Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

fontspec update breaks lualatex?

95 views
Skip to first unread message

Rembrandt Wolpert

unread,
Sep 29, 2015, 10:03:41 PM9/29/15
to
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA256

The simple script

\documentclass[11pt,twoside,letterpaper]{article}
\usepackage{luatextra}
\defaultfontfeatures{Ligatures=TeX}
\newfontfeature{Microtype}{protrusion=default;expansion=default;stretch=
20;shrink=20;step=1}
\directlua{fonts.protrusions.setups.default.factor=.3} % .4, .5

\setmainfont[Microtype,ItalicFont={Trinite No2 Italic Cond},
BoldFont={Trinite No2 Roman Cond}]{Trinite No2 Roman Con
d}

\begin{document}

A simple sentence.

\end{document}

fails with:

! Undefined control sequence.
<argument> \xetex_supressfontnotfounderror: D

1.9 ... nite No2 Roman Cond}]{Trinite No2 Roman Cond}

The previous fontspec work/ed/s fine. Any advice apart from
downgrading to previous fontspec?

Many thanks,
Rembrandt

This is on Mac OSX

-----BEGIN PGP SIGNATURE-----
Comment: GPGTools - http://gpgtools.org

iQEcBAEBCAAGBQJWC0L0AAoJEFWqP1hIXLxauiQH/A35WuA6url3dPCQw/LikZvg
5zHNcfY7asfLZRNaXRAATlzHe28jzITFuZnu+tabuNaMKBNGI8HHsQ/+jpv5Rwod
eo07dhQ+gvJ9seBhRSIHlSCQdw0OTKNwoVtcPizCBNCV8N6vbicun6aj6OvTSkw0
0nFZ1SN1sdodKcCGY1pNqIvayl3a6fNzTaEAl+z0tN4z0+KMc0poPslqZ9wm4Fqs
BRT1qXE/oKgr2ZlvKxkZvmGfXK8uXu39xjvH+M+xvXuehEZyTinZmIPHrRERxCCv
3Swh8pGkHu689dRAttMdP+Uwow5m6HuT7GX4P/jBeb0mESJIhPU8aKFXXFOr6Lc=
=MED2
-----END PGP SIGNATURE-----

Denis Bitouzé

unread,
Sep 30, 2015, 1:52:13 AM9/30/15
to
Le 30/09/15 à 04h03, Rembrandt Wolpert <r.f.w...@gmail.com> a écrit :

> The simple script
>
> \documentclass[11pt,twoside,letterpaper]{article}
> \usepackage{luatextra}
> \defaultfontfeatures{Ligatures=TeX}
> \newfontfeature{Microtype}{protrusion=default;expansion=default;stretch=
> 20;shrink=20;step=1}
> \directlua{fonts.protrusions.setups.default.factor=.3} % .4, .5
>
> \setmainfont[Microtype,ItalicFont={Trinite No2 Italic Cond},
> BoldFont={Trinite No2 Roman Cond}]{Trinite No2 Roman Con
> d}
>
> \begin{document}
>
> A simple sentence.
>
> \end{document}
>
> fails with:
>
> ! Undefined control sequence.
> <argument> \xetex_supressfontnotfounderror: D
>
> 1.9 ... nite No2 Roman Cond}]{Trinite No2 Roman Cond}
>
> The previous fontspec work/ed/s fine. Any advice apart from
> downgrading to previous fontspec?

No need to downgrade. A fix in on the way but, in the meantime, you can
just add the following lines before loading `fontspec':

--8<---------------cut here---------------start------------->8---
\ifdefined\suppressfontnotfounderror
\expandafter\let\csname xetex_suppressfontnotfounderror:D\endcsname
\suppressfontnotfounderror
\else
\expandafter\let\csname xetex_suppressfontnotfounderror:D\endcsname
\luatexsuppressfontnotfounderror
\fi
--8<---------------cut here---------------end--------------->8---

More on this here:

┌────
http://tex.stackexchange.com/questions/269786/unicode-math-broken
└────

> Many thanks,

You're welcome.
--
Denis

Joseph Wright

unread,
Sep 30, 2015, 3:47:22 AM9/30/15
to
Fix is now in TL: sorry for the issue.

Joseph

--
Joseph Wright

Jeffrey Goldberg

unread,
Sep 30, 2015, 12:01:23 PM9/30/15
to
On 2015-09-30 07:47:16 +0000, Joseph Wright said:

> Fix is now in TL: sorry for the issue.

Thank you for the quick fix and for all that you and the team do in
developing and maintaining fontspec and unicode-math.

But now the whining: This is the second time in a couple of months that
an update broke things for me (and presumably a lot of people) because
the developers didn't test against current TeXLive, but seem to have
tested against stuff in the mysterious future.

And now for aimless musings:

I'm working on a project for a fairly complicated document in which I am
the only one who actually has used TeX. So I provide a nice Makefile for
my colleagues who are in no position to cope with TeX error messages.

One of the things that I've learned in working on a complicated document
with non-TeXies is how much I am looking forward to LaTeX3. I made a
poor initial choice of basing my class on tufte-book, which really
doesn't play well with other packages. As a consequence, instead of
being able to show off to my colleagues, "Hey, look. This package makes
it easy to do some stuff that makes our lives easier", I have to try to
patch things or reconstruct them.

Of course the problems of packages not playing well together isn't
limited to Tufte LaTeX, but it provides an extreme example. LaTeX3 isn't
going to make the problem go away overnight, but it is going to stop the
problem from getting worse.

