[XeTeX] Bug: unicode-math's \stemathfont segfaults on TeX Live 2009

[Gerrit Ansmann]

Hello,

following the recommendation of Will Robertson I am describing my problem here. A former post can be found here: http://github.com/wspr/unicode-math/issues#issue/96

I am using Debian Squeeze with TeX Live 2009 installed not via Debian’s package management system, but with the net installer (full install).
 From this I force-removed fontspec with tlmgr, just to install a recent version (GitHub or CTAN, I tried both) according to instructions (which included extracting euenc.tds.zip into the TDS tree at that time). I furthermore updated xltxtra to the GitHub version and installed unicode-math from GitHub according to instruction.

Now compiling

\documentclass{article}
\usepackage{xltxtra}
\usepackage{unicode-math}
\setmathfont{Asana Math}

\begin{document}
test $\Omega ξ + a \sum ε \cdot ∃ $
\end{document}

with xelatex ends in a Segmentation Fault:
[…]
(/usr/local/texlive/2009/texmf-dist/tex/latex/fontspec/fontspec.cfg)))
(/usr/local/texlive/2009/texmf-dist/tex/xelatex/unicode-math-table.texSegmentation fault

After removing the \setmathfont command it compiles without errors. (I assume it to be expected behaviour, that the output of the math part is reduced to $+ a$.)

If I use \setmathfont within the document this still happens and if I insert nonsense commands before this, the compiler first complains about them (and segfaults after I accept the complaint)—that’s why I am pretty sure, the problem iswithin \setmathfont. Using other fonts (Neo Euler, XITS Math) does not help either, nor does using \setmathfont{Asana-Math.otf} instead.

This did not occur on a TeX Live system installed on the same system via Debian’s package manager and with a unicode-math version from half and a week ago (3ca9b3f) as well as the fontspec version shipped with it. Said unicode-math version results in the same problems on the manually installed TeX-Live, so do newer ones.

Since another user talked about a similar problem (http://github.com/wspr/unicode-math/issues/closed#issue/88 — last post), I assume this not to be highly individual.

I am not on the mailing list right now, but have no problem to join, if this is wished for better communication.

Greetings,
Gerrit Ansmann