Scenario: I'm in zsh and I want to edit a long shell command in an already open instance of Emacs. I've set $VISUALto emacsclient -n -c. When I run C-x C-e from zsh, a new Emacs frame opens, populated with contents of the zsh line. After editing and saving the buffer, I hit C-x k, and buffer closes. The zsh line remains unchanged. Why isn't zsh recognizing the changes made in Emacs?

Emacs 24.5.1, zsh 5.0.8, Mac OS X 10.10.4


Set VISUAL to emacsclient -c (or some other variation without -n).

The option -n causes emacsclient to return as soon as it's contacted the running Emacs instance to tell it to edit the file. The program calling the editor (here, zsh, but this applies equally to any other program that invokes $VISUAL) knows that you've finished editing because the program that it invoked has exited, which, with -n, happens essentially immediately.

You need to arrange for the program invoked through VISUAL to exit when you've finished editing, i.e. when you close the buffer in Emacs. That's exactly what emacsclient without -n does: it contacts the running Emacs instance to tell it to open the file, then sits around until Emacs replies to tell emacsclient that the editing session is finished. The normal way to finish an editing session is to press C-x # (server-edit). This notifies emacsclient and closes the buffer, as well as the frame when it was opened by emacsclient -c. Closing the buffer also notifies the emacsclient process that the session is finished, but you get a warning prompt (“Buffer … still has clients, kill it?”), which you can disable by overriding the function server-kill-buffer-query-function (RMS thinks it's dangerous). In a frame opened by emacsclient -c, pressing C-x C-c (save-buffers-kill-terminal) saves the file, closes it and closes the frame.

| improve this answer | |
  • -n will also prevent emacsclient from working as an editor for git. – cg433n Jul 8 '15 at 22:07
  • A tangent (and I'll put it in another question if you want), but what uses does the -n flag have? – Squidly Jul 9 '15 at 9:45
  • 3
    @MrBones Say I'm in a file manager or on the command line and I want to edit a file. Not “edit a file and then do something with it”, just start editing it. If I ran a separate instance of Emacs per file, I'd run emacs myfile & (or the Windows equivalent), putting the Emacs process in the brackground. With emacsclient, I could run emacsclient myfile &, but that leaves an emacsclient process whose sole job is to wait, then exit. Instead I run emacsclient -n myfile, so the emacsclient process goes away by itself. This is so common I have a shell alias for it: alias e='emacsclient -n' – Gilles 'SO- stop being evil' Jul 9 '15 at 10:04
  • @Gilles I realised that I only ever run emacsclient in the terminal, which is why I couldn't see the point of -n. Thanks – Squidly Jul 9 '15 at 12:06
  • 2
    @MrBones, I would say running emacsclient in a terminal is the precise reason for using -n: after opening sending the file to the open Emacs frame, emacsclient terminates and one may continue to work in the terminal. This is exceptionally handy. – Daniel Jul 9 '15 at 18:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.