@DavidCarlisle You're right... it's the bad (or good) habit of trying not to depend on the OS, anyway it's a great project, I hope to see it in action soon :)

@DavidCarlisle Fine with me: we are not tied to anything

@JosephWright (sorry for always bothering you with my L3 related questions): How could I write a test in l3build which needs multiple files (because I have to somehow write tests for the class/package option package)?

@Skillmonlikestopanswers.xyz Put them in a support folder inside testfiles: see for example l3kernel and tests for file input

@JosephWright thanks, I'll take a look :)

I get an overfull hbox warning from the following code (and I'm totally lost on why):

\input{regression-test}

\RequirePackage{expkv-opt}

\START
\OMIT
\makeatletter
\newcommand\setoptions[2][\relax]
  {%
    \unless\ifx\relax#1%
      \def\@currext{#1}%
    \fi
    \expandafter\def\csname opt@\@currname.\@currext\endcsname{#2}%
  }%
\ekvdef{known set}{keyA}{}%
\ekvdefNoVal{known set}{keyB}{}%
\setoptions[pkg]{keyA,keyB=value,keyC={value, with a comma}}%
\TIMO

Forget about it, found my stupid mistake :)

Morning @DavidCarlisle

Interesting discussion about TeX and website (mis)design: twitter.com/patrickgundlach/status/1278584101289811968

@JosephWright will look...

@JosephWright well one can certainly discuss some design decision, but I don't think that he is right when claiming, that you can't find the central information. On all sites the "we are a group supporting TeX" and "what is TeX" is prominent and easy to find.

@UlrikeFischer also there is the issue that new users who need to know what tex is don't start (or join or use) the user groups, they come here....

@JosephWright you saw the error I mentioned yesterday?

@UlrikeFischer No

@JosephWright there is a variant missing: \cs_generate_variant:Nn\prop_put:Nnn {cxx}

@UlrikeFischer REally?

@UlrikeFischer github.com/latex3/latex3/commit/…

@JosephWright oh, then I need to recreate the formats to get this.

@UlrikeFischer Yes: for testing, I'm just adding the appropiate variants at the start of my test file

@JosephWright today texlive update has taken care of this.

@JosephWright what's happened with www.tex.ac.uk and redirects to texfaq for the old cgi-bin/texfaq2html url, I thought they were working before?

@DavidCarlisle It was: @StefanKottwitz any ideas?

@DavidCarlisle Perhaps something to do with having to move the DNS? But that was working ...

$ORIGIN tex.ac.uk.
*                       300 MX    10     mail.texfaq.net.
ftp                     300 CNAME        texfaq.net.
www                     300 CNAME        texfaq.net.

@DavidCarlisle ^^^ DNS record at Mythic Beasts

@JosephWright texfaq.net? (not .org?)

@DavidCarlisle Ah, that would be the issue :)

@JosephWright stick to .xyz :-)

@DavidCarlisle :)

@DavidCarlisle DNS fixed, give it ... a little while

@DavidCarlisle Hmm, still not right

@UlrikeFischer The selecting shoudl work, it's the set up I've not done, but at present I've been focussed on trying to get export/multi-models right, as DeviceN/Separations really needs to build on it

@JosephWright not valid for www.tex.ac.uk. The certificate is only valid for the following names: texfragen.de, www.texfragen.de blame @UlrikeFischer she's German:-)

@DavidCarlisle I get a 404 from GitHub Pages :(

@JosephWright oh I don't get that far, perhaps dns takes a while (actually when I set up latexcgi dns I had to quit my work vpn before I saw it that day) but /cgi-bin/texfaq2html?whatever are supposed to work at texfaq.org

@JosephWright it works with spot and one value: \color_set:nnn {myspot1} {spot} {0.5}, but not with another name and with two values.

