bug#25547: 25.1.91; emacsclient -c creates frames on the wrong display

[Eliza Velasquez via "Bug reports for GNU Emacs, the Swiss army knife of text editors" <bug-gnu-emacs@gnu.org>]

[-- Attachment #1: Type: text/plain, Size: 1406 bytes --]

Oops, forgot to reply all on my last message. Attached is the patch.

On Wed, Nov 4, 2020 at 8:44 PM Stefan Monnier <monnier@iro.umontreal.ca> wrote:
>
> > It looks like I was able to make progress today. Just by modifying
> > server.el, I've gotten `emacsclient -ce "..."` to create a new frame
> > based on the currently selected frame's terminal when called in a dumb
> > terminal.
>
> Great!
>
> > Unfortunately, the command for calling ediff is still very
> > unpleasant:
> >
> > emacsclient -ucF "((delete-frame-on-ediff-quit . t))" \
> >   -e "(ediff-merge-with-ancestor \"${local}\" \"${other}\" \"${base}\"
> > nil \"${output}\")" \
> >   -e "(add-hook 'ediff-quit-hook (lambda () (when (frame-parameter nil
> > 'delete-frame-on-ediff-quit) (delete-frame))))"
>
> I think the problem is that emacsclient lacks the equivalent of `-f`,
> i.e. calling a particular ELisp function where the remaining arguments
> can be consumed by that function.  That would solve the problem of
> quoting (your above script will stumble when encountering
> a file name with quotes in it or with a final backslash).
>
> > So perhaps introducing a wrapper script for ediff merges would still be useful.
>
> We should be able to add a convenience `batch-ediff-merge-with-ancestor`.
> Bonus points if you manage to make it work with `emacs --batch -f` as
> well as with `emacsclient --<newflag>`.
>
>
>         Stefan
>

[-- Attachment #2: 25547.patch --]
[-- Type: application/octet-stream, Size: 2286 bytes --]

diff --git a/lisp/server.el b/lisp/server.el
index a660deab8e..b6d0136d51 100644
--- a/lisp/server.el
+++ b/lisp/server.el
@@ -354,9 +354,11 @@ server-delete-client
 
       (setq server-clients (delq proc server-clients))
 
-      ;; Delete the client's tty, except on Windows (both GUI and console),
-      ;; where there's only one terminal and does not make sense to delete it.
-      (unless (eq system-type 'windows-nt)
+      ;; Delete the client's tty, except on Windows (both GUI and
+      ;; console), where there's only one terminal and does not make
+      ;; sense to delete it, or if we are explicitly told not.
+      (unless (or (eq system-type 'windows-nt)
+                  (process-get proc 'no-delete-terminal))
 	(let ((terminal (process-get proc 'terminal)))
 	  ;; Only delete the terminal if it is non-nil.
 	  (when (and terminal (eq (terminal-live-p terminal) t))
@@ -918,6 +920,19 @@ server-create-window-system-frame
            (server-send-string proc "-window-system-unsupported \n")
            nil))))
 
+(defun server-create-frame (nowait proc &optional parameters)
+  (add-to-list 'frame-inherited-parameters 'client)
+  (let ((frame (make-frame `((client . ,(if nowait 'nowait proc))
+                             ;; This is a leftover, see above.
+                             (environment . ,(process-get proc 'env))
+                             ,@parameters))))
+    (server-log (format "%s created" frame) proc)
+    (select-frame frame)
+    (process-put proc 'frame frame)
+    (process-put proc 'terminal (frame-terminal frame))
+    (process-put proc 'no-delete-terminal t)
+    frame))
+
 (defun server-goto-toplevel (proc)
   (condition-case nil
       ;; If we're running isearch, we must abort it to allow Emacs to
@@ -1264,6 +1279,10 @@ server-process-filter
 					   terminal-frame)))))
 		    (setq tty-name nil tty-type nil)
 		    (if display (server-select-display display)))
+                   ((equal tty-type "dumb")
+                    ;; Dumb terminals should create frames without
+                    ;; preference for tty or window system (bug#25547)
+                    (server-create-frame nowait proc frame-parameters))
 		   ((or (and (eq system-type 'windows-nt)
 			     (daemonp)
 			     (setq display "w32"))