Here is a hard to debug problem: Equations system with a black rectangle in LaTeX. Turns out the problem arose from a stray \rowcolor command far earlier than the appearance of the problem. Hacker News discussion.

@UlrikeFischer chat.stackexchange.com/transcript/message/54762959#54762959 should now be fixed

@UlrikeFischer I'm hoping that the color export and multi-model stuff is now right: should be down to having to round out Separation/DeviceN support

@UlrikeFischer Please test :) I'd love to be able to merge ...

I see that TUG2020 has over 100 people down to 'attend'

@JosephWright I'm just looking at the changes. Why is the tip of color-models not in color-multi-models?

@HaraldHanche-Olsen In my (very humble) opinion this is a bug, rowcolor is doing something strange with the color stack. Expecting the row to unset the thing, and the fact it does not exist fumbles things...

@UlrikeFischer I've not merged across, that's all: I want to look at Bruno's cctab changes next

@JosephWright I'm now installing the color-multi-models branch and will try to run some tests.

@UlrikeFischer Great :)

@UlrikeFischer Like I say, I hope we have the multi-model stuff now working, and just the details of 'new model' stuff to sort

@JosephWright Don't you mean "people, bears and ducks"?

@samcarter_is_at_topanswers.xyz Indeed: beings

@samcarter_is_at_topanswers.xyz Why do Bär and the professore not count as people?

@UlrikeFischer Is Bär giving a talk? ;)

@UlrikeFischer The very important guests have to be named explicitly, they are good advertisements to make others register just to see them

@JosephWright Yeah, it's so nice!

@samcarter_is_at_topanswers.xyz regarding the question about tikz with expl3: the replace commands must expand the \l_bb_tl, e.g: \exp_args:NNnx\tl_replace_once:Nnn \l_aa_tl {<} { \exp_not:N\tikzmark{\l_bb_tl} }.

@UlrikeFischer Thanks! I'll pass this on. (Maybe @Skillmonlikestopanswers.xyz or JouleV can write up an answer?)

@Rmano there can't be a bug in \rowcolor

@DavidCarlisle :)

@DavidCarlisle Oops, you're right :O

@DavidCarlisle \CT@end should probably contain a \global\let\CT@do@color\relax? Or would that have bad side effects?

@UlrikeFischer no idea what the code does, can't look now:-)

@DavidCarlisle well we could wait if someone opens an issue or if they prefer to mention the problem only in blog entries in the wide internet ...

@UlrikeFischer if they raise an issue it could go in the queue behind all the longtable ones

@UlrikeFischer I need to transfer it to latex3 or set up a new gh organisation so that I can trick slaves to do the work recruit helpers to fix issues.

@HaraldHanche-Olsen ^^^^^^^

@DavidCarlisle Where was the issue registered?

@DavidCarlisle ooh row row row your colour gently down the stream

@UlrikeFischer The user says thanks!

@Rmano I was wondering i f@DavidCarlisle had an actual issue logged: 'Hacker News' not something I monitor ;)

@JosephWright Yes, I supposed that @DavidCarlisle comment what a kind of tongue-in-cheeck that the original blogger should open an issue...

Speaking of Hacker News: lua.org/manual/5.4/readme.html#changes

@JosephWright Are any of these changes significant?

@FaheemMitha Don't think so for us

@JosephWright cool!

@JosephWright Ok.

ooh there's a monster

@Thor hi mr monster

@PauloCereda aaaahh, a monster! /frightened

judging from my two seconds experiment, giant cursors are really impractical!

@PauloCereda Hello there! I'm calling to inquire about abuse of the adjustbox class!

@Skillmonlikestopanswers.xyz <3

@Thor ooh a monster question

@JosephWright There are to-be-closed variables which can be quite nice for working with nodes. But I don't think that there are any plans to update the version in LuaTeX any time soon.

@Rmano I noticed that if you use \draw (0,0) to[D] ++(2,0) to[short,-*] ++(0,-2); in circuitikz the two paths wouldn't be conjoined and you get an ugly corner.

@Skillmonlikestopanswers.xyz I don't see it... ^^^^

@Rmano currently creating an MWE.

