@ShreevatsaR I think LMTX is not the name of the engine but the name of the Context version (or maybe the complete Context distribution) which uses luametatex.

For unicode-math, how can I write a single instance of the uppercase R for once, on the go, in a font other than the main math font without having to write \setmathfont[range=\BbbR] in the preamble since I don't need a permanent change?

For example, if I have \setmathfont{TeX Gyre DejaVu Math}[range={bb,bbit}, Scale=MatchUppercase] in the preamble, how can I type R in Cambria Math for once inside some equation?

@MarcelKrüger That's also how the manual uses these terms.

@ShreevatsaR Actually, there's ConTeXt MkII which can use pdfTeX or XeTeX as backend, although it's deprecated for a while now. That would be invoked by texexec.

@UlrikeFischer About?

@JosephWright did you define a mixed colorspace first?

ooh a duck

@UlrikeFischer Am I right in thinking that colorspace can't mix spot colours with arbitrary process colors? I just tried \color{red!50!foo}{Text} and got TeX capacity exceeded

@UlrikeFischer Errors with ... most combinations

@JosephWright did you define a mixed color space first?

@UlrikeFischer No, that was rather the point

@UlrikeFischer I was expecting the code to switch to the CMYK fallback as that's the only way to handle mixing here

@UlrikeFischer I'm working out what mixing approaches I need to set up: first, I want to cover \color{foo!50}, which needs to work

@JosephWright I don't think that it is so sophisticated. Don't forget that is has to hook into xcolor commands and use them and that isn't easy.

@UlrikeFischer Right, yes

@UlrikeFischer You can't define a mixed color space if the 'other' colors are not CMYK ...

@UlrikeFischer I'll work on bits and pieces today: probably I'll address teh 'set up' stuff first

@JosephWright well it depends a bit what you mean with "mixed colors". On the one side we have the "professional" view that someone will perhaps want to setup a very specific mix of pantone colors as a deviceN color space or whatever. On the other side we will have users which simply wants that "red!50!foo" doesn't explode and gives a more or less right color in a link border.

I have a tendency to say that "professional" colors should be predeclared as "mixcolor" and that expressions should use rgb or cmyk fallbacks.

@UlrikeFischer Yes, I see that

@UlrikeFischer I'll work on stuff: this is complex/tricky and likely needs me to alter how the current path works. At the moment, I'm not sure I've got an easy 'bailout' route, and that's needed if the target model might not even exist

@UlrikeFischer I was thinking 'mixed' spaces for use would be \color_model_new:nnn { name } { deviceN } { ... }

@UlrikeFischer Sounds about right: I think I'll look to implement that

@UlrikeFischer We might need to special-case the \color{foo!50} business: that does need to work. I think it's as soon as you hit a second colour (other than white), it's time to fallback to cmyk

@JosephWright yes. I just got noticed today about the context manual: pragma-ade.com/general/manuals/colors-mkiv.pdf

@UlrikeFischer Oooh, better read that

@UlrikeFischer I may need a few days to get this all in order: it's pretty tricky stuff

@UlrikeFischer Probably I'll try to add something for the object creation, but it's non-trivial to set up

@UlrikeFischer Perhaps with just 'tough, you have to have hooks in place'

@UlrikeFischer for deviceN, does a CMYK equivalent make sense? I'm thinking perhaps not, which impacts on the generic interface

\color_model_new:nnn {name}{family}{parameters} as command sounds right, the exact format of the parameters can then be dependant on the keyword/family. That should be quite flexible and extensible.

@UlrikeFischer Yup

@UlrikeFischer Just reading the bottom of page 4 of the colorspace manual: suggests mixing spot colors is doable in an expression ... or I think it is

@UlrikeFischer OK

@JosephWright imho we should have for every color at least a cmyk fallback. Not necessarly for the internal setup in the pdf but so that e.g. html export works and that hyperref can use it. So I would always setup some function that converts the deviceN values to cmyk (it doesn't need to give the "right" cmyk color, only some cmyk at all).

