Re: Emacs's set-frame-size can not work well with gnome-shell?

[Dmitry Gutov]

On 18.01.2020 5:05, Dmitry Gutov wrote:
> Alternatively, if we can't make resizing child frames work reliably, 
> maybe prohibit it on the API level? And then work on making sure that 
> deleting and re-creating a new child frame, and showing a buffer there, 
> is fast enough.

BTW, this already gives me a bit better performance than the original 
resize-test (0.7s for 10 iterations vs 1.0s for the same, and a lot of 
that 0.7 is spent in GC):

(defun open-test (buffer width height)
  (display-buffer-in-child-frame
   buffer `((child-frame-parameters
             . ((width . ,width)
                (height . ,height)
                (top . 50)
                (left . 50)
                )))))

(defun resize-test (frame height)
  (let ((width (frame-width))
        (buf (window-buffer (frame-first-window frame))))
    (delete-frame frame)
    (open-test buf width height)))

(setq-local test-buffer (get-buffer-create "test child-frame"))
(setq-local test-frame (window-frame (open-test test-buffer 40 10)))

(benchmark 10 '(setq-local test-frame (window-frame (resize-test 
test-frame 10))))

But an overlay-based popup is an order of magnitude faster still. And 
that's important because we really can't make the user wait for 70-100ms 
for Emacs to react to input.