble

[Bruce Ingalls]

Kai Großjohann wrote:
> Bruce Ingalls <bingalls@SpAmBlOcK.fit-zones.com> writes:

>>I tried appending any 2 chars to the end, and this worked.
> 
> Amazing.
> 
>>Now that I can get plink working under M-x shell, I tried adding
>>garbage to the end of my password in Tramp, but that did not work.
>>Unfortunately, this is not a good indicator, as Plink only works for
>>me via protocol 1, and I'll have to fool with Tramp settings for this.
> 
> What's the command line you use to tell plink to use protocol 1?
> I'll add an entry plink1 to tramp-methods for you.

-1 just like ssh.
> 
>>I understand your unease. The other solution is to pass in passwords,
>>using Plink's -pw command line option.
> 
> Whee.  That doesn't seem better.

I forgot to mention, that you don't need to append garbage, when passing the
password on the command line.
BTW, I tested plink with both cygwin & the native cmd.exe shells, and they
behave equally. I still must retest with XEmacs, but I don't expect 
different
behavior.

Re: w32 ssh trouble

[Kai Großjohann]

Bruce Ingalls <bingalls@CUT-this-SPAM-BLOCK.fit-zones.com> writes:

> Kai Großjohann wrote:
>> Bruce Ingalls <bingalls@SpAmBlOcK.fit-zones.com> writes:
>
>>>I tried appending any 2 chars to the end, and this worked.
>> Amazing.
>> 
>>>Now that I can get plink working under M-x shell, I tried adding
>>>garbage to the end of my password in Tramp, but that did not work.
>>>Unfortunately, this is not a good indicator, as Plink only works for
>>>me via protocol 1, and I'll have to fool with Tramp settings for this.
>> What's the command line you use to tell plink to use protocol 1?
>> I'll add an entry plink1 to tramp-methods for you.
>
> -1 just like ssh.

Great.  I've now added the entry.  New version in CVS.

So if I understand you correctly, you just have to append two random
chars to your password, and then it will work, right?

Whee.  I wonder what's going to happen.  It's really weird.

>>>I understand your unease. The other solution is to pass in passwords,
>>>using Plink's -pw command line option.
>> Whee.  That doesn't seem better.
>
> I forgot to mention, that you don't need to append garbage, when
> passing the password on the command line.

Understood.

Hm.  Does it work to set tramp-password-end-of-line to "xx\n"?  It's
a really really really dirty trick.  But who knows...

Alas, Tramp is not flexible enough to allow different values of
tramp-password-end-of-line for different methods.
-- 
This line is not blank.

Re: w32 ssh trouble

[Kai Großjohann]

Bruce Ingalls <bingalls.NO_SPAM@fit-zones.com> writes:

> I do not have these shell problems with w32 XEmacs.
> However, I have to figure out how the tramp syntax works.

/[user@host]path/to/file
/[method/user@host]path/to/file

The square brackets are literally part of the filename syntax.

I have preliminary changes for using the unified syntax under XEmacs,
but I don't know how well or badly it works -- I don't normally use
XEmacs, and it seems no XEmacs developer has tested it.

-- 
This line is not blank.