@UlrikeFischer Ah, probably I need to do more testing :)

@UlrikeFischer I see that my latest checkin failed as we are at the deprecations-to-remove point, so I'll need to do a bit on master before I can get anything to pass in the branch(es)

  ,boolTF boolA = \myboolATF
  ,boolTF boolB = \myboolBTF



    \myboolATF{S1 S2 S3}{S0}%
    \myboolBTF{\iftrue{boolA}{AAA}{BBB}}{YYY}%

there is place for improvement here ;-)

better ;-)

@DavidCarlisle the cpanm tool (cpanminus) is my tool of choice nowadays for perl as it installs to the current user or your dir of choice by default. It also have a nice way of listing dependencies such that when you install those then if it is not in the base installation it will install it for you (compiles it from scratch, so build tools are needed, bit scary if you want to install your own database drivers)

@DavidCarlisle IIRC, there are ISPs/regions in which .xyz are blocked (topanswers had such an issue, I think)

@cis the boolean value is only stored in the \myboolATF macro by letting it to \@firstoftwo or \@secondoftwo, so you'd use \myboolBTF{\myboolATF{AAA}{BBB}}{YYY}.

@Skillmonlikestopanswers.xyz Works!

\documentclass{article}
%\usepackage{xcolor}
\usepackage{expkv}
\usepackage{expkv-def}

\ekvdefinekeys{cmd}{
  ,boolTF boolA = \boolATF
  ,boolTF boolB = \boolBTF
}
\newcommand\cmd[1][]
  {%
    \begingroup
    \ekvset{cmd}{#1}%
    \boolATF{S1 S2 S3}{S0}%
    \boolBTF{\boolBTF{\boolATF{AAA}{BBB}}{YYY}}{}%
    \endgroup
  }

\begin{document}
\cmd[boolA=true, boolB=true]

\cmd[boolA=true, boolB=false]

\cmd[boolA=false, boolB=true]

\cmd[boolA=false, boolB=false]
\end{document}

@Skillmonlikestopanswers.xyz really? it works for me, so I don't care:-)

quack

@JosephWright hm, did I remember right that l3build doesn't copy support files with umlauts?

@UlrikeFischer It should work, but I've not tested all this

@JosephWright it doesn't work, I have two files in the support dir, and the one is then in the build dir and the other (grüße.txt) not.

@UlrikeFischer I'll have to look at that

@JosephWright but it is not pressing, I will use filecontents for now in this test.

@cis Your code doesn't make sense, if \boolBTF is false the first will use the second value (empty), if it is true it'll use its first value (\boolBTF{\boolATF{AAA}{BBB}}{YYY}), but the second \boolBTF will never input YYY, because to get it called it has to be true. So the code you posted is equivalent to \boolBTF{\boolATF{AAA}{BBB}}{}, which doesn't seem like what you want to do. Instead \boolBTF{\boolATF{AAA}{BBB}}{YYY} should be what you intend.

@JosephWright Is there a option for l3build to skip the format generation when testing LaTeX2e and use a preinstalled format?

@cis also, you can use an \ifnum test to get logical AND, logical OR, or logical XOR tests:

\documentclass{article}

\usepackage{expkv-def}

\ekvdefinekeys{cmd}
  {%
     boolTF boolA = \boolATF
    ,boolTF boolB = \boolBTF
  }

\newcommand*\cmd[1][]
  {%
    \begingroup
    \ekvset{cmd}{#1}%
    boolA: \boolATF{true}{false}\par
    boolB: \boolBTF{true}{false}\par
    Or-test:
    \ifnum\boolATF{1}{}\boolBTF{1}{}0>0
      true%
    \else
      false%
    \fi
    \par
    And-test:
    \ifnum\boolATF{\boolBTF{1}{}}{}0>0
      true%
    \else
      false%
    \fi
    \par
    Xor-test:

@MarcelKrüger --rerun?

@MarcelKrüger Not really, as the cases where we need a custom format tend to go badly wrong without it

@JosephWright OK, then I will have to teach l3build to build luametatex formats... Is there a list of all engine options which l3build might try to use and which have to be supported in order to run l3build?