So, again, I really appreciate that fontspec and unicode-math are being
designed with LaTeX3 conventions. But that is still in the mysterious
future for many production systems.

Cheers,

-j

Joseph Wright

unread,
Sep 30, 2015, 12:56:40 PM9/30/15
to
On 30/09/2015 17:01, Jeffrey Goldberg wrote:
> But now the whining: This is the second time in a couple of months that
> an update broke things for me (and presumably a lot of people) because
> the developers didn't test against current TeXLive, but seem to have
> tested against stuff in the mysterious future.

We (the LaTeX team) do work pretty hard to test both against the current
release of third-party stuff, both in terms of 'looking forward' and
'looking back' changes (i.e. do changes we want to make break other
packages, do changes in other packages break with updates to expl3,
etc.) Here, 'third party' includes stuff such as fontspec written by
members of the team but not part of the code maintained by the team as a
whole.

Unfortunately, we do miss stuff. In the current case, I'd made a change
anticipating some issues we know might otherwise come up, Will spotted
something not quite right and I changed some code but forgot there we
*two* places I needed to do that. As we'd already got things 'just about
ready' and as there is other stuff we are testing at the same time, this
slipped past both of us.

Partly for this reason we've been working (again) to extend our test
suite for LaTeX3 code. We've now added an automated test system, which
we are slowing going to grow to give us at least basic coverage of
third-party material, especially commonly-used packages such as
fontspec. This should reduce the likelihood of mistakes in this area.
[See https://travis-ci.org/latex3/latex3 for current status,
https://github.com/latex3/latex3 for the current code including
development fixes, etc.] Additional tests which fit into the framework
are always welcome: the LaTeX-L list is the best place to suggest new ones.

[We are well-aware that fontspec in particular is more-or-less essential
for using XeLaTeX and LuaLaTeX.]

Importantly, one of the reasons for the changes we wanted to make is
that there will be a new LaTeX2e release *tomorrow* which is aimed at
addressing long-term support for LuaTeX. That will be an overall benefit
to users, and to developers of LuaLaTeX packages, but it will need some
adjustment to code using LuaTeX. A number of members of the team have
been working with package authors to get this all in place.

I think it's important to note that most of the time either the team or
individual package authors (as appropriate) fix issues within 24-48 h.
The team are extremely mindful of the need for reliability with existing
documents, but there is a balance between that and allowing for the fact
that XeTeX and LuaTeX are under active development both at the engine
level and in macro support, and changes are needed at the format level,
etc., to support that. That's a key reason we are making a format update.

Joseph

Jeffrey Goldberg

unread,
Sep 30, 2015, 7:10:11 PM9/30/15
to
On 2015-09-30 16:56:32 +0000, Joseph Wright said:

> On 30/09/2015 17:01, Jeffrey Goldberg wrote:
>> But now the whining: This is the second time in a couple of months that
>> an update broke things for me (and presumably a lot of people) because
>> the developers didn't test against current TeXLive, but seem to have
>> tested against stuff in the mysterious future.
>
> We (the LaTeX team) do work pretty hard to test both against the current
> release of third-party stuff, both in terms of 'looking forward' and
> 'looking back' changes (i.e. do changes we want to make break other
> packages, do changes in other packages break with updates to expl3,
> etc.)

Thank you for that. And now that I've taken a look at the expl3 docs,
I can imagine how difficult that can be.

I've just had a bad day and was particularly whiny.

>
> Unfortunately, we do miss stuff.

It happens. I understand.

> In the current case, I'd made a change
> anticipating some issues we know might otherwise come up, Will spotted
> something not quite right and I changed some code but forgot there we
> *two* places I needed to do that. As we'd already got things 'just about
> ready' and as there is other stuff we are testing at the same time, this
> slipped past both of us.

I understand. I've broken things that way, too.

>
> Partly for this reason we've been working (again) to extend our test
> suite for LaTeX3 code. We've now added an automated test system, which
> we are slowing going to grow to give us at least basic coverage of
> third-party material, especially commonly-used packages such as
> fontspec.

Excellent.


> This should reduce the likelihood of mistakes in this area.
> [See https://travis-ci.org/latex3/latex3 for current status,
> https://github.com/latex3/latex3 for the current code including
> development fixes, etc.] Additional tests which fit into the framework
> are always welcome: the LaTeX-L list is the best place to suggest new ones.

I have to confess that I haven't really been paying much attention. I've
only recently started to move from pdf(la)tex to xe(la)tex, from bibtex
to biblatex/biber. All of these are much needed change, but it also means
that I struggle with compatibility issues with tools that I don't fully
grok.

(I've also shot myself in the foot by basing my current class on tufte-book,
so even in pure 2e terms, I've made package competibiltiy difficult.)

> Importantly, one of the reasons for the changes we wanted to make is
> that there will be a new LaTeX2e release *tomorrow* which is aimed at
> addressing long-term support for LuaTeX. That will be an overall benefit
> to users, and to developers of LuaLaTeX packages, but it will need some
> adjustment to code using LuaTeX. A number of members of the team have
> been working with package authors to get this all in place.

Cool.
>
> I think it's important to note that most of the time either the team or
> individual package authors (as appropriate) fix issues within 24-48 h.

Yes. It is great.

It just hit me at a frustrating time.

And the more I look at expl3 and learn about LaTeX3, the happier I get.

So thanks again, and thanks for the insight into the process.

Cheers,

-j


0 new messages