@UlrikeFischer yes although for hyperref/html rgb would be more natural than cmyk, but on the other hand the naive cmyk-rgb conversion is probably OK for those uses so cmyk fallback in all cases is probably OK

@JosephWright btw I looked a bit at svg. It seems to use a css-syntax like rgb-icc(255,255,0, #SpotColor,'Pantone 355-Green C',0.30, #CMYK,.7,0,0.5,.6) (and various variants). Should be not to difficult, if we get the generic color storing right.

@DavidCarlisle yes, rgb is more natural. The main question is if it makes sense to store for all colors always both.

@DavidCarlisle My initial thinking was these 'professional' colours are print-based, so a cmyk equivalent made most sense

@UlrikeFischer We can argue the approach here: we can convert rgb <-> cmyk and with multi-model support, the user could supply both

@UlrikeFischer Sure, that's a backend thing so it's unaffected by the set up I'm trying to get right just at the moment

@UlrikeFischer probably not worth it for most uses, just using cmyk or doing a simple translation is probably OK for uses other than professional printing

@DavidCarlisle My thinking too

@UlrikeFischer, @DavidCarlisle Expect more rebased/re-written PR data: I'm trying to avoid a long load of commits to change the plans :)

@JosephWright you also need to support Phil using direct \special syntax as documented last century for a driver he's not using

@DavidCarlisle well we sometimes see questions where people discuss the conversion, e.g. tex.stackexchange.com/questions/436012/…. So probably we need an interface where people can say exactly which rgb or cmyk they want and avoid automatic conversion.

@DavidCarlisle ;-)

@DavidCarlisle :)

@DavidCarlisle He's right about the need for clear docs: getting things right with dvipdfmx is interesting, I'm still not 100% sure about stroke versus fill there (PDF much much easier)

@UlrikeFischer There are two parts to that: we can provide 'multi-model' input for fixed conversions, plus we might want to expose the formulae (both xcolor and ConTeXt seem to)

@JosephWright exposing the formula is one side. The other is if a specific color definition should be possible: Say (mock up syntax) \definecolor{myred}{rgb=1 0 0, cmyk=0 1 0 0,spot1= 0.3} and then \color[rgb]{myred} would give 1 0 0 and \color[cmyk]{myred} magenta.

@UlrikeFischer My initial syntax plan was to go with the xcolor syntax, so {rgb/cmyk/spot1}{1 0 0 /0 1 0 0/0.3}, at least in part as it means we can just pickup the same inputs

@JosephWright yes I was surprised that the color stack push/pop specials we used in the graphics-def files did not seem to be in any of the (x)dvipdfm(x) manuals.

@DavidCarlisle I may have to rethink that for l3backend: depending on how we end up dealing with separations in PostScript, it might not work in dvipdfmx, so we'd need to use the PDF-based specials

@JosephWright yes syntax is fine, the question is, it is more difficult to handle later if rgb is sometimes there and then needs a conversion on use? Would it be easier if an optional conversion is done and stored directly at the color definition?

@JosephWright I have been wondering about this when I saw the question. How is it then with the color stack?

@UlrikeFischer I wonder if I should address this first, then get back to the models: in some ways, the 'support multiple models' part is easier (it's purely an internal question for l3color, not backend involved)

@UlrikeFischer That's the issue, from memory: that's why I used the dvips syntax for dvipdfmx

@JosephWright If is doesn't work, we should make a feature request. I want more than one color stack anyway.

