For some reason ob-R refuses to compile when it requires ob-comint.
When (require 'ob-comint) is not included in ob-R.el everything
compiles without error, but warnings are thrown because the
arguments to a macro defined in ob-comint are mis-interpreted as
functions.
When (require 'ob-comint) is added to ob-R.el then it throws errors
complaining that the last argument to a function is nil and should
be a string. I don't understand this error at all and can't fix it.
* lisp/org.el (org-babel-load-languages): this variable controls which
languages will be loaded by org-babel. It is customizable through
the customize interface.
(org-babel-do-load-languages): load those languages in
org-babel-load-languages and disable those with nil cdr's
* lisp/babel/ob.el (org-confirm-babel-evaluate): variable used to
control evaluation of code blocks, default value it t, meaning all
code block evaluation requires confirmation
(org-babel-confirm-evaluate): function used to request confirmation
of code block evaluation from the user
(org-babel-execute-src-block): this function is the single point of
entry for evaluation of code blocks (whether initiated through lob
call, through direct code block evaluation, or as part of file
exportation). Every time this function is called it will now
request confirmation from the user. The newly added
`org-confirm-babel-evaluate' variable can be used to configure this
behavior.
(org-babel-no-eval-on-ctrl-c-ctrl-c): This variable can be used to
inhibit evaluation of code blocks with C-c C-c.
* lisp/org.el (org-ctrl-c-ctrl-c): added documentation of code block
evaluation behavior
* lisp/babel/ob-keys.el (org-babel-key-bindings): adding keybindings
for executing code blocks and for opening their results
* lisp/org.el (org-entry-get-with-inheritance): New argument LITERAL-NIL.
(org-entry-get): Pass `literal-nil' into
`org-entry-get-with-inheritance'.
(org-todo): React to nil values of the LOGGING property.
* lisp/org.el (org-switchb): Renamed from `org-iswitchb'. Improve
docstring.
(org-iswitchb): New alias.
(org-ido-switchb): Make alias point to `org-switchb'.
* lisp/org.el (org-time-string-to-absolute): Ignore cyclic repeater
when displaying items on todays agenda date.
Ignore the cyclic repeater when displaying items on today's agenda
date. If you have a weekly task and miss the date the agenda view
will show more than a week late now instead of resetting on the
cyclic repeating date. This makes it much more obvious when you
missed a repeating task after the repeater.
* lisp/org-html.el (org-export-html-preprocess): Call org-format-latex,
possibly with a protect-only argument.
* lisp/org.el (org-format-latex): New argument PROTECT-ONLY.
with the switch #+OPTIONS: LaTeX:verbatim ,
LaTeX code will be exported verbatim to HTML, so that jsmath can grab
and convert it.
Proposed by Christian Moe.
* lisp/org-macs.el (org-not-nil): Return the value if not interpreted
as nil.
* lisp/org.el (org-entry-get):
(org-entry-get-with-inheritance): Interpret the value "nil"
as nil for properties.
Bernt Hansen writes:
> Carsten Dominik <carsten.dominik@gmail.com> writes:
>
> > On Jun 25, 2010, at 3:23 PM, Robert Goldman wrote:
> >
> > > Question: what is the proper way to get a NIL into a property? Are
> > > we
> > > to use () instead of "nil"? Or are property values always interpreted
> > > as strings?
> > >
> > > Apologies in advance if this is a stupid question!
> >
> > Not a stupid question at all.
> >
> > There is no way, currently. Property values are string - the only
> > way to make
> > org-entry-get return nil is to not have the property defined at all.
>
> I've wanted a similar thing in the past for the LOGGING property where
> the parent task has special logging set via the LOGGING property but I
> want to undo that for some of the child tasks so they use the default
> logging setup.
>
> Having a way to undefine a property would be good in general I think.
-Bernt
* lisp/org-macs.el (org-not-nil): New function.
* lisp/org.el (org-block-todo-from-children-or-siblings-or-parent):
Use `org-not-nil' to interpret a property value of nil.
Robert Goldman writes:
> I have found what I believe to be a bug in handling ordered subtasks.
> Here is the behavior:
>
> I have a top level set of tasks that is ordered.
>
> One of the outline items below the top level set is a grab bag of tasks
> that will be performed in parallel. So this task is NOT ordered
> (ORDERED: nil).
