it needs double "C-g" to quit corfu completion in org-mode buffer

[RNAer]

Normally, I can type C-g to run corfu-quit to exit the completion. Weirdly, in org-mode buffer, I need to type it twice. Trying to figure it out why, I added the following hook to see what's happening when I run C-g.

(add-hook 'post-command-hook
          (lambda ()
            (message "Command executed: %s" this-command)))

I find that the first C-g is calling keyboard-quit instead of corfu-quit (see the crude recording) and this only occurs in org-mode buffer. I am wondering if anyone knows why and how to fix it.

Untitled.mov

I am using GNU Emacs 30.0.93 (build 2, aarch64-apple-darwin24.2.0, NS appkit-2575.30 Version 15.2 (Build 24C101)) and latest corfu and builtin org-mode. Note that this problem persists even I purge all my org-mode settings.

[minad]

I don't see this problem. Can you reproduce this from emacs -Q?

[minad]

The inefficiency of Flyspell was one of the motivations for Jinx.

[kayzhang]

@minad Sure. I'll definitely give Jinx a try later.

[ISouthRain]

Hello?
Is there any good solution to this problem?
Does this look like a flyspell-mode issue?
I disabled flyspell-mode OK, it really fixed the problem.
Is there a better solution?

[ISouthRain]

(setq flyspell-delay-use-timer t) ;; fix need double "C-g" to quit corfu

[shahinism]

For me disabling flyspell didn't solve the issue, so probably there is something else on its way. Same behavior, logging Quit and flashing modeline. For the time being, I'm just binding it to <escape>.

[hanwenguo]

Sorry if disturbing anyone, but I have encountered a similar yet different issue. The behavior seems similar: if I press C-g at some point (being too early or too late won't trigger the issue) after triggering corfu popup, that C-g won't work.

However, the differences are:

• With the messaging setup from OP, the logged command for that C-g is nil instead of keyboard-quit.
• After (set-input-mode t nil 0 20), the same issue with C-g still happens. (C-t will always trigger command nil then, but that is natural since C-t is not bound in corfu-map.)

Although coming up with a guess is generally bad when reporting bugs, but for this one, my guess is that it's somehow the inverse of the above discussion: the command loop is waiting for keyboard input and C-g thus acts as an ordinary input character, which is mapped to nil in some weird keymap activated for some weird reason.

Would appreciate any suggestion on debugging.

[minad]

Do you even observe this in a minimal setup starting from emacs -Q with only Corfu loaded, or only few other modes?

[hanwenguo]

Yes. It seems that this is some issue between Prot's spacious-padding and corfu-popupinfo-mode. Below is a minimal example with emacs -Q:

(use-package spacious-padding)

(use-package corfu
  :config
  (setq corfu-auto t
        corfu-auto-prefix 1 ;; just to help with debugging
  )
  (corfu-popupinfo-mode 1)
  (setq corfu-popupinfo-delay 1.0) ;; just to help with debugging
)

(global-corfu-mode 1)
(spacious-padding-mode 1)

To trigger, type anything and try to press C-g just before (or after? I'm not sure as the time frame is very short) the moment the info popup is about to appear.

Relevant or not, I'm on Emacs 31.0.50 feature/igc and corfu 9d5fc8f. I'll test with HEAD corfu again. EDIT: still happening on current HEAD (79bb30e)