@UlrikeFischer Probably, yes: colour support is much better in pdfTeX than elsewhere, and it's awkward to abstract some of it out

experts, I am still confused about all of this math in section titles. I read one can use math in section titles these days without having to use \cprotect. Then why when I add unicode-math package, it gives error, and when I remove it the error goes away? I only use lualatex. Here is MWE

\documentclass{article}
\usepackage{unicode-math}
\usepackage{amsmath,mleftright}
\usepackage{hyperref}
\begin{document}
\tableofcontents

\section{$\cos\left(  A+B\right)  $ and $\sin\left(  A+B\right)  $}%
test

answers for classic math mode can not be expected to always apply to unicode-math, but why are you using left and right here they just mess up the spacing) (amsmath should be loaded before unicode-math, although recent releases load it anyway

@DavidCarlisle but the main problem here is \sin and \cos. The hyperref code is quite a pain.

@UlrikeFischer yes I know I was about to blame you for that bit

@UlrikeFischer @Nasser simpler example is

\documentclass{article}
\usepackage{unicode-math}

\usepackage{hyperref}
\begin{document}
\tableofcontents

\section{$\cos $}%
test
\end{document}

@DavidCarlisle couldn't you blame me for something that I can solve instead??

@UlrikeFischer world pandemics?

@JosephWright I look at xdvipdfmx, spot color in itself works fine (with my example code). But as suspected the problem is the color stack. On the next side it is lost and the commands don't allow to add the color model only the sc value. I think I will write to the texlive list and look if someone has an idea.

@Nasser cprotect completly hides the argument before hyperref. It solves the error but destroys the bookmark content.

@Nasser were you using unicode-math in your example? (although I wouldn't use cprotect anyway there isn't a good reason to prefix every use in the document)

@DavidCarlisle No I was not using unicode-math in the first example. This must be why.

@Nasser but as Ulrike says cprotect isn't really a solution to your problem anyway.

@UlrikeFischer   But without unicode-math, I can't install the font you send me link to and use it as mainfont. I spend one hr and finally installed it, but same problem exist. MWE

\documentclass{article}
\usepackage{unicode-math}
\setmathfont{Asana-Math}[Scale=MatchLowercase]
\setmainfont{Baskervaldx-Reg.otf}
\usepackage{amsmath,mleftright}
\usepackage{hyperref}
\begin{document}
\tableofcontents

\section{$\cos x$}%
test
\end{document}

And when I add cprotect as I used to do before, like this

@Nasser you can use \usepackage{fontspec} (it is loaded by unicode-math) then \setmainfont will work.

@UlrikeFischer I'll try \usepackage{fontspec} to see if this will help with bookmarks not lost and see....

I have never seen something more confusing than using fonts in Latex.

But using the opentype font will not resolve the conflict between hyperref and unicode-math. If you want to use unicode-math, and keep the bookmarks intact you have currently only the option to use \textorpdfstring (Mico's answer) or \pdfstringdisablecommand (egreg's answer).

@Nasser ^^^

@UlrikeFischer but I was following your suggestion in your comments to try the OPENTYPE version of Baskervaldx since you said the other one is meant to work only in pdflatex and not lualatex.

@Nasser yes, but this was only a side-remark. Using the package is wrong, but it is not related to the hyperref/unicode-math conflict.

@UlrikeFischer Ok. Then for now Mico solution is only one to use. No other options for me. I can't do \pdfstringdisablecommand. There should really be build-in solution in Latex for this. I mean sometimes one needs to have math in section titles, and keep bookmarks without having to go through all this gymnastics just to do this.

@Nasser saying something should work isn't a very constructive comment.

@DavidCarlisle "something should work isn't a very constructive comment" humm. Then what should one say instead? Not say anything if something does not work? If you mean I should pull my sleeves and code something, I do not know inner latex coding at all. If Latex was Pascal or any other normal looking language, yes, but Latex syntax is too hard and for me is not a normal programming language. I have no idea how you experts do these things all day.

@Nasser we don't do them all day, we do them in free time in the evenings, if we have free time. (most of us)

@DavidCarlisle I think Latex programmers must be born with that skill. Like LISP programming. It needs some special kind of brain to handle the strange syntax that one must be born with. I could not get used to Latex inner syntax myself, even though I been using it (as user only) for long time now.