@MarcelKrüger The engine list is in checkengines, you also need to use specialformats

Just realized, that as a micro optimization, having spaces around the key and value in expkv's space stripper is faster than without spaces (still a macro expecting user input shouldn't introduce additional spaces, as the space stripper only strips one space on either end). (This observation is not true for l3keys and packages using expl3's space stripping code)

Someone with admin rights can you please have a look at: tex.stackexchange.com/q/552001/3929 doesn't the person have access to edit their own question (which is of very poor quality)

@daleif As long as they'd stayed logged in

@JosephWright then I don't understand why the person insist that they cannot editor their question. The MWE was quite important (they placed it in the chat) as it easily showed the problem

@daleif 'Can' in a technical sense and 'manage to do it' are not the same

@JosephWright pointing to .org is good, and check the cgi-bin/texfaq2html path

@cis Verbatim-in-argument

@cis You cannot wrap up verbatim inside an argument

@cis because the moment #1 is read as an argument it gets tokenized (fixed category codes) and the category code changes of \verb and verbatim can't apply anymore.

@JosephWright OK, is there a short way to solve?

@cis You'd have to do it the other way around.

@Skillmonlikestopanswers.xyz Ehhh, what? :()

@cis first read verbatim and then reapply catcodes.

@cis I'm writing an MWE.

@JosephWright @DavidCarlisle Just fix the DNS. Both for tex.ac.uk and www.tex.ac.uk

@StefanKottwitz Hopefully I have, but now GitHub is not happy: at least it's Not Your Fault

@JosephWright Let me know if we should have it separate again with full control and doing frequent mirroring from the Github source

\documentclass{article}

\usepackage{xparse}

\ExplSyntaxOn
\NewDocumentCommand\myitem{v}
  {
    \item \texttt { #1 } \\
    \tl_rescan:nn {} { #1 } % this is like reading #1 again
  }
\cs_new:Npn \mycmd { This~is~a~macro }
\ExplSyntaxOff

\begin{document}
\begin{itemize}
  \myitem{\mycmd}
\end{itemize}
\end{document}

@cis ^^^^

@Skillmonlikestopanswers.xyz Oha, I was afraid of it. Thanks anyway.

@Skillmonlikestopanswers.xyz There is a solution with tcolorbox as well:

\documentclass{article}
\usepackage[most]{tcolorbox}
 \tcbset{colback=white, colframe=white, fontupper=\ttfamily}

\newcommand\aaa{the command}
\begin{document}
\begin{tcblisting}{}
(1) \aaa
\end{tcblisting}
\end{document}

@cis guess what that's doing internally :)

@cis well, of course there is equivalent code outside of expl3 as well. \tl_rescan:nn is essentially not so different from \scantokens (but using \scantokens directly is a major pain in the ..., so be glad that you can use \tl_rescan:nn) and changing category codes to read something verbatim was a known technique long before xparse as well. Still, imho, expl3 makes all this way more convenient.

@Skillmonlikestopanswers.xyz The only inconvenience (depending on the use case, of course) of \tl_rescan:nn is that it reads things with a fixed catcode setting, and we haven't decided on a good interface yet (nor if there will be one)

@PhelypeOleinik I don't really see this as an inconvenience most of the time. And using \scantokens correctly is a much bigger inconvenience.

@Skillmonlikestopanswers.xyz That is true. As I said, depends on the use case. Most of the time it's fine

@PhelypeOleinik Catcode tables ;)

@JosephWright :-)

@JosephWright I saw the many commits you and Bruno pushed... :)

@JosephWright I'll push a test for Bruno's comment on interleaving cctabs with groups

@JosephWright Not too much can be done without overusing \aftergroup, but at least a pretty error message

@Skillmonlikestopanswers.xyz More to do yet

@PhelypeOleinik OK, go for it

@PhelypeOleinik I'll sort the deprecations before I take on my actions there

@JosephWright cctab deprecations?