>
> The problem is that the blocking behavior from ordered tasks seems to be
> inherited from the top level task list into the second level of the
> outline, even though the ORDERED property at the second level is
> explicitly overridden.
>
> I am attaching an org file that displays this issue. To see the
> problem, put your cursor on the "Bar" task and attempt to change its
> status to DONE.
The problem was here that the value of the property is the string
"nil", which is of course not nil.
This patches introduces a special case to interpret "nil" as nil.
* lisp/org.el (org-timestamp-change): New optional argument UPDOWN.
Use this to identify calls from org-timestamp-up/down, so that we can
skip by rounding minutes in this case.
(org-timestamp-up):
(org-timestamp-down):
(org-timestamp-up-day):
(org-timestamp-down-day): Call org-timestamp-change with the
updown argument.
This function used to look only if there was a prefix argument in the
current interactive call. Now we have an explicit marker indicating
that the function is called from the updown commands.
* lisp/org-agenda.el (org-agenda-action): Make `c' key call org-capture.
* lisp/org-capture.el: New file.
* lisp/org-compat.el (org-get-x-clipboard): Function moved here from
remember.el.
* lisp/org-mks.el: New file
* lisp/org.el (org-set-regexps-and-options): Allow statistic cookies as
part of complex headlines.
(org-find-olp): New argument THIS-BUFFER. When set, assume that the
OLP does not contain a file name.
* lisp/org.el (org-mode): Set `comment-start' instead of changing the
syntax of the `#' character.
Leo writes:
> Setting a comment starter without a corresponding comment ender is
> problematic and the # creeps in mysteriously under auto-fill. For
> example, in my current running emacs, this happens almost certainly in
> all org files that has # in their header. The only (temporary) solution
> seems to reboot emacs (which is painful and disruptive).
>
> If you try eval (forward-comment 1) at the beginning of an org file that
> has some "#+..." it will move to the end of file (the whole file is
> regarded as one single comment). So when auto-fill a long text, it will
> find the common prefix to be #.
>
> In addition, I don't think org mode has clear comment syntax or ideas on
> what to do with it.
>
> I can't see any gain from (modify-syntax-entry ?# "<") so I am proposing
> removing it entirely and get rid of this mysterious and annoying bug
> once and for all.
>
> The attached patch may (though I think it is quite safe) cause some bugs
> but those will be fixable unlike the one mentioned above.
>
> Best wishes,
>
> Leo
(org-agenda-get-scheduled):
* lisp/org.el (org-time-string-to-seconds):
For deadline and scheduled agenda display ignore the cyclic repeater
when calculating how many days late the task is. If you have a weekly
task and miss the date the agenda view will show more than a week late
now instead of resetting on the cyclic repeating date. This makes it
much more obvious when you missed a repeating task after the repeater.
Thanks to Bernt Hansen for this patch.
* org.el (org-raise-scripts): Do not fontify sub/superscripts of text
with face `org-special-keyword'. Makes property keys as :LAST_REPEAT:
display correctly.
This patch adds a modification hook to remove inline-image overlays if
the underlying text is modified. This prevents blind editing of text
behind images.
* lisp/org.el (org-raise-scripts): Do not act in links.
Matt Lundin writes:
> Git commit 70d24c5d03 causes underscores
> in hyperlinks to display parts of link and description as subscripts.
> E.g., this link...
>
> [[http://www.samplepage.com/an_underscore][Some description words]]
>
> ...displays the word "Some" in the description as a subscript. With M-x
> visible-mode, the substring "underscore][Some" is displayed as a
> subscript.
* lisp/org.el (org-edit-special): Make sure source code editing goes
before table formula editing.
* lisp/org-table.el (org-table-fedit-map): "C-c '" will now also exit
the formula editor.
* lisp/org.el (org-read-date-analyze): Fix regular expression for
matching american dates
Daniel E. Doherty writes:
> In playing around with the date prompt (C-.), I ran across the following
> puzzling behavior from rather simple inputs.
>
> I entered the following on June 1, 2010. Here is a date entered as
> "3/15": <2011-03-15 Tue>. It interpreted it as the upcoming March 15 as
> expected.
>
> But here is a date entered as "5/21": <2021-06-05 Sat>. Note how it
> interpreted the "21" as the year 2021, not at all what I expected from
> the documentation or the analogous "3/15" example.
>
> Maybe there is some underlying logic here that I'm not getting. Perhaps
> it has to do with how 2-digit years are interpreted?
