@DavidCarlisle sorry. strictly speaking, the check didn't fail because it didn't run. but it built the documentation using a local copy of the package. but the package can't be loaded without error. so even without the test, the build should have failed.

I've uploaded some stupid things to ctan, it can't possibly have loaded the version of the package I think, but I can't think how it can have used anything else.

@DavidCarlisle but the checks also did run earlier today and passed. I must be doing something very wrong ...

@UlrikeFischer e.g. github.com/cfr42/nfssext/actions/runs/13231737868/job/…

Is anyone aware if babel recently changed how \captions<language> is processed? I maintain a template that stopped working with the newest LaTeX verison

(sorry wrong link, the tex-template is publications.copernicus.org/Copernicus_LaTeX_Package.zip )

hm. apparently babel stopped processing \@classoptionslist and started processing \@raw@classoptionslist instead in the last couple of months

@Lupino yes, you should change that too. See my edit tex.stackexchange.com/a/284002/2388

@UlrikeFischer thanks, got it to work. Do you know by any chance when this change was made (for \@iflatexlater{???}{<new>}{<old>})

nevermind, i query with \ifdefined\@raw@classoptionslist…\fi

@Lupino classoptionslist gets passed through an edef to remove white space (terrible idea, I can't think who implemented it that way) which means it is more or less impossible to use key=value arguments if the values may have commands in them. The raw@... lists were added some time ago to give classes and newer kv option handler code the chance to get hold of the options as supplied before the classic 2e option handler mangles them.

@DavidCarlisle which still can't be reliably processed, because now you need to survive an \edef while your values aren't passed through an \edef.

@DavidCarlisle that package is eons old and went through several hands before i took it over. but with the next upate, the edef is no more, at least at this particular instance…

In TL 2025 pretest with \includegraphics{example-image.<ext>} I get, using pdflatex, pdfTeX warning (ext1): too small image resolution ignored, <ext> being any of pdf, jpg or png. Is it the same for everybody?

@egreg should be already reported on the mailing list, @UlrikeFischer (or was it @cfr?) mentioned the same yesterday, iirc.

@egreg tug.org/pipermail/pdftex/2025-February/thread.html

@samcarter OK

@Skillmon well yes but you only need to not error (and most commands thse days can be made robust enough to do that), it's not perfect but there are some prices to be paid for not waiting for latex3

@DavidCarlisle I was very patient with L3, not my fault that you didn't manage to release it before I started using LaTeX :P

@egreg "You want \ooalign." that had to be you:-)

@DavidCarlisle Exactly my thoughts when I saw that answer.

@Skillmon "What about accessibility?"

@mickep Unicode is lagging behind in math symbols: they seem more interested in emoticons.

@mickep ooh a11y

@mickep <mtable intent="adele-product">....

@PauloCereda b7t

@DavidCarlisle oh no

@DavidCarlisle Oh, I missed that part ;)

@mickep If that is not what is generated from @egreg's answer code, blame @UlrikeFischer

@DavidCarlisle I should check. And we should merge also your intent branch ...

Is there a reason why bashful doesn't work in LuaLaTeX for me? It doesn't mention any incompatibilities in the docs AFAICT

@yo' example?

@DavidCarlisle ohhh

@yo' does it use shellescape (or otherwise special case luatex) or does it directly use \write18 (which won't work) I could check later, not now...

@yo' it uses write18, load shellescape and then it should make it work again:-)

@DavidCarlisle I think it uses write18, and loading shellesc,bashful made all the difference.

Thanks both!

(btw, I'm still surprised that there's \fmtversion but still you need ifluatex and friends to determine lualatex vs xelatex vs pdflatex. Unless I miss something obvious again.)

\documentclass{article}
\usepackage{shellesc,bashful}
\begin{document}

\bash[stdout]
 echo "hello world"
\END
That 's it!


