Package xcolor Warning: Incompatible color definition on input line 4050.


Package xcolor Warning: Incompatible color definition on input line 4051.


Package xcolor Warning: Incompatible color definition on input line 4052.


Package xcolor Warning: Incompatible color definition on input line 4053.


Package xcolor Warning: Incompatible color definition on input line 4057.

@UlrikeFischer ^

@DavidCarlisle where?

@UlrikeFischer pdflatex listings.dtx

@DavidCarlisle I can look later, but it probably loads color first and defines colors in color syntax.

@UlrikeFischer oh listings has a current date so is still maintained, good:-)

@DavidCarlisle how current?

@UlrikeFischer Package: listings 2023/02/27 1.9 (Carsten Heinz) so this year

@DavidCarlisle xcolor could perhaps try to convert such colors or color could define also a l3 color...

@UlrikeFischer yes that's what I'm thinking, we could no doubt update listings but it might be an indication of where we could improve compatibility somewhere. what's changed though?

I guess this is bad:

{restoring \\color@.=macro:->\xcolor@ {}{0 g 0 G}{gray}{0}}
{restoring \current@color=macro:->0 g 0 G}

@DavidCarlisle why do think that something changed?

@UlrikeFischer well I wouldn't have published a package with that many warnings, so I assumed that they didn't occur when listings.dtx was made, but perhaps that's just me being optimistic:-)

@DavidCarlisle ah it could be the new code that updates also the dot color in normalcolor

@UlrikeFischer yes \\color. and \current@color do show up a lot in the trace

@DavidCarlisle I don't dare to check how many warnings we overlook in the various docu build with l3build.

@UlrikeFischer there was a time when I ensured that you could build source2e with no overfull box or other warnings...

@JosephWright Thanks for including my suggestion into l3draw!

@DavidCarlisle it isn't new. The dtx loads the lstdoc.sty which loads color and defines the color darkgreen and then it loads xcolor. And so every use of darkgreen (in the code snippets) gives an incompatible color message. The easiest work around is to load xcolor earlier in the dtx.

\documentclass{article}
\usepackage{color}
\definecolor{darkgreen}{rgb}{0,0.5,0}
\def\rstyle{\color{darkgreen}}
\usepackage{xcolor}
\begin{document}
{\rstyle xxx}
\end{document}

@UlrikeFischer ah I didn't think the bits I were seeing had changed that much, but I thought it easier to ping you rather than trace further:-) it's so tempting to make color load xcolor but I'm not sure if that would help or harm eventual move to l3color

@UlrikeFischer couldn't we make xcolor not do this:

> \XC@current@color=macro:
->0 0.5 0 rg 0 0.5 0 RG.

@DavidCarlisle I think it would make it it easier if there were only one (legacy) color format, I only don't know if there are packages around which will fall over if a color has suddenly the format \xcolor@ {}{0 0.5 0 rg 0 0.5 0 RG}{rgb}{0,0.5,0}.

@DavidCarlisle ? In which context?

@UlrikeFischer any that are used, someone would surely have complained by now if they were incompatible with xcolor

@DavidCarlisle only if xcolor is loaded before. And in classes that doesn't happen easily. But I think one could try. What is with plain? That would need to stay with the old syntax wouldn't it?

@UlrikeFischer in your example above it's basically setting \XC@current@colro to \current@color if it hasn't got a real xcolor definition but it could recognise at least pdftex formats and define the \xcolor@... version... perhaps (but making color.sty do that as well, or as I say just making color be xcolor might be simpler

@UlrikeFischer if we make any changes in that direction for graphics or color I think we need a rollback color-2023-12-09.sty then make graphics-pln use those frozen versions and not pull in more and more expl3 based code

@DavidCarlisle ah yes. It could split up the values and redefine the color in most cases, it would be engine dependant, but there aren't so many formats to handle.

@DavidCarlisle but I do think that loading xcolor would be easier. Three color formats are problematic anyway.

@UlrikeFischer yep

@DavidCarlisle and by the way the keyval change seems to have gone quite smoothly ...

@UlrikeFischer hyperref next..

@DavidCarlisle I looked yesterday at it. As a first step I need to change the key module names in the pdfmanagement driver to get in line with main hyperref.

