wp.jochen.hayek.name/blog-en

Category: ssh

  • emacs/tramp/win/putty: “Fatal: Received unexpected end-of-file from server”

    Short story – what (released) versions to choose:

    The combination emacs/tramp/putty on Windows does not work with the released PuTTY-0.63. Use either PuTTY-0.62 or a PuTTY development snapshot, r10016 works for me.

    • http://www.gnu.org/software/emacs/windows/ – the latest released version was fine for me: 24.3.1 – I am always happy to get the very latest released version
    • tramp – as packaged with the emacs at the location just referred to
    • PuTTY to be found at http://www.chiark.greenend.org.uk/~sgtatham/putty/download.html
    • PuTTY-0.63 faults with accesses as shown below
    • PuTTY’s development snapshot does not fault for me, as 0.63 does – r10156 is, what I picked up – I am happy to try newer ones, if asked to
    • Windows 7 Professional – because that’s the current company standard

    Long story:

    After the Windows upgrade at work (WinXP->Win7) I certainly installed the newest available emacs (i.e. emacs package) and PuTTY (0.63) as well.

    Current issue: On opening a file accessed through “/pscp:SESSIONNAME:”, I keep getting this message:

    Fatal: Received unexpected end-of-file from server

    I tried the “/plink:SESSIONNAME:” tramp access instead, and it worked (I mean: it didn’t yield this message), but the combination tramp/plink keeps leading to corrupted (remote) files, as I described before [link], so that’s not really an option.

    S.K.R. de Jong wrote [link], that the quoted message stems from an improper reaction of PuTTY, and Simon Tatham conceded the bug [link], writing the bug would be fixed with r10016.

    Because I  found Simon’s statement and announcement only during my later investigations, I decided to give the older PuTTY-0.62 a try, and this problem actually disappeared. Of course I love leading / bleeding edge, but right now this seems to stand in my way. So I will have to post the scenario on some PuTTY forum / mailing list, but until that will lead to an improvement here, I shall stick to PuTTY-0.62.

    When I had found Simon Tatham statement on this issue [link], I drew new hope, that a more modern version of PuTTY addressing the latest security issues  would also solve my problem – so I am giving the current development snapshot (to be found here) a chance, i.e. “r10156”. Actually I have never been eager to employ a development snapshot of PuTTY, so that is my “first“. I subscribed for the PuTTY-announce mailing list, so I shall be able to replace the development snapshot by a properly released one.

    If that will not successful:

  • Wednesday afternoon / evening is my time scheduled for extended technical read-ups

    But once again serious issues at work were in my way.
    This time my work PC’s WinXP->Win7 migration was the issue. emacs/tramp/putty in its newest (released) versions still did not want to work together as neatly as before. I will have to make use of putty-0.62 instead of 0.63, but the differences do not seem to hurt me right now. I do have a proper work set-up again (after like 3 days with lots of interrupts), but in the meantime (apart from some real work) I also improved a few issues with wiki articles and described a few Win7 migration related issues, that may serve a few colleagues quite well.

  • startup PuTTY’s authentication agent “Pageant” with private keys

    PuTTY’s authentication agent pageant combines the functionality of ssh-agent and ssh-add, and if you start it up, it makes sense to add the necessary private keys on its command line. You can do this using a script, but in Windows it’s convenient to does this through a “.lnk” shortcut. And if you place that “.lnk” file in Windows’ user startup directory, you won’t forget to start it manually.

    Here is a nice description of how to make “… Pageant automatically load keys on startup“:

    Basically:

    Recipe (Win <10):

    • within your Windows auto-start-up folder (Start > All Programs > Startup — right-click > Open) …
    • create a Windows “shortcut” to Pageant.exe, it gets called pageant.exe.lnk!
    • edit the shortcut’s properties!
    • tell the command line, where to “Start in:“! %USERPROFILE%\AppData\Roaming\putty or %USERPROFILE%\etc\putty (that’s where you created resp. left your private keys using PuTTYgen)
    • add the private keys on the command line within “Target:” (YOUR.ppk)– and the paths may be relative to the directory you mentioned in “Start in:“