\documentclass{standalone}

\usepackage{circuitikz}

\begin{document}
\begin{circuitikz}
  \def\dist{2}
  \draw
    (0,0) coordinate (O)
    node[rground]{}
      to[open,v<=$u_{e}$,o-o]
    ++(0,.75*\dist)
      to[R=$R$]
    ++(\dist,0) coordinate(m)
      to[short,*-,i<=$I_{n}$]
    ++(.25*\dist,0)
    node[op amp, anchor=-](opv){}
    (opv.+)
      to[short]
    (opv.+-|m)
      to[short]
    (m|-O)
    node[rground]{}
    (opv.out)
      to[short]
    ++(.25*\dist,0) coordinate(o)
      to[short,-o]

@Rmano not really minimal but^^^^

@Skillmonlikestopanswers.xyz will look

@JosephWright no just in reply to @Ulrike's comment here chat.stackexchange.com/transcript/message/54804436#54804436

@Thor yes?

@DavidCarlisle So 'ignore' :)

@Skillmonlikestopanswers.xyz it's probably a bug --- sometime happens to the point that there is an official hack (manual 5.4 Line joins between Path Components). I'll put here a minimal example with the hack...

\documentclass[border=10pt]{standalone}
\usepackage[RPvoltages]{circuitikz}
\begin{document}
\begin{tikzpicture}
  \def\dist{2}
    \coordinate(o) at (2,0);
    \draw (0,0) --
    ++(0,.75*\dist)coordinate(tmp)
      to[D,i=$i_{D}$,v=$u_{D}$]
    (tmp-|o)
       % to
       to [short, .-*] % hack to fix the short here...
       % to[short,-*]
    (o)
    ;
\end{tikzpicture}
\end{document}

...sorry ^^^^ will see if I find where the problem is (but it's here from the start of time, eons ago).

@barbarabeeton A robotic sloth: heise.de/news/…

@Thor hm in which phase does the div/0 happen? Python or TeX processing?

@JosephWright setting \l_color_fixed_model_tl has side effects on the \color_show:n command. (the colors themselves are fine):

\documentclass{article}
\usepackage{l3color,xcolor}

\begin{document}
\ExplSyntaxOn
\color_set:nnn {mycr} {cmyk/rgb}{1,0,0,0/1,0,0}
\definecolor   {mycr} {cmyk/rgb}{1,0,0,0/1,0,0}

\color_set:nnn {myrc} {rgb/cmyk}{1,0,0/1,0,0,0}
\definecolor   {myrc} {rgb/cmyk}{1,0,0/1,0,0,0}


\color_set:nn  {mixA} {myrc!50!mycr}
\colorlet      {mixA} {myrc!50!mycr}

xcolor:~{\color{mixA}mixA}\par
l3color:~{\color_select:n{mixA}mixA}\par
\color_show:n {mixA}

\tl_set:Nn \l_color_fixed_model_tl {cmyk}
\color_set:nn  {mixB} {myrc!50!mycr}

@Thor Can you make a minimal working example that would allow us to reproduce the problem?

@UlrikeFischer You remind me, it should be a string ...

@JosephWright but beside this I think your mixing is better than xcolor. It does less unneeded conversions between cmyk and rgb.

@UlrikeFischer :)

@samcarter_is_at_topanswers.xyz I can, but it would be reductible to an empty adjustbox environment with the option "center". :)

@UlrikeFischer What do you think the outcome should be of show here? I can see an argument for both possible cases!

@JosephWright it should at first show the right values, e.g. here the cmyk is missing a 0:

The color mixC has the properties:
>  model  =>  rgb
>  rgb  =>  1 0 0
>  cmyk  =>  0 1 1.

@UlrikeFischer Ah, yes: now that is strange

@UlrikeFischer Also an extra value for the cmyk case

:54805467 and in the cmyk case the cmyk value it show the cmyk for rgb: The color mixB has the properties:
>  model  =>  cmyk
>  cmyk  =>  0 1 1 0
>  rgb  =>  0 1 1 0.
<recently read> }

@Thor scaling tables is evil anyway but you could test if the output is there (is it always an external file) and skip if not. \IfFileExists{something.pgf}{do what you were doing}{skip this}

@JosephWright oh, and actually the color is then wrong, it only didn't show up in my example as I redefined it. If you switch back to rgb it is then blue.