@UlrikeFischer your example works if pdftex.def looks like

\providecommand\xcolor@[4]{#2}
\def\c@lor@@rgb#1,#2,#3\@@#4{%
  \c@lor@arg{#1}%
  \c@lor@arg{#2}%
  \c@lor@arg{#3}%
%  \edef#4{#1 #2 #3 rg #1 #2 #3 RG}%
   \edef#4{\noexpand\xcolor@{}{#1 #2 #3 rg #1 #2 #3 RG}{rgb}{#1,#2,#3}}%
  }
  % \@namedef {\color@xdarkgreen}{\xcolor@ {}{0 0.5 0 rg 0 0.5 0 RG}{rgb}{0,0.5,0}}

.. not sure if that's wise but if we want a half way house short of really making color load xcolor

@DavidCarlisle I'm not sure either, really loading xcolor looks saner. But before we should remove the global colors, and one need to check the package options. E.g. monochrome is currently load-only, so it would get lost in a document loading color first.

@JosephWright I don't really understand it. Yesterday disabling l3bitset worked, but now it doesn't. \@expl@finalise@setup@@ is set to empty in ltfinal and so it has not effect at all.

@UlrikeFischer yes more of the front end changes if we really load xcolor, the above is a minimal change to the backend so it uses xcolor saved format without bringing in the extra color mixing and options things. No rush but a possibiliity....

@DavidCarlisle yes, I think it would be a good first step and I don't see why it should harm. Unless someone does some low-level extraction of the values ... Would that work with plain?

@UlrikeFischer yes it should, I think.

@DavidCarlisle hyperref (hycolor) seems to work fine too. So ...

@UlrikeFischer I could make a PR to graphics-def

Just reading the new Komödie: The date for the Dante meeting is given as 4.4. - 6.4. 2024

@samcarter Wir Deutschsprachige müssen uns irgendwo treffen,

@DavidCarlisle Natürlich! Und das an geschichtsträchtiger Stelle de.wikipedia.org/wiki/Wielandgut_(O%C3%9Fmannstedt)

@UlrikeFischer breaks xcolor though :( (at least in the simple form above)

@DavidCarlisle how? I tried a simple example and that worked.

@UlrikeFischer yes so did I I need to backtrack but now I'm getting \xcolor@ {}{\xcolor@ {}{0 0.5 0 rg 0 0.5 0 RG}{rgb}{0,0.5,0}}{rgb}{0,0.5,0} as xcolor is using the function to generate the system-specific second brace group and this has changed it to make it make all four groups

@UlrikeFischer ie when xcolor.dtx says % If the model is supported and |\rangeRGB| equals 255, we stick to the algorithm provided by the driver. it expects \color@RGB to make 0 1 1 rg 0 1 1 RG` not an \xcolor@ call.

@DavidCarlisle actually I get that too, but the color worked nevertheless as it expands everything simply twice.

@UlrikeFischer yes but I think that's perhaps too scary to rely on

@UlrikeFischer I may be able to inject an ifsomething test to lose the inner call (the alternative of course would be to modify color.sty to add the xcolor@ wrapper but I'll see...

@DavidCarlisle is it only rgb or also other models?

@UlrikeFischer all

@UlrikeFischer I stupidly started by writing an lvt test file

@UlrikeFischer \edef#4{\ifx\colorlet\@undefined\expandafter\noexpand\fi\xcolor@{}{#1 #2 #3 rg #1 #2 #3 RG}{rgb}{#1,#2,#3}}% ???

@DavidCarlisle we could adapt xcolor first to expand the argument?

@UlrikeFischer I think having it contain the four fields fully expanded is logically the right thing so I'd rather not

@DavidCarlisle ? I meant can't xcolor not do that (so create four full expanded fields)?

@UlrikeFischer oh in its define, yes it could

@UlrikeFischer I hate tex

@DavidCarlisle ?

macros with these \meaning not \ifx equal

FAIL: rgb
macro:->\xcolor@ {}{0 0.5 0 rg 0 0.5 0 RG}{rgb}{0,0.5,0}
macro:->\xcolor@ {}{0 0.5 0 rg 0 0.5 0 RG}{rgb}{0,0.5,0}

@DavidCarlisle quite unfair

@DavidCarlisle oh no

@UlrikeFischer that expl3 stuff may yet turn out to be useful....