\end{document}

@yo' I see you figured it out but ^^ works

@yo' iftex package gives you \ifluatex etc, if you don't want a package, the format has already distinguished things \c_sys_engine_str is the current engine \sys_if_engine_luatex:TF is true in luatex, etc

@yo' it comes down to, who do you trust, @JosephWright or me?

@DavidCarlisle :)

@DavidCarlisle doesn't iftex just access the L3 layer?

@Skillmon no (and it works in plain tex, so loading expl3 would be quite heavy cost)

@DavidCarlisle :D

@DavidCarlisle \c_sys_engine_str is what I was looking for, thanks again!

@yo' so that answers the trust question :(

@DavidCarlisle \ifcsname c_sys_engine_str\endcsname <implement using expl3>\else <implement using David's brain>\fi?

@DavidCarlisle well, as long as it works in Overleaf's TeX Live 2024...

@Skillmon not really worth it, usually it only does an ifx on something anyway, and there are some legacy tests for vtex etc that aren't necessarily in the format. It predates expl3 being in the format (and parts of it probably predate expl3) so it is what it is.

@Skillmon not me. I've not tried the pre-test.

@cfr my brain only stored insufficient data about the person who wrote it. I only remembered the gender and that it was a frequent chat user. So could also have been @samcarter now that I think about it.

@Skillmon I mentioned it to samcarter in top answers.

Went with this, sorry @DavidCarlisle

\ExplSyntaxOn
\str_set_eq:NN \l_tmpa_str \c_sys_engine_str
\str_replace_once:Nnn \l_tmpa_str { tex } { latex }
\tl_set:NV \compilerName \l_tmpa_str
\ExplSyntaxOff

@UlrikeFischer or this... :)

@UlrikeFischer thanks for reminding me :)

@egreg -- Unicode has rather strict guidelines for accepting symbols (believe it or not) and it's a definite advantage to have an "in" on the UTC. I suggest that you persuade @DavidCarlisle to champion this. It's best backed up with an example (where the notation is defined) in an item published by a recognized publisher.

@yo' put a \debug_on:n { all } in front of that code block.

Ah, cleaning up old posts on the main site ... So, it is important to look at the date of the original post =D

Oh, crap! I got a tick which will probably prevent me from getting 55,055 =(

@JasperHabicht I know how you feel. I missed 66666 two times

@Skillmon Well, there is still some room for changes until 55555 ...

@Skillmon Two times even!

@JasperHabicht yes, I donated some Rep to David to have a second chance, and missed that one as well

@Skillmon That's a nice idea! Better that downvoting egreg's answers (as often suggested by a certain someone)

How reliable is gs at identifying broken PDFs?

The following errors were encountered at least once while processing this file:
        CMap unrecoverably broken

   **** This file had errors that were repaired or ignored.
   **** The file was produced by:
   **** >>>> LuaTeX-1.18.0 <<<<
   **** Please notify the author of the software that produced this
   **** file that it does not conform to Adobe's published PDF
   **** specification.

@cfr this message is imho wrong. bugs.ghostscript.com/show_bug.cgi?id=708042. ghostscript has been corrected (and luatex in texlive 2025 no longer created the empty beginbfrange).

@UlrikeFischer ah! thank you. so it is just a question of eventually gs will get updated. good.

@UlrikeFischer that means we can not worry that using gs will break search etc.

how can I find out how I managed to generate documentation with l3build for a package which cannot even be loaded, if I can't reproduce the problem locally? I can't reproduce in a chroot or not in a chroot or, generally, at all. but l3build generated a ctan archive on github from the correct sources. the svn numbers match what ended up on ctan. the erroneous code is typeset in the PDF. and other documentation also built with it. but this is obviously BS and not possible at all ....

[that is, I can't reproduce the error of not getting errors locally.]

@cfr but it is not true that your package can not be loaded, it errors but in scrollmode it continues.

@cfr but beside this, your style seems not to come from the current folder when building the doc: /home/runner/work/nfssext/nfssext/build/local/cfr-lm.sty. That looks a bit odd.