Unfortunately v2.14 won't help because it was just a rebuild and has no new features.  We are working on a major overhaul of the MATH table that will address the font-specific problems.

In the meantime, you might want to look at

    https://sourceforge.net/p/xetex/bugs/175/

and

    https://sourceforge.net/p/xetex/bugs/173/

and the comments in the implementation of get_ot_math_kern() in

    https://github.com/TeX-Live/texlive-source/blob/trunk/texk/web2c/xetexdir/XeTeXOTMath.cpp

Note especially comment (3) in the superscript branch.

Cheers,
David.




From: luigi scarso <luigi.scarso@gmail.com>
Sent: Tuesday, March 18, 2025 8:51 AM
Cc: dev-luatex@ntg.nl
Subject: [Dev-luatex] Re: Imprecise language in the OpenType Math specification leading to implementation differences amongst math engines
 


On Tue, 18 Mar 2025 at 13:42, Mikael Sundqvist <mickep@gmail.com> wrote:
Hi Apoorv,

On Tue, Mar 18, 2025 at 7:17 AM Apoorv Potnis <apoorvpotnis@gmail.com> wrote:
>
> Dear Developers of LuaTeX,
>
> I am writing this mail to report a bug in the OpenType Math specification regarding imprecise language, found by Khaled Hosny: [MATH] The algorithm for applying MathKernInfo is incorrect · Issue #1147 · MicrosoftDocs/typography-issues.
>
> LuaTeX follows the letter of the specification, while XeTeX follows the intent of the specification. This has created compatibility issues for font developers.
>
> Thank you.
>
> --
> Apoorv Potnis

Can you please provide an example where it is clear that the problem
is the luatex engine and not the font? That is, one example of this
situation that shows well in xetex but not in luatex.

The reason that I ask, is that the examples I have considered so far,
where this seems to matter (r_j for example in stixtwomath in TL25),
the xetex output is not good either. I suspect that the corner kerns
are not good in the font there, and therefore ask for a reasonable
example to test this on (in TL25).


Just seen
STIX Two, v2.14 INTERIM Latest
tiroj released this 18 hours ago


--
luigi