>  r (the letter r)
>  g (the letter g)
>  b (the letter b)

>  r (the character r)
>  g (the character g)
>  b (the character b)

\tl_analysis_show:c

@DavidCarlisle who stringifies? xcolor?

@UlrikeFischer yes I just changed my pdftex.def to match (it probably never matters but makes testing easier if the tokens match)

@DavidCarlisle do we need to change something in xcolor?

@UlrikeFischer no I don't think so:

PASS: rgb

PASS: RGB

PASS: cmyk

PASS: gray

END-TEST-LOG

I'll check something in to a branch in graphics-def in a bit

@UlrikeFischer * [new branch] XC-color -> XC-color

@UlrikeFischer just pdftex.def can do the others if we decide it's a good idea

@DavidCarlisle I still get warning for named colors:

\documentclass{report}

\usepackage[dvipsnames,usenames]{color}
\expandafter\show\csname \string\color@Orange\endcsname
\usepackage{xcolor}
\begin{document}
cccc\color{Orange}xxx
\end{document}

@UlrikeFischer ah I'll check thanks

@UlrikeFischer oh that's just I stopped too early (the named test in the lvt) is commented out:-) fix coming up..

@UlrikeFischer grr gets edef'ed twice so the \xcolor@ vanishes. clear where it goes but not so clear if I can reasonably fix in the .def file, might need to change color.sty

@UlrikeFischer ooh `\protected

@UlrikeFischer fixed although the test fails as numeric differences 0.5 v 0.50

@UlrikeFischer I just added \protected which wouldn't work with classic plain tex but probably I don't care etex and pdftex would be OK

Hello hello, googled around and did not find an immediate answer so came here to ask a short sanity-check

Is there a way to define a new math style in xelatex, using a system OT font?

So define a command like \mathcal2{#1} which would draw #1 in the selected font

@cfr On the baseline issue. I'm inclined to remove one of them because I agree it's not really more logical to have two, but at the same point having code attached to a terminal node that is logically applying to the root is why I'm uncomfortable with just having one. Is there any way to get the same effect with code on the root?

@nathdwek Perhaps this is what you have in mind? tex.stackexchange.com/q/131866

Thanks, had not found that question yet

I was going to do what the asker suggests and hijack an unused alphabet

@DavidCarlisle ;-) looks ok now.

Now bit worried about the ascii range thing, since I have to typeset a rho

@nathdwek well while I would trust David's answers: the question and the answer is nearly 10 years old and could be a bit outdated. Do you use unicode-math?

Not yet

Honestly, I'm also not finding a font that satisfies me, so I might give up on the idea

@nathdwek but the question and answer is using it. Without unicode-math things are very different.

For context, I'm running ways of denoting different "versions" of the same variable, (transformed from/to different domains), and I'm using underbar for the last domain, but not a fan of it. So thought maybe there could be a nice font out there to distinguish it efficiently from other domains

but not a fan of what I am seeing in 2 short minutes

@UlrikeFischer hence "not yet" and not "not"

I'm on xelatex with mathspec for the moment, but from what I read it's not like I would completely need to change my flow and document if I were to add unicode-math to the soup

@nathdwek well I'm not a fan of mathspec tex.stackexchange.com/a/487319/2388

mmmh, interesting

@UlrikeFischer I'll take it into account next time I start a document :)

Or maybe tonight already, but for the moment just trying to polish some slides :))

@nathdwek Is it something specific you do not like with the fonts you looked at so far?

Yeah just that they're not distinguishable enough from other versions

the challenging bit is \rho

Looks too similar in the reasonable fonts I have found

I have it as regular, bold, capitalized and bold capitalized already

Maybe a \mathbb equivalent for \rho would be nice, but I don't think it would be very readable especially on slides

@UlrikeFischer I added luatex, xetex, and dvips drivers, xcolor doesn't really support named model just gets turned in to cmyk, so currently the test fails for dvips and xetex, not sure whether should just make color do the same and give up on named.

if I can even find a font that can typeset \rho in \mathbb taht is

@nathdwek yes but not with that name, you can't have 2 in a name

Sorry @DavidCarlisle I don't know what you are referring to exactly

@nathdwek I don't recall seeing greek blackboard bold.