>
> What's going on here? I am using org-version 6.36trans on emacs 23.1.
What was going on here is that the regular expression for matching
american-style dates was wrong. It was looking for month numbers in
the second field and day numbers in the first field - wrong, of
course.
* lisp/org-macs.el (org-rm-props): Add org-emphasis to the properties
that must be removed.
* lisp/org.el (org-do-emphasis-faces): Add org-emphasis property to
items that have emphasis done wit font-lock.
(org-fontify-entities): Do not do anything in commented lines.
(org-unfontify-region): Decompose the region as well, because we do
composition during font-lock.
(org-raise-scripts): Do nothing inside an emphasis string.
Reported by Eric Fraga in http://article.gmane.org/gmane.emacs.orgmode/25940
* lisp/org-compat.el (org-string-match-p):
(org-looking-at-p): New functions.
* lisp/org-table.el (org-table-align): Handle raised text with
invisible characters.
* lisp/org.el (org-script-display): Add raise properties for tables.
(org-raise-scripts): Handle raising differently inside tables.
Pretty display of subscripts and superscripts no longer messes up
table alignment. This is achieved by two things:
1. Inside tables, the raised characters are not made smaller, they
remains at the same size. Instead they are raise/lowered more, by
a full half character height to still be clearly readable as
subscript or superscript.
2. The invisible characters are taken into account when computing the
field width.
* lisp/org.el (org-ctrl-k-protect-subtree): New option.
(org-kill-line): Protect hidden subtrees if the user wants it.
* doc/org.texi (Headlines): Mention the special behavior of C-k
in headlines.
Scott Otterson writes:
> For what must be the dozenth time, I've just accidentally deleted a
> large tree by typing C-k while in a headline.
>
> This is really easy to do because emacs users have "C-k deletes to the
> end of the line" worn deeply into their neural pathways -- it's so
> automatic for me that the keystroke is close to subconscious. A
> mistaken C-k is especially hard to detect because org-mode displays
> the result exactly like what your subconscious expects, that is, a
> collapsed headline is deleted to the end -- and the tree underneath is
> wiped out with no noticeable warning.
>
> Feature request: add an option preventing tree deletion with C-k
> without user confirmation. Actually, I'd like an option to prevent it
> period.
>
> If this option is already in there, then you're encouraged to tell me
> to RTFM. But then also please tell me where it is, because I can't
> find it.
Carsten replies
> This is now possible due to the variable
> `org-ctrl-k-protect-subtree'. But I predict that you are going to set
> it to nil again soon :D
* doc/org.texi (Subscripts and superscripts): Document that `C-c C-x \'
will also format sub and superscripts.
* doc/orgcard.tex: Document that `C-c C-x \'
will also format sub and superscripts.
* lisp/org.el (org-use-sub-superscripts):
(org-pretty-entities-include-sub-superscripts): Move here from
org-exp.el.
(org-set-regexps-and-options): Parse subscript option and set
`org-use-sub-superscripts' as a local variable.
(org-match-sexp-depth):
(org-create-multibrace-regexp):
(org-match-substring-regexp):
(org-match-substring-with-braces-regexp): Moved here from org-exp.el
(org-set-font-lock-defaults): Call `org-raise-scripts'.
(org-remove-font-lock-display-properties): New function.
(org-unfontify-region): Call
`org-remove-font-lock-display-properties'.
(org-script-display): New constant.
(org-raise-scripts): New function.
When turning on entity display with `C-c C-x \', sub- and superscripts
will also be displayed in a smaller font, and raised/lowered.
* lisp/org-latex.el (org-export-latex-fontify): Avoid fontifying
several stars in a row.
* lisp/org.el (org-emphasis-alist): Mention
`org-export-docbook-emphasis-alist' in the docstring.
* lisp/org.el (org-store-link): Don't call org-store-link
interactively when called through remember from the agenda
Samuel Wales writes:
> For months, I found ID properties where they didn't belong.
> Finally I found that they are added when I do M-x
> org-remember in an agenda buffer, to the headline under
> point. Feels like a bug. [...]
The new function `org-agenda-file-p' checks if a given file is an org
agenda file. Such a function is very useful in hooks, for example if
you want to export agenda files automatically when saving:
(defun th-org-mode-init ()
;; Update appointments and export to iCalendar when saving.
