[BUG] org-auto-repeat-maybe: error "Can’t expand minibuffer to full frame" and missing log note

[Bhavin Gandhi <bhavin7392@gmail.com>]

On Mon, 13 Jun 2022 at 23:01, Bhavin Gandhi <bhavin7392@gmail.com> wrote:
> > 1. emacs -Q
> > 2. Create test.org with following content
> >
> > * TODO Repeating entry
> > SCHEDULED: <2022-05-01 Mon ++2d>
> >
> > 3. C-c C-t on the entry, shows the above error. And shifts the scheduled
> >    date correctly.
>
> When this error appears, the log note entry about state change is not
> added. I found this issue for the first time when I saw the state
> change log notes were missing.
>
> If I do C-c C-t again on the same entry, we get the note.
> - State "DONE"       from "TODO"       [2022-06-13 Mon 22:48]
>
> This is missing when we hit the error mentioned in the first mail.

Confirmed.
Trying to change the subject for Woof! for the first time, I hope I get
it right.

I remember reading this code around a year back, where we were fixing
two bugs.
1. Bug: Duplicate logbook entry for repeated tasks:
   https://list.orgmode.org/orgmode/CAOn=hbcaW1R6vtun-E2r4LS=j3dp=VjqmjGtzy8UC1SyPArKbA@mail.gmail.com/
2. Bug: org-agenda-undo does not work with repeated tasks [9.4]:
   https://list.orgmode.org/87v98a8mes.fsf@gnu.org/

Currently, when we call `org-auto-repeat-maybe', it adds
`org-add-log-note' to post-command-hook using `org-add-log-setup'. As
soon as we reach the question about 10 repeater intervals (the call to
`y-or-n-p' from `org-auto-repeat-maybe'), the post-command-hook gets
executed.[1]

So, we end up calling `delete-other-windows' while we are waiting for
the user to respond with y/n. And this causes error "Can’t expand
minibuffer to full frame".

I don't know how to fix this issue yet, but I'm planning to add a test
case for completing repeating tasks which need 10+ repeating intervals
to shift the date.

Maybe this is off-topic:
I'm not sure why [1] happens, my assumption was Emacs will wait till all
the functions finish executing, and then run the post-command-hooks.
this-command remains org-todo (C-c C-t) throughout all the other calls
done by org-todo as long as they are not commands
i.e. org-auto-repeat-maybe etc. So, I tried following snippet:

(defun my-test-command ()
  (interactive)
  (message "my-test-command: setting post-command-hook")
  (add-hook 'post-command-hook 'my-own-hook)
  (y-or-n-p "Some question: ")
  (message "my-test-command: message after setting the hook")
)
(defun my-own-hook ()
  (remove-hook 'post-command-hook 'my-own-hook)
  (message "my-own-hook: hopefully executing at the end")
)

In this case as well, as soon as it calls y-or-n-p, I can see
"my-own-hook: hopefully executing at the end" in the minibuffer (while
it is still waiting for user input). Is it expected? I read (info
"(elisp)Command Overview") but couldn't understand this behavior.

-- 
Bhavin Gandhi (bhavin192) | https://geeksocket.in