Yeah...

@mickep just to be clear on my idea: I am not finding any satisfying solution with "classic" families \mathbb, \mathfrak, \mathcal, etc

So my idea was just to load a custom system installed TT font to write my "special" \rho with

@nathdwek Below are a few different \rho from stix two.

@nathdwek you asked for \mathcal2 which you can't have, you can have \mathcalb or \mathscr or \mathcalnew ....

But just playing around with libreoffice write, I am not really finding such a font anyway

@DavidCarlisle ah, clear now what you meant by "2", thanks a lot for taking the time for considering the question

@nathdwek they are not fonts but generic font names that may refer to anything, just as in text \bfseries isn't a font ot's just the bold version of whatever font you are using

absolutely

Is it discouraged to nest l3draw environments? For example, could I place a draw environment in a coffin that is placed inside a draw environment? I noticed that the bounding box will be affected in an unexpected way, but if this is not meant to work anyways, it might not be a bug really ...

That is why I am trying above to clarify between families (should have said alphabet from looking at your answer in the other question), and regular old font

@mickep I appreciate the second one, but I think if I try to use those to visually distinguish from \bm{\rho}, I'm just going to lose the audience

Ah, I found the answer: \draw_suspend_begin:...

@nathdwek Maybe the real problem is that you do have to many \rho.

@DavidCarlisle yeah I think we agreed on that from the start, probably I had not formulated it correctly initially

@mickep always fun to run out of symbols and subscripts and superscripts and stars and daggers and tildes in the middle of a reasoning

@nathdwek er.... obfuscate or clarify:-) I am not finding any satisfying solution with "classic" families \mathbb, \mathfrak, \mathcal, et they are just names so whatever font you eventually find you could assign it to one of those names, and it isn't clear how many you have tried so far unicode-math would allow 10 or so different unicode math fonts for each of them, if you also look to classic tex math fonts ctan has dozens and dozens of font packages

@DavidCarlisle ah yeah very clear now how I used "classic" families to mean the LM Roman version of them

@nathdwek \mathcolor{red}{\rho} ?

That's on me :)

@nathdwek If you had to pay me 1€ for each new symbol you used, I'm sure you would use a bit fewer. Maybe it could even help the readers. :P

@mickep at least my field does not mix \nu and v

:)

@mickep hey if I only charged 99 cents would I get the money instead?

@DavidCarlisle fun, but I think the referee would call it "Tacky"

Then Elsevier would probably charge us 5000 euros in editing and publishing fees and change it to unreadable green and put it out anyway

@nathdwek try texdoc colortbl for a masterpiece in subtle use of color.

@DavidCarlisle You are too cheap...

I'll keep it in mind for next time though :)

maybe I could use LM Roman mathcal, and draw the curliest of all curly rhos myself and just learn how to insert that in my equations :)

For my next procrastination project

Thanks a lot for the advice, I learned a bunch

Also broke my slides using unicode-math :/

@mickep Actually to your point about helping the readers

That is the tightrope to walk isn't it

I could just use a different symbol altogether to mean "rho but in the stacked domain". But if I find a distinguishable version of rho, that still looks like a rho, but with "stacked" vibes, that's clearer

emphasis on "if" I find it

@DavidCarlisle I appreciate the self-deprecating humor, and all your more serious contributions to tex and friends :)

@UlrikeFischer I know it's a slightly ill-posed question without the full context of the document, but should I just load unicode-math in place of mathspec, or should I load fontspec as well?

Getting the following error:

! Extended mathchar used as mathchar (14688526).

unicode-math will load fontspec for you but it doesn't harm to use both, don't use mathspec at the same time.

@nathdwek ignore the colours which are just to help separate columns but the italic math rho easily available from unicode math are

Thanks @DavidCarlisle, underbar it is

@nathdwek that means you (or something) used a Unicode \Umathchar value in a classic tex \mathchar assignment

Any advice on how to track that down? Gotta admit I am bit in over my head

pastebin.com/Pv40WwWg

@JasperHabicht It's currently a more-or-less direct rip-off of pgf, which doesn't nest - but I would need an example to work on it in more detail

@JosephWright No, the example is in fact in the manual already where \draw_suspend_begin: is explained. This is exactly my use case (just with a coffin instead of a box).

