Strange interaction between beamer, siunitx and unicode-math

[itoijala]

The following:

\documentclass{beamer}
\usepackage{siunitx}
\usepackage{unicode-math}
\begin{document}
\begin{frame}
  $\mathbf{foo}$ % works with \symbf
  \setmathfont{XITS Math}
  $x$
\end{frame}
\end{document}

gives the error:

Runaway argument?
9@nil try@load@fontshape expandafter ifx csname curr@fontshape endcsname ETC.
! File ended while scanning use of \split@name.
<inserted text> 
par 
<*> foo.tex

[josephwright]

As you might expect, siunitx is only parenthetically involved here: you can get the same effect with

\documentclass{beamer}
\usepackage{unicode-math}
\begin{document}
\begingroup
  \setbox0=\hbox{\ensuremath{\mathsf{}}}
  \setbox0=\hbox{\ensuremath{\mathtt{}}}
\endgroup
\begin{frame}
  $\mathbf{foo}$ % works with \symbf

  \setmathfont{XITS Math}$x$
\end{frame}
\end{document}

(the method siunitx uses to pick up the active math font families). I'll see if I can trace this further.

[CommunicationAnimale]

It seems that the same issue prevents using \smashoperator or \cramped (from mathtools) even outside beamer :

\documentclass{article}

\usepackage{mathtools}
\usepackage{unicode-math}

\begingroup % or \usepackage{siunitx}
\setbox0=\hbox{\ensuremath{\mathsf{}}}
\setbox0=\hbox{\ensuremath{\mathtt{}}}
\endgroup

\begin{document}
\begin{align*}
    Z_\beta &= \smashoperator{\sum_{\sigma \in \left\{-1, +1\right\}^N}} e^{-\beta H(\sigma)} \\
    Z_\beta &= \sum_{\sigma \in \left\{-1, +1\right\}^N} e^{-\beta H(\sigma)}
\end{align*}
\end{document}

gives

Removing the \begingroup ... \endgroup statement makes the black bar go away.

[wspr]

I'm sleep deprived but I've looked at this again and can't see where the issue is coming from. Will continue to investigate. (My answer to the OP would be: don't change the maths font mid-slide!)

Thanks @CommunicationAnimale for the additional comments.

[wspr]

@CommunicationAnimale — I've just tested your example with the most recent release of the package and it now appears fixed. Hopefully you see the same — please let me know (by opening a new issue) if you see it re-occur.