(when (org-agenda-file-p)
(add-hook 'after-save-hook 'th-org-agenda-to-appt t t)
(add-hook 'after-save-hook 'org-export-icalendar-this-file t t)))
(add-hook 'org-agenda-mode-hook 'th-org-agenda-mode-init)
Org-refile-cache fails when org-refile-use-outline-path is set to file.
Specifically, org-refile-cache-check-set throws a markerp error when it
encounters file targets, since they have nil instead a marker object.
This patch applies the test only to targets with markers (i.e.,
headings).
The function org-get-refile-targets was building org-refile-cache even
if org-refile-use-cache was set to nil. This caused every refile
attempt to call org-refile-cache-clear and to produce the message
"Refile cache has been cleared."
Tom writes:
> if I have a heading like this:
>
>
> ** TODO test task
> stuff
> SCHEDULED: <2010-05-15 Sat 07:35 +1d>
>
>
> Then an empty line is inserted below the heading (before "stuff") if
> org-indent-mode is on and logging is set like this:
>
>
> (setq org-log-repeat nil)
> (setq org-log-done 'time)
>
>
>
> I tested it with a clean config using only the settings above.
This patch fixes the problem first reported by Tassilo Horn in
[mid:87y6fhxc47.fsf@thinkpad.tsdh.de]. Problem was that
`org-refile-cache-get' returned an invalid refile target table after
the refile cache was cleared.
Baoqui Cui writes:
> "robut@iinet.net.au" <robut@iinet.net.au> writes:
>
> I very much like the idea of native inline image display in Org-mode but can't
> seem to make it work.
>
> Given a 6.36 snapshot or 6.36 release and these org file contents
>
> * Test image
> Test image
> [[Screenshot.png]]
>
>
> I hoped org would display that image after C-c C-x C-v. Rather Org-mode returns
> "No images to display inline".
>
> I've tried different ways of linking that image, different image formats,
> relative vs complete paths, and my regular .emacs vs a near empty one and
> always the same result. If I toggle iimage-mode the image displays fine per se
> but does not affect how Org-mode works.
>
> Seems clear I am missing something simple. What?
>
> I like the idea of inline image display too, but hit the similar
> problems. After reading the code in org.el, I found that the inline
> image file link has to start with either "file:" or "./".
>
> For example, the following two links are OK:
>
> [[file:~/images/myImage.png]]
> [[./figures/org-mode-unicorn.svg]]
>
> but the following two are not:
>
> [[Screenshot.png]]
> [[~/images/myImage.png]]
>
> Here is a small patch that seems to work well for me, but I'd like
> Carsten to check whether it may break anything
Patch by David Maus, who writes:
> Org enters an infinite loop when `org-replace-escapes' is called with
> a table containing a replace string that contains the escape sequence
> it should be replaced with.
>
> Example:
> ,----
> | (org-replace-escapes "%m" '(("%m" . "87zl0qq1f3.wl%maus.david@gmail.com")))
> `----
>
> I stumpled upon when I tried to store a link to a internet message
> whose message id contained the sequence "%m" (perfectly valid for a
> message id) while using "%m" as message description.
>
> Attached patch fixes this by
>
> 1. detecting such 'self reference' and replacing the offending
> sequence in the replace string by a string with a text property
> that contains the original sequence
>
> 2. replacing occurences of substrings with this text property by the
> original sequence.
This behavior is now parallel to the treatment of outline nodes.
This commit also introduces another change. When an outline node or a
plain list item is folded by outline and contains hidden children,
M-left/right will refuse to act on this item. You must either open
the tree, or use the subtree commands M-S-left and M-S-right.
Based on a patch by Matti De Craene, but significantly modified after
a discussion involving Bernt Hansen and others.
The definition of "makes sense is here:
- either the user is logging repeats (org-log-repeat)
- or the entry contains clock data, in which case the LAST_REPEAT is
needed to display clocking time properly.
Request by Dan Griswold, with some support from Bernt Hansen
The target state can now be fixed locally with the REPEAT_TO_STATE
property, or globally with the variable `org-todo-repeat-to-state'.
This was a request by John Wiegley.
By default, title, author, date and email lines appear in dark blue
with the initial keywords greyed out. The title is in a larger font
than the others. This is implemented by the following new faces:
org-document-title
org-document-info
org-document-info-keyword
In addition, the variable org-hidden-keywords can be used to make the
corresponding keywords disappear.
