odt writer results in misinterpreted formulas in LibreOffice

[cryptarch]

Explain the problem.

I have a LaTeX document which uses the amsmath align environment. I need to convert it for inclusion in a LibreOffice odt document. Due to the way LibreOffice interprets MathML as formulas, characters are rendered as upside-down question marks, indicating missing arguments.

Example source:

\documentclass{article}
\usepackage{amsmath}
\begin{document}
hello world
\begin{align*}
    y &= {(x+1)}^2 \\
      &= x^2 + 2x + 1.
\end{align*}
\end{document}

Example output generated by pdflatex:

Command for conversion from LaTeX to odt:

pandoc -f latex -o main.odt main.tex

How the converted document looks in LibreOffice Writer:

LibreOffice documentation indicates this is expected behavior from their side:

• Formulas over multiple lines says to use the newline operator for multiline formulas.
• Aligning formulas using equals sign says to provide dummy arguments to the equals "operator" using a {}={} pattern. (This more closely matches my use case compared to newline.)

Therefore, unlike #6714 or #9235, I don't think the incorrect rendering would be accepted as a valid report by them this time.

I believe pandoc should be taught about LibreOffice's newline operator and/or the {}={} trick when generating mathml for odt. I understand if this isn't made default behavior; I suggest a flag such as --libreoffice-compat or so could accommodate the other software's expectations.

Pandoc version?
3.7.0.2

[jgm]

For reference, here is the MathML produced by pandoc and included in the ODT:

<math display="block" xmlns="http://www.w3.org/1998/Math/MathML">
  <mtable>
    <mtr>
      <mtd columnalign="right"
      style="text-align: right; padding-right: 0">
        <mi>y</mi>
      </mtd>
      <mtd columnalign="left"
      style="text-align: left; padding-left: 0">
        <mo>=</mo>
        <msup>
          <mrow>
            <mo stretchy="false" form="prefix">(</mo>
            <mi>x</mi>
            <mo>+</mo>
            <mn>1</mn>
            <mo stretchy="false" form="postfix">)</mo>
          </mrow>
          <mn>2</mn>
        </msup>
      </mtd>
    </mtr>
    <mtr>
      <mtd columnalign="right"
      style="text-align: right; padding-right: 0"></mtd>
      <mtd columnalign="left"
      style="text-align: left; padding-left: 0">
        <mo>=</mo>
        <msup>
          <mi>x</mi>
          <mn>2</mn>
        </msup>
        <mo>+</mo>
        <mn>2</mn>
        <mi>x</mi>
        <mo>+</mo>
        <mn>1</mn>
        <mi>.</mi>
      </mtd>
    </mtr>
  </mtable>
</math>

LibreOffice does not seem to be rendering this correctly, and that does seem to be a bug (in LibreOffice), since the spec for ODT calls for math to be represented in MathML. The documentation you cite is about some kind of user-facing markup that is not MathML and not part of the internal representation and not in any case produced by pandoc.

If there is a problem with the XML pandoc produces, we would be glad to hear it.

[jgm]

Here is how this MathML renders in Firefox, for comparison: