Hi Morgan,
Yes, the reproducer works for me. I re-ran it today from emacs -Q with
both GNU Emacs 31.1 and 32.0.50, and Org 10.0pre0.20261108.969:
step 1 (only a): (("\\(a\\)" "currentColor"))
step 2 (a cached, b new): (("\\(a\\)" "currentColor") ("\\(b\\)" "NO FILL"))
One thing that matters: it must start with an empty preview cache
(hence the fresh org-persist-directory in the command). If \(b\) is
already cached from an earlier run, it is not regenerated and the bug
does not show up. That may be why the previous reproducer was hard to
trigger.
Command (the script and the patch are attached again):
REPRO_DIR=/tmp/repro emacs -Q --batch \
--eval '(progn (add-to-list (quote load-path) "/path/to/org/lisp")
(setq org-persist-directory "/tmp/repro/persist/"))' \
-l org-latex-preview-repro.el
Interactively: with a dark theme, preview a fragment, then type a new
fragment after it in the same paragraph and preview again. The new one
is rendered black.
With the attached patch, step 2 gives currentColor for both fragments.
Thanks,
Niccolò
On Wed, Nov 11, 2026 at 3:55 PM Morgan Smith <[email protected]> wrote:
>
> > With the new org-latex-preview and the dvisvgm process, a fragment can be
> > rendered without any color, so it shows black (invisible on a dark theme).
> > It happens when the fragment is generated in the same call as a fragment
> > before it that is already cached, and both have the same foreground and
> > background.
>
> This sounds a lot like the following report so I'm going to combine the
> discussions.
>
> Supersedes:
> https://list.orgmode.org/CALtzAB03oDSsw=LThTTbuNFXVfsL9rbjqM079n66v=oPxSZ=h...@mail.gmail.com
>
> Karthik was unable to get the reproducer from the previous report to
> work. Can you get this report's reproducer to work?
>
> Thanks,
>
> Morgan