Visual indicator not always correct

[098799]

Visual indicator (red cursor) is not working on startup of emacs. Also when switching/killing buffers I frequently mistakenly get white cursor for when I'm in fact in the ryo mode.

One clear example is:

• I'm in ryo mode (cursor is red).
• I enter magit popup, my cursor is white and I'm not in ryo mode (correct).
• I press q to leave the magit popup and I'm back in ryo mode, but incorretly with white cursor.

My use-package config hooks regarding which modes have ryo enabled by default are as follows (and I don't want to add magit popup to that as it has it's own set of single letter shortcuts):

:config (add-hook 'text-mode-hook #'ryo-modal-mode) (add-hook 'prog-mode-hook #'ryo-modal-mode) (add-hook 'fundamental-mode-hook #'ryo-modal-mode) (add-hook 'conf-unix-mode-hook #'ryo-modal-mode)

Thank you for all the work you've put into developing this package. I'm using it for a few months now and my config grew up to 150 lines of shortcuts. I like it a lot and I really don't feel like switching, but not knowing if I'm in the ryo mode or not at a glance makes it sometimes cumbersome to use.

[Kungsgeten]

Hi! Sorry that I haven't replied here, must have forgotten about it.

To be honest I don't know why the visual indicator sometimes doesn't work. The check for it is used in post-command-hook which is run almost no matter what you do.

In the case of Magit I've done some research. The function magit-restore-window-configuration ends with the following code:

(when (and winconf (equal frame (window-configuration-frame winconf)))
      (set-window-configuration winconf)
      (when (buffer-live-p buffer)
        (with-current-buffer buffer
          (setq magit-previous-window-configuration nil))))

If I try and comment out this, it seems like the cursor color switching works. @tarsius Do you know why magit-restore-window-configuration doesn't seem to call post-command-hook when it is done?

[tarsius]

No lisp code should ever call post-command-hook. Emacs calls that hook somewhere in the command-loop, which is implemented in C.

[Kungsgeten]

@tarsius Yes ofcourse. I just find it strange that the quoted code somehow has the effect that post-command-hook isn't run after the function has been executed (and we've returned to the original buffer).

[tarsius]

I just find it strange that the quoted code somehow has the effect that post-command-hook isn't run after the function has been executed

I strongly doubt that it has that effect. In fact pressing q in a popup buffer does not even call this function. Also note that Magit no longer uses magit-popup it now uses transient. So if you suspect that this issue has something to do with the Magit's popups, then you should try to reproduce the issue with a recent snapshot of Magit, which uses transient.

[Kungsgeten]

@tarsius I'm using the latest version of Magit when testing, so transient is used. I know that q doesn't run magit-restore-window-configuration directly, but it does so via magit-mode-bury-buffer and magit-bury-buffer-function by default. I tried using magit-mode-quit-window instead, which seems to work.

I suspect that set-window-configuration perhaps runs post-command-hook before everything has actually been changed (or something like that). Anyway thanks for your replying and keep up the good work!

[098799]

It's not only magit's behavior that is affected. I can't provide additional examples until Monday unfortunately.

[Kungsgeten]

There's no rush :) But please see if you can give other examples and perhaps I can try to see if there's a common factor among them.

[098799]

Another case: I'm in ryo, I open a file (counsel-find-file) that doesn't have ryo hook enabled and the indicator is still red, but ryo-mode is off (until I press any key, then it's updated).

[Kungsgeten]

I also get this bug. Unfortunately I don't know how to fix it. One "solution" could be to use a different cursor for RYO, for instance like this:

(setq ryo-modal-cursor-type 'hbar)
(setq ryo-modal-cursor-color nil)

[098799]

Perhaps that should be the out-of-the-box behavior, since it's the only correct one. Fixes magit issues as well, btw.

[eldyer]

I have the same issue, but the use of a different cursor indeed fixes it.

[BlankSpruce]

I'm not sure if it covers all the cases but here's something I've discovered:
For current ryo-modal I couldn't recreate the first example provided in this issue but I can observe wrong cursor color when I do the following sequence:

• ryo-modal-mode on in the buffer
• C-x C-f
• call embark-act, press ! (to select shell-command)
• call any shell command (like echo 123)
• shell commands ends, ryo-modal-mode is still on, wrong cursor color until I do any movement command

Now if I add the following to my init.el:

(defun foobar ()
  (when ryo-modal-mode
    (ryo-modal--cursor-color-update)))

(add-hook 'buffer-list-update-hook 'foobar)

the color of the cursor is changed as expected as soon as shell command ends. I don't know any major disadvantages of this approach though.

[098799]

Thanks @BlankSpruce. It's been a few years and I've gotten used to only caring about the shape of the cursor, which I've customized to be vim-like (thin on insert mode, i.e. ryo-modal-off). This has been working fine.

All in all my Emacs setup has matured a little bit and i don't change packages and config too often anymore. Ryo-modal has retained it's spot as one of the most important packages in my configuration, kudos @Kungsgeten !