@PhelypeOleinik No, the general ones: they expire and stop use compiling docs (deliberately)

@Skillmonlikestopanswers.xyz That is my abstract example. That is probably not understandable as I imagine it? My problem is the last option.

% arara: lualatex
% arara: lualatex
% arara: ghostscript: {  resolution: 250  , device: png16m}

\documentclass{article}
\usepackage{verbatimbox}
\usepackage{tikz}
%\usepackage{xcolor}
\usepackage{expkv}
\usepackage{expkv-def}

\tikzset{
CommonStyle/.style={baseline=0em, anchor=base, remember picture},
SymbolTextColor/.style={text=#1},
SymbolTextColor/.default={black},
}


\newcommand\TheSymbol{%
\noindent\begin{tikzpicture}[CommonStyle]
\node[] (tempcoordinate-0) {S0};
\end{tikzpicture}}

\newcommand\TheMap{\foreach \n in {1,2,3}{%

The solution is probably very simple to program.

I could ask the question on the main page. But I could probably only take out one of the two options 'arrows' or 'startpoints' to shorten the example.

@cis there are multiple solutions here. You could create a choice key for stepstrokes which only accepts true or false but does more than a boolTF (namely also setting other keys).

@cis but does it make sense to evaluate map at all if stepstrokes is true? Your output routine for stepstrokes does already incorporate also the map routine, so wouldn't it be more sensible to ignore the map completely as soon as stepstrokes is true?

@cis you can't do that with a bool key, those only accept the values true and false. What you could do is using a data key and adding a special flag as the default if no value is given, then you wouldn't use the \myarrows as the \mycolor data key (so not directly forwarding to \color) but check whether the value is equal to the special flag or not, and only pass it to \color if it is, else you just draw the arrows.

@cis or you set the arrows key up to be a meta key and forward any value to another set of keys which only affect the arrows, that could result in a a syntax like \mycmd[arrows={color=red, thick}] (similar to what pgf does, but you couldn't access the key arrows.set(color) like in pgf, so no \mycmd[arrows/color=red] as expkv isn't using a directory tree like structure for its keys).

Also I realized I should add a feature to expkv, namely a user definable handler for undefined keys in a set...

operator: what’s your emergency
me: my fridge fell on me
operator: is anything broken
me: some eggs maybe

Lol

Btw the joke is a bit lame ..

Jk

@yo' LOL

@yo' hahahaha

uff I didn't break pdftexcmds -- at least for now ;-)

@UlrikeFischer we'll know whom to blame :)

@UlrikeFischer did you break something else instead?

@DavidCarlisle not yet, but I could try to create a pull request for colortbl ;-)

BTW, we're just discussing how and when to enable the strict (halt-on-error) mode on Overleaf...

@JosephWright latexcgi live on learnlatex

@yo' switch over to learnlatex.org interface?

@DavidCarlisle ooh ;-)

@UlrikeFischer if it all goes wrong I'll blame you.

@DavidCarlisle ;-)

@egreg anyone, why does [nonewpage]{imakeidx} cause a warning that makeindex needs to be run by hand?

\documentclass{article}
\usepackage[nonewpage]{imakeidx}
\makeindex
\begin{document}
Some text about Foo\index{foo}.
More text\index{baz!bar}.
Even more text\index{alpha@$\alpha$}.
More text about a different part of baz\index{baz!wibble}.

\clearpage
Some text about Foo\index{foo} again, on a different page.
Even more text\index{beta@$\beta$}.
\printindex
\end{document}

@UlrikeFischer I even read the doc and it doesn't mention this:-) not sure why imakeindex can't force a makeindex run next time by writing to the aux file? I should look at at the code one day, not tonight though:-)

@DavidCarlisle i have been wondering too if the new shipout would work, but not tonight either ;-)

@UlrikeFischer alternatively I'll have to do what the doc at latexcgi says I plan to do which is to add a % !TEX makeindex possibility to force a makeindex run (gets more like arara every day, but written in a proper language:-)