Running (prettify-symbols-mode) twice adds two entries to font-lock-keywords

Running (prettify-symbols-mode) twice adds two entries to font-lock-keywords

[Clément Pit--Claudel]

[-- Attachment #1.1: Type: text/plain, Size: 2704 bytes --]

Hi emacs-devel,

Do bug reports with a proposed fix go on the bug tracker, or on this list? I've seen patches discussed here, so I'm not too sure. Apologies if this is the wrong place (please do let me know!)

Changing the lists of symbols prettified by prettify-symbols-mode is more tricky than it should be; I think this is a bug in prettify-symbols-mode. Here is a test case:

(with-current-buffer (get-buffer-create "fl")
  (erase-buffer)
  (emacs-lisp-mode)
  (prettify-symbols-mode -1)
  (setq prettify-symbols-alist '(("A" . ?B)))
  (prettify-symbols-mode)
  (setq prettify-symbols-alist '(("A" . ?C)))
  (prettify-symbols-mode)
  (insert "A A A")
  (pop-to-buffer (current-buffer)))

Looking at this code, I expect “C C C” to be shown. Yet it's not the case; instead, “B B B” is shown.

On the other hand, this works:

(with-current-buffer (get-buffer-create "fl")
  (erase-buffer)
  (emacs-lisp-mode)
  (prettify-symbols-mode -1)
  (setq prettify-symbols-alist '(("A" . ?B)))
  (prettify-symbols-mode)
  (prettify-symbols-mode -1)
  (setq prettify-symbols-alist '(("A" . ?C)))
  (prettify-symbols-mode)
  (insert "A A A")
  (pop-to-buffer (current-buffer)))

I think the line indicated by !!! in the following code sample is wrong:

    (define-minor-mode prettify-symbols-mode
      (...)
      (if prettify-symbols-mode
          ;; Turn on
          (when (setq prettify-symbols--keywords (prettify-symbols--make-keywords))
            (font-lock-add-keywords nil prettify-symbols--keywords) ;; !!! HERE
            (setq-local font-lock-extra-managed-props
                        (append font-lock-extra-managed-props
                                '(composition
                                  prettify-symbols-start
                                  prettify-symbols-end)))
            (when prettify-symbols-unprettify-at-point
              (add-hook 'post-command-hook
                        #'prettify-symbols--post-command-hook nil t))
            (font-lock-flush))

Indeed, turning on prettify-symbols-mode twice in a row causes two entries to be added to font-lock-add-keywords. This is not an issue in general, except if one changes prettify-symbols-alist between two calls. In that case font-lock-keywords gets two distinct entries for prettify-symbols, and the second one is ignored.

The proposed fix is to move the following snippet up, so that it runs in all cases, instead of just running when prettify-symbols-mode is turned off.

    (when prettify-symbols--keywords
      (font-lock-remove-keywords nil prettify-symbols--keywords)
      (setq prettify-symbols--keywords nil))

Cheers,
Clément.


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 836 bytes --]

RE: Running (prettify-symbols-mode) twice adds two entries to font-lock-keywords

[Drew Adams]

> Do bug reports with a proposed fix go on the bug tracker, or on this
> list? I've seen patches discussed here, so I'm not too sure.
> Apologies if this is the wrong place (please do let me know!)

My understanding (but someone will correct me, if mistaken) is
that such things are _accepted_ either here (emacs-devel) or as
a bug report, but that using the bug tracker (i.e., reporting a
bug, with the fix/patch) is generally preferred/encouraged.

For one thing, that separates such stuff from this list.  For
another, the bug is tracked...

Re: Running (prettify-symbols-mode) twice adds two entries to font-lock-keywords

[John Wiegley]

>>>>> Clément Pit--Claudel <clement.pit@gmail.com> writes:

> Do bug reports with a proposed fix go on the bug tracker, or on this list?
> I've seen patches discussed here, so I'm not too sure. Apologies if this is
> the wrong place (please do let me know!)

I prefer the bug tracker, so that they do not get lost. Issues and patches
deserve a "ticket number". This guarantees they will be looked at *some day*,
and focuses discussions around them; while messages only sent here may get
lost in the emacs-devel archives forever.

-- 
John Wiegley                  GPG fingerprint = 4710 CF98 AF9B 327B B80F
http://newartisans.com                          60E1 46C4 BD1A 7AC1 4BA2