This one is confusing. `org-read-property-name', the function that handles completion for `org-set-property', and `org-pcomplete' only take into account the value of `org-buffer-property-keys'. `org-buffer-property-keys' returns a list of all property keys that are used in the current buffer, including any properties that are defined using the "#+PROPERTY" keyword. That is why "#+PROPERTY: UNIQUETESTPROP BLANK" works. `org-buffer-property-keys' does not take into account `org-global-properties' and as far as I can tell that is not a bug.
`org-global-properties' only allows you to set property property/value pairs that can be inherited by subheadings. So currently, AFAIK, there is no way to change the properties that will be completed in a buffer besides actually using them or defining them with the "#+PROPERTY" keyword. However, I am sure that a patch would be welcome. Le lun. 9 nov. 2026 à 12:23, Raymond Zeitler <[email protected]> a écrit : > > 0. Set* org-global-properties to '(("MEETINGID" . "BLANK") ("PASSCODE" . > "BLANK") ("UNIQUETESTPROP" . "BLANK")) > 1. Restart Emacs and org-agenda > 2. With point on an org file heading, invoke org-set-property (C-c C-x p). > 3. Press <TAB> at the Property prompt. > 4. Completion buffer shows several property names, including > CLOCK_MODELINE_TOTAL_ALL and VISIBILITY_ALL, but not any of the names defined > in org-global-properties. > > However, adding the following line at the top of the buffer and reloading the > file produces the desired effect, namely that pressing UNI<TAB> is completed > to UNIQUETESTPROP. > #+PROPERTY: UNIQUETESTPROP BLANK > > My goal is to maintain a consistent collection of property names among all my > org files. > > Is this a bug? If not, what am I doing wrong? Thanks > > * The setting is achieved through customize; custom-set-variables has this: > '(org-global-properties > '(("MEETINGID" . "BLANK") ("PASSCODE" . "BLANK") > ("UNIQUETESTPROP" . "BLANK"))) > > As well, I have the settings (global-completion-preview-mode t) and > (completion-styles '(basic partial-completion emacs22 substring)), although > as completion works for the buffer-defined property, this should be > irrelevant. > > (version)"GNU Emacs 30.1 (build 2, x86_64-w64-mingw32) of 2025-02-23" > (org-version)"9.8.10" > > - Ray > > Raymond Zeitler > [email protected] > https://ray-on-emacs.blogspot.com
