bug#73446: 29.1; Curious behaviour with with-current-buffer and read-string

bug#73446: 29.1; Curious behaviour with with-current-buffer and read-string

[Al Haji-Ali]

Consider this code snippet:

,----
| (with-current-buffer (get-buffer "*scratch*")
|   (message "START point %S %S" (current-buffer) (point))
|   (goto-char 10)
|   (message "BEFORE point %S %S" (current-buffer) (point))
|   (with-current-buffer (get-buffer "*Messages*") ;; some random buffer
|     (read-string "Test: "))
|   (message "AFTER point %S %S" (current-buffer) (point)))
`----

Assuming that *scratch* has a point at something other than 10 and *is visible in a window*, I get this output in Messages

,----
| START point #<buffer *scratch*> 2952
| BEFORE point #<buffer *scratch*> 10
| AFTER point #<buffer *scratch*> 2952
`----

I was expecting (point) AFTER to be 100, as it was BEFORE.

This only happens when `read-string` is called and when `with-current-buffer` is nested, and only when *scratch* is visible in a window.

I can easily solve this issue with `save-excursion`, but I thought this might be an unintended behaviour worth reporting.

Best regards,
-- Al

bug#73446: 29.1; Curious behaviour with with-current-buffer and read-string

[Eli Zaretskii]

> From: Al Haji-Ali <abdo.haji.ali@gmail.com>
> Date: Tue, 24 Sep 2024 09:19:56 +0100
> 
> 
> Consider this code snippet:
> 
> ,----
> | (with-current-buffer (get-buffer "*scratch*")
> |   (message "START point %S %S" (current-buffer) (point))
> |   (goto-char 10)
> |   (message "BEFORE point %S %S" (current-buffer) (point))
> |   (with-current-buffer (get-buffer "*Messages*") ;; some random buffer
> |     (read-string "Test: "))
> |   (message "AFTER point %S %S" (current-buffer) (point)))
> `----
> 
> Assuming that *scratch* has a point at something other than 10 and *is visible in a window*, I get this output in Messages
> 
> ,----
> | START point #<buffer *scratch*> 2952
> | BEFORE point #<buffer *scratch*> 10
> | AFTER point #<buffer *scratch*> 2952
> `----
> 
> I was expecting (point) AFTER to be 100, as it was BEFORE.
> 
> This only happens when `read-string` is called and when `with-current-buffer` is nested, and only when *scratch* is visible in a window.

I can't reproduce this, but see switch-to-buffer-preserve-window-point.

bug#73446: 29.1; Curious behaviour with with-current-buffer and read-string

[Al Haji-Ali]

> I can't reproduce this, but see switch-to-buffer-preserve-window-point.

Oh, is it maybe emacsen specific?

I have `switch-to-buffer-preserve-window-point` set to t, but setting it to `nil` gives the same result. Here's how I can reproduce the problem:

1. Open 'emacs -Q'.
2. Paste the following into the scratch buffer

(progn  
  (defun test-bug-73446 nil    
    (let ((buf1 "test1")
          (buf2 "test2"))
      (switch-to-buffer (get-buffer-create buf1))
      (insert (make-string 20 ?H))
      (goto-char 1)
      (switch-to-buffer (get-buffer-create buf2))
    
      (let ((switch-to-buffer-preserve-window-point t)
            start before after)
        (display-buffer buf1 'display-buffer-use-some-window)
        
        (with-current-buffer (get-buffer buf1)
          (setq start (point))
          (goto-char 10)
          (setq before (point))
          (with-current-buffer (get-buffer buf2) ;; some random buffer
            (read-string "Test: "))
          (setq after (point)))
        (list start before after))))
  (message "%S" (test-bug-73446)))

3. Press `C-x c-e` (or call `eval-last-sexp`) 
3. Press return when the prompt "Test: " appears.
4. I see the output (1 10 1), instead of (1 10 10) as I would expect.

Best regards,
-- Al

bug#73446: 29.1; Curious behaviour with with-current-buffer and read-string

[Eli Zaretskii]

> From: Al Haji-Ali <abdo.haji.ali@gmail.com>
> Cc: 73446@debbugs.gnu.org
> Date: Tue, 24 Sep 2024 14:12:43 +0100
> 
> 
> 1. Open 'emacs -Q'.
> 2. Paste the following into the scratch buffer
> 
> (progn  
>   (defun test-bug-73446 nil    
>     (let ((buf1 "test1")
>           (buf2 "test2"))
>       (switch-to-buffer (get-buffer-create buf1))
>       (insert (make-string 20 ?H))
>       (goto-char 1)
>       (switch-to-buffer (get-buffer-create buf2))
>     
>       (let ((switch-to-buffer-preserve-window-point t)
>             start before after)
>         (display-buffer buf1 'display-buffer-use-some-window)
>         
>         (with-current-buffer (get-buffer buf1)
>           (setq start (point))
>           (goto-char 10)
>           (setq before (point))
>           (with-current-buffer (get-buffer buf2) ;; some random buffer
>             (read-string "Test: "))
>           (setq after (point)))
>         (list start before after))))
>   (message "%S" (test-bug-73446)))
> 
> 3. Press `C-x c-e` (or call `eval-last-sexp`) 
> 3. Press return when the prompt "Test: " appears.
> 4. I see the output (1 10 1), instead of (1 10 10) as I would expect.

If this is what you do, then it's the expected behavior: test1 is
displayed in another window, and so its window-point overrides what
you do.  See the node "Window Point" in the ELisp manual.

This is not a bug.

bug#73446: 29.1; Curious behaviour with with-current-buffer and read-string

[Al Haji-Ali]

On 24/09/2024, Eli Zaretskii wrote:
> If this is what you do, then it's the expected behavior: test1 is
> displayed in another window, and so its window-point overrides what
> you do.  See the node "Window Point" in the ELisp manual.
>
> This is not a bug.

Ah I see. It is interesting that even if the window displaying the
buffer is shown but seemingly not selected (since the focus is on the
minibuffer during `read-string`), the point gets reset to the
window-point when Emacs waits for user input.

I vote to close this issue and apologies for the noise.

Best regards,
-- Al

bug#73446: 29.1; Curious behaviour with with-current-buffer and read-string

[Eli Zaretskii]

> From: Al Haji-Ali <abdo.haji.ali@gmail.com>
> Cc: 73446@debbugs.gnu.org
> Date: Tue, 24 Sep 2024 16:07:42 +0100
> 
> 
> On 24/09/2024, Eli Zaretskii wrote:
> > If this is what you do, then it's the expected behavior: test1 is
> > displayed in another window, and so its window-point overrides what
> > you do.  See the node "Window Point" in the ELisp manual.
> >
> > This is not a bug.
> 
> Ah I see. It is interesting that even if the window displaying the
> buffer is shown but seemingly not selected (since the focus is on the
> minibuffer during `read-string`), the point gets reset to the
> window-point when Emacs waits for user input.
> 
> I vote to close this issue and apologies for the noise.

Thanks, I'm closing it.  But there's no need to apologize: Emacs is a
very complex beast, and no one knows all of its subtleties by heart.