    Recipe (…):

    • %USERPROFILE%/AppData/Roaming/Microsoft/Windows/Start Menu/Programs/Startup/ : Pageant.lnk
    • …

    It’s helpful to dump the details of pageant.exe.lnk using one of these nice tools:

    PuTTY’s authentication agent “Pageant” has its counterparts in: ssinf h-agent + ssh-add

  • emacs/tramp/win/putty: “byte-code: couldn’t find a proper ‘ls’ command”

    The PC at “my” desk at my customer’s site just got migrated from WinXP to win7, and I have to re-install and re-configure a couple of packages, and now I have trouble re-adopting tramp.

    Having to deal with tramp always brings Michael Albinus to my mind. I only met him during one week in 1988 (1989?), when I had only lived for a couple of months in Berlin. It was quite an enthusiastic time then, and we were so much younger then. Would he let me invite him on a couple of glasses of rather drinkable red wine in the future back home in Berlin?

    Alright, back to the issue itself! Yes, after the Windows upgrade I certainly installed the newest available emacs (i.e. emacs package) and PuTTY (0.63) as well.

    • byte-compile …/emacs/lisp/net/tramp-sh.el
    • exit emacs
    • remove ~/.emacs.d/tramp, i.e. the “tramp connection history” resp. the “tramp connection details cache“
    • start emacs

    and it seemed to find a proper “ls” command again.

    I.e. I didn’t follow the other advise, demanding the installation of the newest tramp package.

    Having solved the “ls” issue, I faced these messages:

    Unable to use key file "…id_rsa.ppk" (unable to open file)
Fatal: Received unexpected end-of-file from server

    The windows registry had a couple of entries named “PublicKeyFile” pointing to an outdated location, where id_rsa.ppk had resided before. Changed all occurrences of that entry within the registry (subtree of PuTTY), and that issue was solved.

  • emacs/tramp/win/putty: “File exists, but cannot be read”

    I had a look here and look there for hours, … – accidentally I looked into the *Messages* buffer and found, that some file access within c:windowstemp failed. The system variables TEMP and TMP are initialized with c:windowstemp, but I have no rights to look into c:windowstemp. Alright, system variables can get overridden. And they actually were overridden. I just changed that to something very, very simple: c:temp. I restarted emacs, and everything was alright again.

  • RoR deploying with Mina, a “really fast deployer and server automation tool”

  • if saving a file through “tramp” is not successful, ending “PLINK.EXE” may be a solution

    PLINK.EXE is part of the PuTTY family, and it’s the utility tramp is communicating through.

    “Burying” the respective tramp buffer in emacs is another one.

    Update 2016-02-09:

    Actually burying the tramp buffer in question is my 1st approach on all platforms (Linux/Mac/Windows), whenever tramp communication seems to be stuck. Afterwards (with a “clean new start”) everything is fine again in like 99% of the respective the cases.

  • Emacs and VC AKA Version Control and an excessive list of handled backends

    I am working for a customer

    • on WinXP
    • with GNU Emacs 24.2.1
    • through Tramp
    • resp. putty plink (“pscp”)
    • on AIX files.

    They are using CVS on AIX, but as I am actually the only Emacs user there, I don’t really bother going through the set-up for using Emacs with CVS on WinXP.

    Recently I looked into *tramp/pscp HOSTNAME*, and I decided I would be able cutting down on my waiting time, if I drastically reduced the list of files and directories being checked through VC in that environment.

    I really don’t mind the VC overhead on OS X or (SUSE) Linux, but in that difficult environment I rather want to proceed as described.

    I found vc-handled-backends and customised it the Emacs way, and I was done.

  • O’Reilly Media book: Linux Server Hacks, Volume One

    It’s not officially called “Volume One”, but since there also appeared a “Volume Two”, it seems to make sense to call this one “Volume One”, right?

    The recipes, you can find here, are also applicable generally in the Unix world.

    This is roughly how you start the ssh-agent, and how your add private keys:

    $ ssh-agent …
$ ssh-add …

    See also:

  • MTPuTTY = Multi-Tabbed PuTTY