Hi all!
I wanted to write a command which could modify an Org entry from the
corresponding agenda item, and for that I tried to use
`(org-with-point-at (org-get-at-bol 'org-hd-marker))` (ChatGPT's
suggestion). But I started noticing that sometimes it did not work and
it tried to edit instead the entry that was originally at point in the
Org buffer. After some debugging I confirmed that this happened if and
only if the original Org buffer was in a non-selected window (e.g., if
the agenda had been invoked from that buffer). I tried to use
`org-agenda-with-point-at-orig-entry` and exactly the same happen.
After some debugging investigation I identified the issue: those
macros only set the current buffer and the buffer point, but not
the window point. If the provided body has an implicit redisplay (and
mine did because it used `completing-read`), Emacs can decide that the
buffer point and window point are inconsistent and overwrite the
buffer point with the window point, so the `goto-char` in those
macros is effectively undone.
Coming up with a minimal reproduction example that manifest this
it's hard, because redisplay logic it's complex and doesn't always
re-syncs the buffer point. With my usual configuration for example,
I could always reproduce the issue just with
```
(org-with-point-at my-marker
(redisplay t)
(message "Point: %s. Original marker: %s" (point) my-marker))
```
But when I tried with emacs -Q, I no longer could. Eventually, I
narrowed down the configuration that made it happen and it turns
out to be flycheck-mode, which is enabled by default in all buffers
(even if Flycheck doesn't have a checker for Org buffers in my
configuration). If Flycheck is enabled, the issue happens, and if it's not
enabled, the issue doesn't happen. I guess Flycheck touches some
buffer state in the background which makes Emacs think the
reconciliation is needed. I tried to come up with some simpler
function to add before the redisplay which forced the points to be
reconciled later without flycheck mode enabled, but I couldn't.
I don't really know enough of Emacs and Org Mode internals to
determine whether this is a bug or just some necessary limitation
(I just learned yesterday about the difference between buffer and
window point after all, because of this), but I think it would be
nice to protect against that scenario in those macros. I tried
using `(with-current-buffer (clone-indirect-buffer nil nil)` inside
them and that fixes it, with no undesired side effect to the best
of my knowledge. If you agree, I can submit a patch, with that
workaround or any other someone more knowledgeable proposes. Also,
if there is concern we could start patching only
`org-agenda-with-point-at-orig-entry`, which is not used that much.
Let me know what you think
Best regards!
------------------------------------------------------------------------
Emacs : GNU Emacs 29.1 (build 1, x86_64-pc-linux-gnu, GTK+ Version 3.24.41,
cairo version 1.18.0)
of 2025-10-14
Package: Org mode version 9.6.6 (release_9.6.6 @
/usr/local/share/emacs/29.1/lisp/org/)