@JosephWright It works well when using \draw_suspend_begin:. So, I would think it is quite robust here.

@nathdwek If I comment out %\setsansfont{Source Sans 3 Light} and put \stop at the end of that pastebin, it runs without error in texlive 2023

Thanks for trying

I have also tried to comment out that line, but does not help

just a second

@nathdwek you get an error? what tex system do you have? is it old?

Yeah, I get an error in the body, hold on I might have been stupid when following line numbers

@nathdwek there is no body there the (far from minimal) example stops before \begin{document}

@JasperHabicht Er, oh that: should work as documented, that's rather the idea :)

@DavidCarlisle I just pasted the preamble, did not expect you or anyone to run it

@JosephWright Yup. I just did not read that far first =D

just maybe to spot conflicting packages

And also can't share everything in the body

But my understanding of the log is that the error is in the following frame

pastebin.com/QEzgqahR

@nathdwek you are kidding? It's far easier to run (a sensible) example rather than try to debug 60 lines of weird code on sight by eye. That's why we ask for proper questions (in the Q&A site) with reasonable examples

@DavidCarlisle Apologies for the mismatch in expectations, if I had written a question I promise you it would have been a proper MWE and more debugging on my side before even asking

@nathdwek no sorry I'm not going to try to guess how to reconstitute disconnected fragments, you can easily make a test document and easily remove every package from the test that is not required to show the issue.

Here in chat, I got a package suggestion and just wanted to try it out, ran into some issues and was just wondering if there was some clash I was not aware

@nathdwek fair enough, but it's 10 in the evening here and debugging by eye is no fun so another day perhaps...

Understand, well now that we're here, I will prepare an MWE because I am curious. But please don't take my earlier question as disrespect, it's just a casual chat that devolved because of curiosity. Someone just suggested to try out a package and I got into it.

@nathdwek I've been chatting on the internet since before UK joined the internet, you'd have to try much harder if you wanted me to take offence. Also you wouldn't be the only one:

:)

Well, seems like the answer is package bm

I imagine unicode-math actually allows to have bold greek in an easier way?

So my lesson is to drop package bm and use \symbf

Do I have that right?

@nathdwek I was just going to say if I join your examples it says bm isn't defined, and if I add bm package I get ! Improper alphabetic constant. with lualatex so you are using xelatex which says ! Extended mathchar used as mathchar (14688526). which I guess is my fault but you really don't need \bm with a unicode math font that has bold italic greek already in the same font, use \symbf{\rho}

yup yup

@nathdwek oh yes you said that already. Exactly,

Thanks for pointing out that I do in fact own a shovel myself

Just had to remember how to dig

and a wonderful evening (?) UK (?) to you

incorrect parentheses but you get the spirit

@nathdwek it's easy to see why bm fails (if you wrote it) but not so easy to know how to fix it, sometimes just making bm detect a uniocde tex and turn itself into \symbf or \symbfit or \symsomething would be the right thing but in othercase it isn't, an automatic drop in replacement that works for all font choices is a bit tricky to iwhat to do, so doing nothing and letting it fail is honest of not that helpful

Thanks for the explanation

If it was not clear the shovel bit was about pushing me to just look into the issue myself

In case the "not that helpful" was directed at that

@nathdwek note \bm{\rho} would be bold italic so perhaps \symbfit better than \symbf in this case like \bm{r} compared to \mathbf{r}

oh yeah absolutely thanks for pointing that out

Not sure if I would have noticed! and yes indeed I need bfit

@nathdwek no I meant leaving bm to drop low level tex errors with unicode math fonts isn't that helpful, I could at least detect it and give a higher level "don't use bm with unicode-math until I get chance to talk to Will" message

mmmh

got it

I appreciate that you're acknowledging this

The original original reason I'm here though is because of the wonderful community and packages, yours included. Don't be too hard on yourself :)

Alright I do need to get these slides dones

@AlanMunn Sorry. Had to eat. I see you switched to for leaves inside for tree, but that means you're executing baseline for every leaf node for every node in the tree. Admittedly, there are only 3 nodes and 2 are already leaves, but that's still 4 baselines. Why not just

    pcorr down/.style={for tree={pcorr,grow=south,anchor=south},for next leaf= {baseline}}

?