On 2026-05-17 13:40, J.D. Smith wrote:
> Replied separately. I /think/ I'm caught up on your review and
> suggestions; let me know if I missed something. Biggest thing now is to
> text v30 vs. v31 (which I haven't yet done), for functionality:
>
> - Enter/exit at correct locations for both (modulo the small difference
> I pointed out w.r.t. the end).
>
> - Editing text near boundaries (just inside, just outside) does what's
> expected for both.
>
> - Changing buffer in windows works as expected for both.
>
> Thanks for the thorough tests and good suggestions!
Thanks for your work, tests continue. Sorry for not focusing on what
you would like to have tested, but I just still keep finding other
issues.
General remarks:
- All following tests done with Emacs 31.
- I have done my copyright assignment to the FSF, so you don't need to
worry about using any patches provided be me.
- Of course, all my patches are strictly noncommittal.
- The reference test document is still:
#+begin_src org
Foo =bar0 bar1 bar2= baz.
#+end_src
Here is a patch, advantages listed below:
@@ -273,14 +273,20 @@ portion."
Operates only when inside an entity wrapped by hidden text and
`org-inside-mode' is enabled. Text will be re-hidden when point leaves
the entity. See `org-inside-appearance' to enable automatic unhiding or
-configure other appearance settings."
+configure other appearance settings.
+This function returns a non-nil value if it has toggled visibilty, nil
+otherwise. This makes it suitable as an entry in `org-ctrl-c-ctrl-c-hook',
+for example."
(interactive)
- (when-let* ((ov (window-parameter nil 'org-inside-overlay))
- (_ (and (> (overlay-start ov) 0)
- (> (overlay-end ov) 0))))
+ (and-let* ((ov (window-parameter nil 'org-inside-overlay))
+ (_ (<= (point-min) (overlay-start ov)))
+ (_ (<= (overlay-start ov) (point)))
+ (_ (< (point) (overlay-end ov)))
+ (_ (< (overlay-end ov) (point-max))))
(let ((inv (overlay-get ov 'invisible)))
(overlay-put ov 'invisible
- (if inv nil 'org-inside--not-hidden)))))
+ (if inv nil 'org-inside--not-hidden))
+ t)))
;;;###autoload
(define-minor-mode org-inside-mode
[ Not sure whether I got all the "<" vs "<=" right - I just hacked
that together rather quickly. ]
- It makes `o-i-t-v' suitable for addition to some abnormal hooks,
which I in particular would be keen to have.
- It avoids touching markers outside of a narrowed region. Not
sure whether you consider this an issue.
- It fixes a bug like this:
+ "make repro"
+ M-: (setq org-hide-emphasis-markers t) RET
+ M-: (setq org-inside-appearance nil) RET
+ open reference Org mode document
+ M-x org-inside-mode RET
+ M->
+ M-x org-inside-toggle-hidden RET (*outside* of the emph text)
+ C-b a couple of times
=> This shows the emph markers when moving into the emph text,
since the call to `o-i-t-h' has "primed" the overlay already to
be visible, IIUC.
That window-vs-buffer thingy will surely be good for more surprises.
Here is one:
+ make repro
+ (setq org-hide-emphasis-markers t)
+ open reference Org mode document
+ M-x org-inside-mode RET
+ Move into emph text and let cursor turn bar
+ C-x 2
+ C-x o
+ M-x org-inside-mode RET
+ C-x o
=> Cursor stays bar in upper window.
I guess whenever the visibility or appearance changes for whatever
reason, you'd need to walk all windows.
BTW, when walking windows (regardless for above issue or for
`org-inside--reset-all'), probably consider walking with
(walk-windows ... nil t)
Otherwise, e.g., *customize*-in-one-frame-org-mode-buffer-in-other
scenarios still can lead to dangling bar shaped cursors when
customizing o-i-a. These operations should be rather infrequent,
so more walking shouldn't be a problem.
The custom spec has issues. Well, at least one, and only in an
exotic (?) Emacs session where cus-start.el has not been loaded
yet and `(get 'cursor-type 'custom-type)' can equal nil:
+ "make repro"
+ M-x load-library org-inside RET
+ M-x customize-variable org-inside-appearance RET
=> Getting stack trace
#+begin_example
Debugger entered--Lisp error: (void-function nil)
nil((nil) bar)
widget-apply((nil) :match bar)
widget-match-inline((nil) (bar))
widget-group-match-inline((group :format "Key: %v" :inline t :args ((const
:args nil :value :cursor) (nil))) (:cursor bar))
widget-apply((group :format "Key: %v" :inline t :args ((const :args nil
:value :cursor) (nil))) :match-inline (:cursor bar))
widget-match-inline((group :format "Key: %v" :inline t :args ((const :args
nil :value :cursor) (nil))) (:cursor bar))
#+end_example
The following crude hack has helped:
@@ -75,7 +75,8 @@ still be used to unhide hidden markers for the entity you are
inside."
(plist
:tag "Specific appearance options"
:options
- ((:cursor ,(get 'cursor-type 'custom-type)
+ ((:cursor ,(progn (require 'cus-start)
+ (get 'cursor-type 'custom-type))
:tag "Cursor Type")
(:face (choice (face :tag "Face Name")
(plist :tag "Attribute List"))
A more fundamental approach probably would be to ask for an
"official" constant providing what the top-level `let' form in
cus-start.el provides as `cursor-type-types'.
[ It can be that I have been seeing pink elephants again here,
caused by test fatigue. At some time I couldn't reproduce the
stack trace shown above, but then it happened again ... ]
I've noticed trailing whitespace in org-inside.el, not sure
whether you or anybody else bothers.
The story will probably continue ...