@UlrikeFischer Yes, something wrong here

@Thor maybe I am not understanding, but ...

\documentclass{article}
\usepackage{adjustbox}
\begin{document}
\adjustbox{center}{}
\end{document}

gives no error here.

@DavidCarlisle That's actually a good point that I noticed and which surprised me; I didn't think adjustbox scaled by default without using the "scale" keyword, but it does do that, and it frustrates me -- ends up looking quite bad.

@Thor if you are not scaling do you need adjustbox? you could use center to centre it?

@JosephWright can a new user not edit their own question? tex.stackexchange.com/q/551671/3929, they wrote it in french and have now posted a new question in English presumably because they could not edit the original?

@Thor as always questions with no example are unclear and time consuming:-)

@Rmano Hmm, no, you're quite right. I added "min width", and that creates the division by zero error. What did I do to create the initial that made me create a whole bunch of actual, worse issue?!

@DavidCarlisle Case in point for sure :)

@Thor well there you go, that means if it is less than ... wide scale it to that minimum, so division by zero on empty

hmmm \adjustbox{center, min width=2cm}{} still works... @Thor

@DavidCarlisle The reason I used adjustbox was because it made it easier to center it and force the width as a factor of \textwidth for certain bad figures.

@Rmano The order matters; if min width comes first it gives division by zero, and if center comes first, it seems to work just fine.

@JosephWright I think \l_color_fixed_model_tl should not be considered when mixing, only when selecting a color: then it should either use an existing value, or convert (and perhaps store this for the next use) or error if the model is not usable.

@Thor centering is trivial \centering for example, and scaling tables is evil, and scaling tikz/pgf is usually best done within the tikz code...

@Thor Ah, ok. If you manage to build a minimal example in the main site somebody can propose workarounds or better solutions... trust the $TEXDEITYs here ;-). (I'm not one of those)

@Thor centre makes it full width, with the content centred, so presumably min-width does nothing in that case unless you specify a min width that is wider than the text block

@DavidCarlisle That seems reasonable. I'll swap them out and take a look.

@Thor @JosephWright is always happy when keyval arguments are order dependent,

@DavidCarlisle ooh positional keyvals

@PauloCereda it's not my fault if affine transformations aren't an Abelian group.

@DavidCarlisle damn those monoids

@daleif A user can edit their own questions, but in this case there are two separate accounts with the same name

@PauloCereda blame this bloke

@daleif Best link them to It looks like you've got two separate accounts, which means you cannot edit your original post or leave comments. The Stack Exchange staff can [merge them together for you](//tex.stackexchange.com/help/user-merge).

I replaced almost all the adjustbox environments with the centre environment. Oooof.

@UlrikeFischer my conclusion too

@UlrikeFischer Five minutes

@JosephWright I'm just installing ...

@JosephWright there is something not right with the currentcolor (.):

\documentclass{article}
\usepackage{l3color}
\usepackage{xcolor}
\begin{document}
\ExplSyntaxOn

\color_set:nnn {myc} {cmyk}{0,1,0.5,0}
\definecolor   {myc} {cmyk}{0,1,0.5,0}
\color_select:n {myc} abc \color_show:n{.}
\color_select:n {.} abc

\color_set:nnn {myrc} {rgb/cmyk}{1,0,0/0,1,0,0}
\color_select:n {myrc} abc \color_show:n{.}
\color_select:n {.} abc

\ExplSyntaxOff
\end{document}

I'm voting for forbidding the space character. (from filenames that is)

@yo' I'mGermansoI'mtotallywithyouonthat!

@yo' can we use zero width space instead?

@DavidCarlisle well, as long as it doesn't mess up with Overleaf's error parser, yes :-)

@DavidCarlisle I'm voting for active tabs expanding to blub.

@yo' my documents have no need for such a parser

@DavidCarlisle but you are you, and users are users :-)

@yo' well which is more important? (think before replying)

@DavidCarlisle users are more important, but you are importanter.

and both are inferior to pineapple duck pizza.

