Re: emacs as a service

[B. T. Raven]

rustom wrote:
> This is about tweaking emacs to run (somewhat like) a service.
> Environment is emacs22 on win-XP.
>
> I describing my current configuration with questions here and there in
> the hope that some (better) solutions will come out.
>
> Normally emacs runs with 1 or more windows (frames in emacsspeak)
> multiple emacs processes have no relation wilth each other.
>
> So in sort the typical setup is:
> 0 or more emacses each having 1 or more frame.
>
> By a server setup I mean the following (which may not be proper usage
> from the os service point of view)
> 0 or 1 emacs having 0 or more frames.
>
> So first thing I did
> --------------
> ; from http://www.emacswiki.org/cgi-bin/wiki/RupertSwarbrick#toc1
>  (defun buffer-used-elsewhere-p (&optional buf)
>    "Returns true if more than one window is attached to BUF in all
>  frames. If buf is nil or not supplied, search
>  for (CURRENT-BUFFER)."
>    (unless buf (setq buf (current-buffer)))
>    (< 1 (length (get-buffer-window-list buf nil t))))
>
>  (defun delete-frame-ex ()
>    "Delete frame as normal, unless we're on the last frame, in
>    which case, exit emacs as we would normally do"
>    (interactive)
>    (if (eq (selected-frame) (next-frame (selected-frame) 0))
>        ;; This is the last frame
>        (save-buffers-kill-emacs)
>     ;; There's more!
>      (unless (buffer-used-elsewhere-p) (kill-buffer nil))
>      (delete-frame)))
>
> (global-set-key "\C-x\C-c" 'delete-frame-ex)
> -----------------
>
> This basically makes C-x C-c kill frames rather than kill emacs except
> for the last one in which case it kills emacs.
>
> Then recently got some solutions from Scott Turner that prompted me to
> try again:
>
> I changed the line
>        (save-buffers-kill-emacs)
> above to
>        (make-frame-invisible (car (frame-list)) t)
> which makes the last frame behavior consistent with the others (and
> emacs unkillable unless one uses explicit M-x kill-emacs).  Basically
> emacs disappears rather than dying.
>
> Q: Obviously the (car (frame-list)) thingy is clumsy. Any better way?
> Also I want to bind the X on the window top right to function delete-
> frame-ex
> rather than the default save-buffer-kill-emacs. How to do that?
>
> To move to a more client-server oriented model.
> The default emacs program that runs is to be replaced by this vbs
> script
> ------------
> Set objShell = WScript.CreateObject("WScript.Shell")
> Set fso = CreateObject("Scripting.FileSystemObject")
>
> strComputer = "."
>
> Set objWMIService = GetObject("winmgmts:\\" & strComputer & "\root
> \cimv2")
>
> Set colItems = objWMIService.ExecQuery("Select * From Win32_Process")
> Dim isRunning
> isRunning = False
>
> For Each objItem in colItems
>   If InStr(objItem.CommandLine, "emacs.exe") Then
>     isRunning = True
>   End If
> Next
>
> parent = fso.GetParentFolderName(WScript.ScriptFullName)
> If WScript.Arguments.Count = 1 Then
>   If isRunning Then
>     callString = parent & "/emacsclientw.exe -n """ & WScript.Arguments
> (0) & """"
>   Else
>     callString = parent & "/runemacs.exe """ & WScript.Arguments(0) &
> """"
>   End If
> Else
>   If isRunning Then
>     callString = parent & "/emacsclientw.exe -n -e ""(raise-frame)"""
>   else
>     callString = parent & "/runemacs.exe"
>   End If
> End if
> objShell.Run(callString)
> -------------
> Q: Whats the diff between emacs and runemacs?

I think that runemacs is just a wrapper used to suppress the appearance 
of the command window. It is after all only 217kB in size.

> Q: The for loop is ugly but my windows/vbs is out of depth here
>
> This script basically does a 4=2x2 way check:
> Is emacs server running or not?
> Is the script called with an argument or not?
> And the default bindings for emacs-ish files (in my case .txt, .org)
> is bound to this script in the registry.
>
> If someone can offer HKEY magic to automate this that'd be nice.

I'm not sure I understand your problem or how it differs from just 
putting emacsclientw in the Windows Registry. You don't need VB to do that:

(server-start)
goes in .emacs (Emacs version 22.1)

ALTERNATE_EDITOR=C:\emacs\bin\runemacs.exe
goes in environment

in HKCR hive
*/shell/open/command
default c:\emacs\bin\emacsclientw.exe -n "%1"

Applications/emacsclientw.exe/shell/open/command
default c:\emacs\bin\emacsclientw.exe -n "%1"

I think that these will propagate to the HKLM hive on startup but beware 
of manual registry surgery. At one point, when I clicked on files 
associated with Emacs in Windows Explorer, all that would happen was 
that the runemacs.exe binary would load into an emacs buffer, a 
consummation not devoutly to be wished for. Also, it's safest to make 
associations from WE with Tools>Folder Options>File Types rather than in 
the registry but you may need ./ (also in HKCR) set to something like 
ft000003, ft000002, or ft000001 for click-opening extensionless files 
(e.g. readme). I think these are automatically generated user file types 
but I can't find any documentation on them. Again beware; some system 
files don't have extensions , even parts of the registry itself.
See:
http://www.emacswiki.org/emacs/EmacsClient#toc6

> Now one fragile aspect of this is that if 2 emacsservers run well the
> system is borked!  How is the (server-start) to be enclosed in a
> (if server-not-started-p ...) ?

Is that... run well or run, well.... ;-)
> And one (entirely windowsy) related question:
> How to catch the system login, logout, shutdown etc events?
> Once I can get some script to give me some control for this I can use
> emacsclient to do appropriate actions.

I am totally out of my depth here. Don't MSwin processes with pending 
open files warn about needed actions in case of logout or shutdown?

> Thanks (if you read so far!)