Sorry for the noise : the bugs do exist, but my previous analysis was false (one paren too much).
More data points : * I installed also MathToWeb, and configured it in emacs (via file-local-variables), with the same problematic result. * The error message "LaTeX to MathML converter not configured" appears exactly twice in org.el: + In org-format-latex + In org-create-math-formula * In both cases, the call has the form : ================================================================= (unless (org-format-latex-mathml-available-p) (user-error "LaTeX to MathML converter not configured"))) ================================================================= * When evaluated from M-: (org-format-latex-mathml-available-p) returns t (as it should...). * I have been so far unable to trace the origin of the faulty evaluations. I'll probably have to run an uncompiled org for this, and I don't (yet) know how to do that. Interim hypothesis : the relevant variables (org-latex-to-mathml-jar-file and org-latex-to-mathml-convert-command) may be inadvertently shadowed somewhere upstream in the call tree. I'm afraid to be a bit out of my depth here... Suggestions ? -- Emmanuel Charpentier Le lundi 24 décembre 2018 à 01:24 +0100, Emmanuel Charpentier a écrit : > Dear list, > > I have a couple of gripes about LaTeX handling in ODT export when one > needs MathML conversion, and therefore a couple of questions: > > The first one is possibly a gripe with the ox-pandoc exporters. The > doc > (info manual) states : > > "Add this line to the Org file. This option is activated on a per- > file > basis. > > #+OPTIONS: LaTeX:t" > > It seems that this option is *NOT* handled by the ox-pandoc exporters > : > when using one of these, the text "LaTeX:t" appears at the top of the > output. > > Hence my first question : what should be responsible for handling > this > option : org mode or ox-pandoc ? > > The second one is that I can't get org to acknowledge the > configuration > of org-latex-to-mathml-convert-command. A bit of peeking int the > source > leads to this function, responsible for testing it : > ===================================================================== > == > (defun org-format-latex-mathml-available-p () > "Return t if `org-latex-to-mathml-convert-command' is usable." > (save-match-data > (when (and (boundp 'org-latex-to-mathml-convert-command) > org-latex-to-mathml-convert-command) > (let ((executable (car (split-string > org-latex-to-mathml-convert-command)))) > (when (executable-find executable) > (if (string-match > "%j" org-latex-to-mathml-convert-command) > (file-readable-p org-latex-to-mathml-jar-file) > t)))))) > ===================================================================== > == > > As written, I can't see how one can use latexmlmath WITHOUT having > ALSO > the MathToWeb jar file somewhere : the test for it is systematic ; > therefore, it fails if the jarfile isn't configured/available. > > The value of executable should be tested, and the test for the > jarfile > should be done *only* if executable is "java" (or, better, belongs to > a > list of "reasonable" java executables or scripts...). > > So my second question is : what do you think ? > > HTH, > > -- > Emmanuel Charpentier > >