Thanks again.
Duncan Greatwood <dgreatwood@gmail.com> writes:
> Hi Michael and Eli -
Hi Duncan,
> sh-5.2$ uname --version
> uname (GNU coreutils) 8.32
> Copyright (C) 2020 Free Software Foundation, Inc.
> License GPLv3+: GNU GPL version 3 or later
> <https://gnu.org/licenses/gpl.html>.
> This is free software: you are free to change and redistribute it.
> There is NO WARRANTY, to the extent permitted by law.
>
> Written by David MacKenzie.
> sh-5.2$
>
> If I may, the extra characters in the emacs error message might not
> come from the uname on windows - that uname seems to work OK on the
> windows side at least - it could be a misparsing in emacs (or even an
> mistake gathering the error message, I suppose).
I also don't believe it comes from uname itself. '[17;120H’' looks
rather like an escape code sequence, perhaps emitted from the underlying
shell.
[DG] Yes, makes sense.
Something like cursor position, window setting, whatever. A
search didn't gave me a clue what's this.
After connecting the remote host via ssh, Tramp sends as very first
command something like
--8<---------------cut here---------------start------------->8---
# exec env TERM='dumb' INSIDE_EMACS='31.0.50,tramp:2.8.0-pre' ENV='' HISTFILE=~/.tramp_history PROMPT_COMMAND='' PS1=///245adade605348c086dac6d8f612435c\#\$ PS2='' PS3='' /bin/sh -i
--8<---------------cut here---------------end--------------->8---
Note the TERM='dumb' setting, which ought to suppress such code
sequences.
[DG] FWIW, I entered this command directly at the "sh" prompt on Windows, and it looks OK AFAIK:
sh-5.2$ exec env TERM='dumb' INSIDE_EMACS='31.0.50,tramp:2.8.0-pre' ENV='' HISTFILE=~/.tramp_history PROMPT_COMMAND='' PS1=///245adade605348c086dac6d8f612435c\#\$ PS2='' PS3='' /bin/sh -i
///245adade605348c086dac6d8f612435c#$
///245adade605348c086dac6d8f612435c#$echo $TERM
dumb
///245adade605348c086dac6d8f612435c#$echo $INSIDE_EMACS
31.0.50,tramp:2.8.0-pre
Somehow, this doesn't seem to work as expected in your case. Perhaps due
to calling /bin/sh. See below for debugging instructions.
> i) the target windows machine is not directly accessible right now, it
> is accessible via an intervening ssh proxy;
That's a valid point. So we need to find a solution.
> ii) SMB seems pretty retro to me in 2024, but maybe that's just me.
Yes, it is just you.
[DG] Fair enough.
On this basis, I configured a mount on the Linux machine of the Windows share using cifs/smb. fstab entry thus:
//WINDOWS-IP/WINDOWS-SHARE-NAME LINUX-MOUNT-DIR cifs credentials=CREDS-FOR-WINDOWS-STORED-IN-LINUX,rw,uid=1000,gid=1000,file_mode=0755,dir_mode=0755 0 0
(where 1000 is the UID and GID of my Linux user)
Then I used the normal tramp /ssh:... form to open LINUX-MOUNT-DIR and then open a file contained on the mount. I.e. With emacs running on macOS, this is going macOS -> (ssh) -> Linux -> (smb) -> Windows. And it seems to work just as you'd hope.
So this seems to be an option for the use case of access via an SSH proxy, provided the user has admin rights on the proxy (the Linux machine in this case) to install on the proxy a CIFS mount of the Windows share.
> From an emacs perspective, it seems a shame not to be able to use ssh,
> given that modern Windows commonly supports ssh and provides a bash
> shell. Depending on your own available bandwidth etc. of course.
That is NOT a shame. MS Windows is a non-free operation system,
therefore it isn't a primary target for Emacs development.
And it doesn't seem to be important. I'm working for Tramp for more than
20 years. During this time, nobody has contributed anything for the sake
of MS Windows. And I don't use MS Windows myself; if possible, I fix
things, that's it.
I, along with many others, appreciate all your efforts.
> BTW, emacs seems to be taking some time before finally producing an
> error message. If there is a way to log what is happening in emacs -
> what tramp is trying, what happens, etc., I'd be happy to. Or LMK how
> I could help otherwise.
Set tramp-verbose to 6 in a new Emacs session, prior the ssh
connection. When it has failed, there is a *debug tramp/ssh ...*
buffer; please send it as attachment.
[DG] Firstly, my apologies - I can't recreate the "uname parsing issue" directly now. Instead, emacs will go to a time out trying to access the Windows machine via the ssh proxy. I don't know what's changed.
In any case, I am enclosing the debug output file. It is 86MB, so I am sharing via Google Drive, hopefully that's OK.
This is with the Windows machine using /usr/bin/sh as the default shell for ssh login. Confirmed at the client command prompt - if I do "ssh WINDOWS", I log in and get the "sh" prompt.
> Best
> -DG
Best regards, Michael.