Magnus Henoch writes:
> This patch has been sitting in my tree for a while... It's a fix to
> org-map-dblocks, to make it use save-excursion instead of remembering
> position values. I need this since I have a dblock function that
> asynchronously updates dblocks from HTTP responses, and some dblocks
> ended up getting updated twice or thrice.
[...]
> My dblock-write function calls url-retrieve, to asynchronously retrieve an
> HTML page. The callback function I pass to url-retrieve will then fill
> in the information I need into the dynamic block.
>
> So in the following case:
>
> * Find start of dblock 1, store as pos
> * Make HTTP request for dblock 1
> * Go back to pos
> * Find end of dblock 1
> * Find start of dblock 2, store as pos
> * Make HTTP request for dblock 2
> * Asynchronous event: HTTP response for dblock 1 arrives, insert lots of
> data in dblock 1
> * Go back to pos
> * Find end of dblock 2
>
> the last step will actually find the end of dblock 1, if the amount of
> data inserted in dblock 1 is great enough that pos suddenly points
> inside it. (Then it will of course find dblock 2 again, request its HTML
> page again, and thus insert the data twice.)
>
> An equivalent fix would be to make pos a marker instead.
Ryan Thompson writes:
> I have found a bug. When the point is at the end of an empty headline
> and you press M-RET (or C-RET) to make a new headline, it deletes all
> the whitespace at the end of the empty headline first, which causes
> the headline to break. I'm not sure if the correct behavior is to
> leave an empty headline, or maybe just do nothing and leave the point
> at the end of the empty headline without creating a new one, but the
> correct thing is definitely *not* to break the headline.
Patch by Jan Böker.
Jan writes:
> What is this?
> =============
>
> This patch changes the way extension regexps in `org-file-apps' are
> handled. Instead of against the file name, the regexps are now matched
> against the whole link, and you can use grouping to extract link
> parameters which you can then use in a command string to be executed.
>
> For example, to allow linking to PDF files using the syntax
> file:/doc.pdf::<page number>, you can add the following entry to
> org-file-apps:
>
> Extension: \.pdf::\([0-9]+\)\'
> Command: evince "%s" -p %1
>
> In a command string to be executed, the parameters can be referenced
> using %1, %2, etc. Lisp forms can access them using (string-match n link).
>
>
> Where to get it?
> ================
> Either apply the patch by hand or
>
> git pull git://github.com/jboecker/org-mode.git org-file-apps-parameters
>
>
> What's next? / Feedback
> =======================
>
> - Find the bugs. Since this messes with links, a central concept of Org,
> I probably have missed some edge cases; so please test this and
> report if it works for you.
>
> I also appreciate any feedback on code quality or the design decisions
> made. I am learning elisp along the way, so you may be able to write
> some changes in a more idiomatic and/or elegant way.
>
> - Add a mechanism for org-mode modules to add default values to
> org-file-apps, similar to the variables org-file-apps-defaults-*.
> This could be used by modules to define their own extensions to the
> syntax of file: links.
>
> - Modify org-docview.el to use this and deprecate the docview: link syntax.
>
>
> What does it (intentionally) break?
> ===================================
>
> This patch introduces a backwards-incompatible change. If LINE or SEARCH
> is given, the file is no longer guaranteed to open in emacs: if IN-EMACS
> is nil and an entry in org-file-apps matches, that takes precedence.
>
> A grep of the lisp/ and contrib/ directories showed that no code in the
> org-mode distribution was relying on this behaviour; whereever LINE or
> SEARCH is given, IN-EMACS is also set to t.
>
> I decided against adding an additional parameter because that would be
> redundant; the original link as seen by org-open-at-point can be
> reconstructed from PATH, LINE and SEARCH.
>
> I am not that sure if this is the right way to do this, but it seems to
> break as little as possible while hopefully avoiding to add too much
> complexity.
Patch by Dan Hackney.
Dan Hackney writes:
> For paragraph text, `org-adaptive-fill-function' did not handle the
> base case of regular text which needed to be filled. This commit saves
> a buffer-local value of `adaptive-fill-regexp' and uses it if none of
> the org-specific regexps match. This allows email-style ">" comments
> to be filled correctly.
John Wiegley writes:
> I have the following data in my Org-mode file:
>
> #+LINK: cegbug https://portal/bugzilla/show_bug.cgi?id=
>
> ** TODO [[cegbug:351][#351]] Bizcard: Fix Maven build setup
> - State "TODO" from "STARTED" [2010-03-01 Mon 14:42]
>
> Now, in the Agenda and in the Org-mode buffer, everything looks fine.