@yo' OK

@cis keyval is the original and best (naturally:-) l3keyval is the future a lot of people use pgfkeys or xkeyval

@cis expkv

@cis depends who you ask. I use now l3keys. @Skillmonlikestopanswers.xyz can give you a comprehensive list ...

@DavidCarlisle @Skillmonlikestopanswers.xyz @UlrikeFischer OK, I will have a look in the manuals. But a tutorial would perhaps be nicer.

@cis there is a section in the documentation of expkv that compares expkv to other available solutions (which I got to actually work).

@cis there is a older tugboat article from Joseph which compares various keyval systems. And here is a big list: tex.stackexchange.com/questions/26771/…

@Skillmonlikestopanswers.xyz @UlrikeFischer Ok, thx. I will read.

@cis But stay at least a mile away from some of the packages in that big list!

@Skillmonlikestopanswers.xyz I see it mentions neither simplekv nor options.

@UlrikeFischer you mean the big list? Should I add them?

@Skillmonlikestopanswers.xyz well I think they belong there.

@UlrikeFischer (already at it)

@JosephWright if an expression starts with a gray-model color like white!40!red, the mix is wrong: It writes three rgb values 1 0 0 g 1 0 0 G into the pdf. But I'm not sure what the right result should be. If we follow the rule that the first color defines the model it should try to convert the red to gray or error, but this probably against user expections.

\documentclass{article}
\usepackage{l3color}
\usepackage{xcolor}
\begin{document}
\ExplSyntaxOn
\color_set:nn {mywhitemixA}{white!40!red}
\color_show:n {mywhitemixA} %wrong
\color_set:nn {mywhitemixB}{red!60!white}
\color_show:n {mywhitemixB} %okay

{ \color_select:n {mywhitemixA} white mix }

{
  \tl_set:Nn \l_color_fixed_model_tl {gray}
  \color_select:n {mywhitemixA} white mixA
  \color_select:n {mywhitemixB} white mixB
}

\end{document}

That is the good thing about the manuals tikz, pgfplots, tcolorbox, .... that a (mostly) short example is given for everything. It is very user-friendly.

@cis expkv is living a niche life. It was released in January this year, so is (to my knowledge) the youngest key=val package, the selling point being a fully expandable parser (which happens to be the fastest key=val parser that supports active commas and equals signs).

@cis there are examples in the expkv documentation in section 1.4. Also you can use expkv-def to have a prettier interface to define new keys (and the documentation of expkv-def also contains examples). In addition to expkv-def there is also the expkv-cs package, which provides macros to create fully expandable key=val macros, and (probably this week) expkv-opt will be released to provide parsing of package and class options.

@cis if you have feedback for my documentations (which are suboptimal, I confess) you can always raise an issue on Github, or email me, or provide it here in chat, or come visit me for a coffee and talk to me in person, or...

@cis but note that there are things that will not be possible with expkv which you can do with pgfkeys, options, ... (but will most likely never need).

@cis also, you could interpret it that the package is so great and free of issues that no one needs to ask (which I doubt myself :) It is, however, bugfree per definition.

@Skillmonlikestopanswers.xyz I find pgfkey not so easy or often cumbersome.

@cis then you'll love the simplicity of expkv (or keyval, which unfortunately gets destroyed by xkeyval)

@cis and if your code uses expl3 and the key=val solution doesn't have to be the fastest (but still be reasonably fast), you can use l3keys. It is very robust and has a nice interface, imho.

@cis whatever you do, you can't do much wrong if you stay away from anything loading catoptions (which means most if not all key=val packages by Ahmed Musa) and xkeyval (because it has some very strange behaviour regarding braces, unbalanced \if statements, ...)

@cis well, this is somewhat covered by the "Standard Use-Case" example in the documentation of expkv.