@Nasser that seems unlikely, but anyway I didn't say that other people should do the coding but just that your choice of words was probably not great way to ask someone to spend some hundreds of hours of volunteer time coding a feature that you happen to need today.

@DavidCarlisle this is here not longer explodes:

\documentclass{article}
\usepackage{unicode-math}

\usepackage{hyperref}
\usepackage{etoolbox}
\makeatletter
\robustify\operator@font
\makeatother

\begin{document}
\tableofcontents

\section{$\cos $}%
test
\end{document}

@UlrikeFischer yes was wondering where um_foo could be made safer. Suggest to Will?

@DavidCarlisle could there be any reason why this is not defined as protected?

@UlrikeFischer well... I suppose in theory something could be expecting it to expand but not that I know of

@UlrikeFischer the classic definition is safe in an edef so arguably the unicode-math one ought to be as well

@DavidCarlisle but is the unicode-math definition expandable? Doesn't look so. If I do \edef\blub{\operator@font} it explodes.

@DavidCarlisle I will add an issue to unicode-math.

@UlrikeFischer I think i'm going to tackle multi-model support then come back to spot colours: I think that might help me clean up some other code

@JosephWright ok, I think that is right.

> The Calvin Klein cologne "Obsession" can attract big cats like jaguars, and tigers, from as far as half a mile away. They then proceed to take long sniffs and cuddle against the source, enjoying the smell longer than they enjoy some of their meals.

WAT

@PauloCereda might as well combine enjoying the smell with enjoying the meal.

@Skillmonlikestopanswers.xyz oh no

@UlrikeFischer that's what I mean, the classic one is safe in an edef unicodfe-math not, so using \protected seems right

@DavidCarlisle I added an issue github.com/wspr/unicode-math/issues/550

@UlrikeFischer yes got the mail thanks, just catching up after duck lunch

@DavidCarlisle oh no

(sorry I was late)

@UlrikeFischer ooh

@barbarabeeton twitter.com/notfunnyelle/status/1270118565564182528

@JosephWright @DavidCarlisle ^^

@PauloCereda -- Skunks are actually quite charming critters, as long as you don't threaten or spook them. If you're going to have one as an indoor companion, it's probably safer for you to have it descented, but then you've removed its protection against other wild critters, and you probably shouldn't let it run free.

@UlrikeFischer ^^ ooh bears

@barbarabeeton ooh they look charming indeed!

@PauloCereda Don quack, ésta imagen es para usted

@PauloCereda Now do that with Australia.

ooh pizza

@HenriMenke ooh kiwis

@PauloCereda :)

@PabloGonzálezL by the way, we are trying to investigate your curious case... so far, we could not reproduce anything... :( I am still looking for an explanation, I will setup a Windows VM...

@HenriMenke ^^

@PauloCereda I think I have the explanation, but I must try again... the only difference between my installations is Java...in windows/fedora it is Oracle, in Ubuntu it is openjava

@PabloGonzálezL Hm it shouldn't make a difference, but in that case I can easily test it. I have literally dozens of Java versions and vendors. :)

@PauloCereda openstreetmap.org/node/1791514845

@HenriMenke OH MY

@PauloCereda It is the only thing that differentiates WSL and win10, in everything else it is the same (TeXLive, etc)

@PauloCereda openstreetmap.org/relation/10627335

@PabloGonzálezL I will check, thanks for the hint. :)

@HenriMenke HOLY COW

It's yet another reason to visit Will :)

And @HenriMenke <3

@PauloCereda Anyway, removing oracle java from fedora 32 has been a headache...

@PabloGonzálezL You mean, you miss Oracle Java?

I think it's been ages since I don't use an Oracle VM... I've been using OpenJDK since version... 6 I guess.

@PauloCereda Have you ever tried out GraalVM? It has a Java JIT and an ahead-of-time compiler so you can compile Java to native code (apparently).

@HenriMenke oh I tried a few examples, but never tried a "real world" case! It's indeed a very nice approach! We might adopt something similar in the future with Kotlin native...