> I can also use C-c C-o if my cursor is within the #<NUMBER> text.
>
> However, if I'm in the Agenda and I hit C-c C-o, it says 'No match'
> after about a second. Is there any reason I can't open these links
> from the Agenda view?
This support was totally broken. It works now again. Unfortunately
it is not possible to edit the table directly in the org-mode buffer
anymore - to edit such a table, you have to use C-c '
This patch implements reading American dates, like
2/5/3 --> 2003-02-05
2/5 --> ????-02-05
Is also fixes a bug that would force the current year when reading a
date like 2/5 (American) or 2-5 (ISO), and in this way would prevent
`org-read-date-prefer-future' to do its job. This bug was reported by
Lukasz Stelmach.
Move the marker for the last refile location to the heading instead of at the
end of the entry. This was causing C-u C-u C-c C-w to end up on the heading
following the newly refiled entry.
This can help to get out of an inconsistent state produce for example
by viewing from the agenda. Reported by Matt Lundin:
> I'd like to report a minor issue with org-agenda-goto and inline tasks.
> Let's say one has the following file:
>
> --8<---------------cut here---------------start------------->8---
> * Here is an entry.
> Blah blah blah blah.
> *************** Here is an inline task.
> *************** END
> Blah blah blah blah blah.
> *************** TODO Here is a second inline task.
> *************** END
> Blah blah blah blah blah.
> *************** Here is a third inline task
> *************** END
> Blah blah blah blah blah.
> --8<---------------cut here---------------end--------------->8---
>
> Let's say one also has the following settings:
>
> --8<---------------cut here---------------start------------->8---
> (setq org-show-hierarchy-above t)
> (setq org-show-siblings '((default . nil) (isearch . t) (agenda . t)))
> (setq org-show-entry-below '((default . nil) (isearch . t) (agenda . t)))
> --8<---------------cut here---------------end--------------->8---
>
> If 1) one tries to jump to the TODO from the agenda and 2) the entry is
> currently folded, org-show-context reveals only the headlines. E.g.,
>
> --8<---------------cut here---------------start------------->8---
> * Here is an entry.
> *************** Here is an inline task.
> *************** END...
> *************** TODO Here is a second inline task.
> *************** END...
> *************** Here is a third inline task
> *************** END...
> --8<---------------cut here---------------end--------------->8---
>
> Invoking org-cycle on the END headline does nothing, since all headlines
> deeper than org-inlinetask-min-level are exempted from cycling. As a
> result, the only way to reveal the text in the entry is to cycle the
> parent twice (first to close, then to reveal).
Ruud Brekelmans writes about problems with spurious footnotes:
> I still find similar behavior when exporting to LaTeX with:
>
> #+BEGIN_LaTeX
> \newcommand{\norm}[1]{\lVert#1\rVert}
> #+END_LaTeX
Emilio Arias writes:
> egallego@babel.ls.fi.upm.es (Emilio Jesús Gallego Arias) writes:
>
> To reproduce save this minimal org file:
>
> #+STARTUP: even
> * A
> :PROPERTIES:
> :ARCHIVE: a
> :END:
> ** B :ARCHIVE:
> Some text
>
> and hit TAB when in the * A headline; then the ** B headline contents
> will be incorrectly shown.
>
> I've found the culprit in org-hide-archived-subtrees:
>
> ,----
> | (defun org-hide-archived-subtrees (beg end)
> | "Re-hide all archived subtrees after a visibility state change."
> | (save-excursion
> | (let* ((re (concat ":" org-archive-tag ":")))
> | (goto-char beg)
> | (while (re-search-forward re end t)
> | (and (org-on-heading-p) (org-flag-subtree t))
> | (org-end-of-subtree t)))))
> `----
>
> The problem is that the RE matches the first archive "property" and
> then does an org-end-of-subtree which skips all the subtrees of the
> parent tree where the ARCHIVE property is located.
>
> I've replaced this part
>
> | (and (org-on-heading-p) (org-flag-subtree t))
> | (org-end-of-subtree t)))))
>
> by
>
> | (when (org-on-heading-p)
> | (org-flag-subtree t)
> | (org-end-of-subtree t)))))))
>
> so org-end-of-subtree is only called if we are really in a headline. I
> think that makes sense.
>