\documentclass{article}

\usepackage{xcolor}
\usepackage{expkv}

\ekvdef{mycommand}{color}{\def\mycolor{\color{#1}}}
\ekvdef{mycommand}{font}{\def\myfont{#1}}
\ekvdef{mycommand}{exponent}{\def\myexponent{\textsuperscript{#1}}}
\newcommand*\mycolor{}
\newcommand*\myfont{}
\newcommand*\myexponent{}
\newcommand\mycommand[2][]
  {%
    \begingroup
    \ekvset{mycommand}{#1}%
    \mycolor
    \myfont
    #2\myexponent
    \endgroup
  }

\begin{document}
\mycommand{text}\par
\mycommand[font=\bfseries]{text}\par

@cis example^^^^

@DavidCarlisle They heard you! twitter.com/roisiproven/status/1277994784368533504?s=20

> Hah yeah that’s true, I’ve always been very scared of LaTeX it’s like they don’t think humans will read the docs

\documentclass{article}

\usepackage{xcolor}
\usepackage{expkv}
\usepackage{expkv-def}

\ekvdefinekeys{mycommand}
  {
    ,store font = \myfont
    ,data color = \mycolor
    ,data exponent = \myexponent
  }
\newcommand\mycommand[2][]
  {%
    \begingroup
    \ekvset{mycommand}{#1}%
    \mycolor\color{}%
    \myfont
    #2\myexponent\textsuperscript{}%
    \endgroup
  }

\begin{document}
\mycommand{text}\par
\mycommand[font=\bfseries]{text}\par
\mycommand[color=red,font=\itshape]{text}\par
\mycommand[exponent=2]{text}\par

@cis similar example, but using expkv-def^^^^

note to myself: Not only should I not google latex, I shouldn't search latex on twitter, too.

@Skillmonlikestopanswers.xyz Ahhhh, good.

@Skillmonlikestopanswers.xyz cis does NOT tweet!

@UlrikeFischer That's by-design and follows xcolor: if the model is gray, flip to the 'next' model so you don't make everything black-and-white (except mixing gray with gray)

@cis I don't as well, just followed the link posted by @yo'

@JosephWright I don't mind the choice of the model, but the values are wrong:

The color mywhitemixA has the properties:
>  model  =>  rgb
>  rgb  =>  1 0 0
>  gray  =>  1 0 0.

The gray has three values, and this is also written to the pdf: 1 0 0 g (not 1 0 0 rg!)

@cis update your TeXLive, data was added a while ago.

Off for a while, will be back later today.

@Skillmonlikestopanswers.xyz Ouuh, once more. OK, I search for the comand tlngr --updtae all or sth.

\newcommand{\ShowStartpointsMap}[2][\MyHuge]{%%%%%%%%
\pgfmathtruncatemacro\Dec{#2}%
\foreach \n in {0,...,\HiraganaDataLastRowNo}{%%%
\pgfplotstablegetelem{\n}{Dec}\of\StartnumbersTable%
\pgfmathsetmacro\dec{\pgfplotsretval}%
\ifnum\Dec=\dec% =====================
\pgfplotstablegetelem{\n}{Strokes}\of\StartnumbersTable%
\pgfmathsetmacro\Strokes{\pgfplotsretval}%
\pgfplotstablegetelem{\n}{soStartpointsStartnumber}\of\StartnumbersTable%
\pgfmathsetmacro\Startnumber{\pgfplotsretval}%
\pgfmathsetmacro\StartnumberStartpoints{\Strokes==1 ? int(\Startnumber+1) : int(\Startnumber+2)}%

@UlrikeFischer ooh blub

@UlrikeFischer Ah, right, oop

@JosephWright ooh oop

@JosephWright object oriented programming?

@Skillmonlikestopanswers.xyz ooh oatmeal pancakes

@cis first we should find a consensus on what the package should support and what it shouldn't, also we should specify some interfaces on how things should be accessed or specified or customized. Sorry, but I feel kind of overwhelmed with the amount of information I get from you in a (to me) unstructured manner. I don't have much free time right now, so you have to bear with me being slow currently (it takes me more than half a week to write expkv-opt now, instead of a single day...)

Again off until later this evening :)

ooh stealth rabbit movements

@cis you may as well use the pgf key system since you are using it already as you have things like \node[text=red] you could use a different package but...

@DavidCarlisle No bad idea.

@UlrikeFischer I'll fix those two issues either later today or tomorrow

@JosephWright fine with me. What should I do if I see typos in the documentation part?

