script and lang missing in font definition.

[u-fischer]

Description

fontspec (2019/01/18 v2.6l) writes "Font "OpenSans-Regular" does not contain any OpenType `Script' information." and then doesn't add script and language tag.

Check/indicate

• [ x] Relevant for XeTeX
• [ x] Relevant for LuaTeX
• [ x] Issue tracker has been searched for similar issues?

Minimal example demonstrating the issue

\documentclass{article}
\usepackage{fontspec}
\setmainfont{OpenSans-Regular.ttf}
\begin{document}
blub
\end{document}

Further details

I 'm not sure if the current behaviour is actually more correct, but it is certainly a change. For some fonts (I found opensans, times and arial) I get no script and language tag in the font features.

The log-file says

Package fontspec Info: Font "OpenSans-Regular" does not contain any OpenType
(fontspec)             `Script' information.


Package fontspec Info: Font family 'OpenSans-Regular.ttf(0)' created for font
(fontspec)             'OpenSans-Regular.ttf' with options [Ligatures=TeX].
(fontspec)              
(fontspec)              This font family consists of the following NFSS
(fontspec)             series/shapes:
(fontspec)              
(fontspec)             - 'normal' (m/n) with NFSS spec.:
(fontspec)             <->"[OpenSans-Regular.ttf]:mode=node;+tlig;"

Trying to force the script with e.g. \setmainfont{times.ttf}[Script=Greek] doesn't work.

otfinfo says:

$ otfinfo -s OpenSans-Regular.ttf
latn            Latin
latn.MOL        Latin/Moldavian
latn.ROM        Latin/Romanian

For other fonts it still works. The main difference I saw is that working fonts have DFLT in their script list.

[u-fischer]

Additional remark: I do have in the luaotfload testsuite suddenly also quite a number of test failures related to unicode-math. A number of boxes have suddenly a different width. A possible cause could be that the math fonts are loaded without the script=math feature:

[texgyrepagella-math.otf]:mode=base;

And the difference is visible:

[RuixiZhang42]

I can confirm that this is a bug introduced by v2.6l.

% !TeX program = XeLaTeX
\documentclass{article}
\usepackage{unicode-math}% fontspec v2.6l
\setmathfont{TeX Gyre Pagella Math}
\begin{document}
v2.6l: $E_{E_E} 2_{2_2}$
\end{document}

Optical sizes in the OpenType math font no longer work :(.

[wspr]

Thanks for the detailed analysis. You're right, it is a change, and it is indeed a bug :( I think I have this fixed now in the working code but want to make doubly sure everything is working properly with unicode-math before I release...

[wspr]

If anyone has the capacity to check the code I'd be much obliged. I'm wary of releasing too soon and missing more bugs!

[u-fischer]

I did run my (luaotfload)-testsuite with the new version (so only with luatex ...). The unicode-math failures are gone, I have one failure because a font id number has changed, but that's imho harmless. I also get a failure because scaling has slightly changed:

\documentclass{article}
\usepackage{fontspec}
\setmainfont{texgyretermes-regular.otf} 
\begin{document}
\fontspec[Scale=MatchLowercase]{texgyrecursor-regular.otf} 

\typeout{font dimen 5: \the\fontdimen5\font}

\typeout{font dimen 8: \the\fontdimen8\font}

\end{document}

This output with 2019/01/18 v2.6l

font dimen 5: 4.49998pt
font dimen 8: 6.07552pt

and with the experimental code

font dimen 5: 4.50005pt
font dimen 8: 6.07559pt

[RuixiZhang42]

@u-fischer 0911d75 contains the exact same changes in font size, e.g., in feat-scale-match.luatex.tlg and feat-scale-match.xetex.tlg. The change has to do with the 1sp, 2sp and 3sp introduced in 1c1f631, f0d6341 and 0911d75. I suspect this is the desired change.

[u-fischer]

@RuixiZhang42 Ah yes. My testfile is a copy from this one.

[wspr]

The ± 1sp, 2sp, 3sp thing was added as an additional measure to avoid XeTeX “caching” the fontname — strictly I could remove them but I hoped the differences were small enough that they wouldn’t affect too much in practice. I could revert if you think even such small changes are problematic :)

[u-fischer]

I personally don't care much about the scaling difference. But I think an update would be good as the missing script is starting to show e.g. https://tex.stackexchange.com/questions/471565/realscripts-fontspec-update-vs-some-opentype-fonts and https://tex.stackexchange.com/questions/471184/problem-with-prime-and-unicode-math-package

[wspr]

Scaling is back to normal. New release sent to CTAN.