@Skillmonlikestopanswers.xyz github.com/circuitikz/circuitikz/issues/417 I report issues and then discuss them with myself. Is it worrying?

@Rmano my condolence. Unfortunately, circuitikz seems way too big for me to really grasp it (and my TikZ foo isn't well as well), so I'm no help here.

@arthur-reutenauer Please see tex.meta.stackexchange.com/questions/8606/…

Does anyone have a recommendation for a Mac programming IDE with good support for javaScript? I usually use BBEdit, but it's not great for jS.

Vim users can stay quiet. :)

@AlanMunn emacs

@Just_A_Man he wont see that ping unless you have mod status on another site so have and special powers

@DavidCarlisle :) Is there a js version of AucTeX?

@AlanMunn who needs js when you have lisp but the answer can be found here lists.gnu.org/archive/html/auctex-devel/2017-04/msg00016.html

@DavidCarlisle Very helpful, but notice my question has existential, not modal force. :)

@AlanMunn that quote was from may 1998 it turns out, Ive been using emacs too long... but I suppose this is the official answer emacswiki.org/emacs/JavaScriptMode

@DavidCarlisle I don't have that many other options except writing him an e-mail. Anyway, if he pops into chat by chance, he will see the message, right?

@Just_A_Man no

@Just_A_Man you could leave a comment under one of his questions eg tex.stackexchange.com/a/325417/1090 is most recent

@DavidCarlisle For some version of 'recent' :)

@AlanMunn oops I think I picked the wrong one, still it would work there as well

@DavidCarlisle I left a comment under my own post in meta; would he see it?

@Just_A_Man no. Basically if the name doesn't pop up if you type @A then the ping will not work.

@Just_A_Man No, this is by design: the site is designed to be 'opt-in', so that people don't get bugged for answers etc.

@DavidCarlisle Ok, I'll remove it then. As for leaving a comment under an unrelated post of Arthur, this seems an inappropriate thing to do, so, I won't do it.

@AlanMunn I see, thx.

@Just_A_Man yes I thought he may have one related to xetex maintenance but I couldn't be bothered to check all the list

@DavidCarlisle For a reason unclear to me, tex.stackexchange.com/users/170/arthur-reutenauer?tab=tags doesn't contain the tag [xetex]. That's strange. After all, the user is a maintainer for xetex. Is the other maintainer on tex.SE?

@DavidCarlisle -- I've got a comment and a counter-example on this question/answer, and would appreciate it if you'd add them. The comment is that at least some document classes define handling of the bibliography so it uses \frenchspacing. The counter-example consists of two names with multi-letter middle names that shouldn't be reduced to a single initial: Nicholas {DeB}. Katzenbach, and Neil {DeG}rasse Tyson.

@Just_A_Man Being a package maintainer doesn't entail being active on the site. And xetex is a very general tag, so it's unlikely that a maintainer would bother to follow it (as opposed to e.g. a tag for a specific package).

@barbarabeeton I can add the comment if you put verbatim what you want to add here. (Still unable to comment?)

@AlanMunn I see. If even the maintainer doesn't follow the tag of their own engine, I don't know what to think. Anyway, a xetex-related post of him would be tex.stackexchange.com/questions/246881/… , but I still don't think it would be appropriate for me to ping him there. What do you think?

@AlanMunn -- Yes, still unable to comment, read comments, vote, or anything else useful. Desired text follows.

@Just_A_Man As I said, the xetex tag is far too general to be worth following. And even if it weren't nobody is obliged to participate in the site.

@barbarabeeton Ok I added it for you.

w.r.t. your posting problems, have you tried logging out of the site and logging back in? I assume you've also tried different browsers?

@AlanMunn -- Yes, have tried just about everything available on this laptop, including different browsers. Only Chrome will even bring up this site; Safari and Firefox tell me they can't/won't connect. (I hope the new laptop arrives soon ...) Thanks for the posting -- looking good.

@Just_A_Man why do you find that strange I don't have tag badges for most of my packages, or for latex, it's all very random:-)

@barbarabeeton Alan's done it I think:-)

@DavidCarlisle -- Yes, he has. Thanks for checking.