@DavidCarlisle 🚫🍍🍕🚫

@CarLaTeX STOP, put pineapple on pizza, STOP?

@UlrikeFischer is Bär's back alright?

@UlrikeFischer ooh

quack

Which key types should a key-val defining macro support? I currently have the following (with e, g and x variants for some of them):

\ekvdefinekeys{test}
  {
    ,code     keyA = \csname tl_analysis_show:n\endcsname{#1}
    ,noval    keyA = \csname tl_analysis_show:n\endcsname{this is code}
    ,bool     keyB = \keyB
    ,boolTF   keyC = \keyC
    ,store    keyD = \keyD
    ,int      keyE = \keyE
    ,dimen    keyF = \keyF
    ,skip     keyG = \keyG
    ,toks     keyH = \keyH
    ,box      keyI = \keyI
    ,qdefault keyI = text
    ,choice   keyJ
    ,         keyJ: 1 = a
    ,         keyJ: 2 = b
    ,         keyJ: 3 = c
    ,default  keyJ = 1

@Skillmonlikestopanswers.xyz Stop putting pineapple on pizza

@CarLaTeX I love pineapple, but it doesn't belong on pizza.

I hate that the SX-chat auto-captures me on input when I type a number. I can't navigate my browser when I'm in this chat :(

@Skillmonlikestopanswers.xyz Might need a clue on what some of these do ..

@Skillmonlikestopanswers.xyz See the recent .cs_Set:Np addition to l3keys?

@DavidCarlisle is it expected that this fails:

\documentclass{article}
\usepackage{graphicx}
\begin{document}
\includegraphics[type=pdf,ext=.pdf]{example-image}
\end{document}

Interesting, haven't used my home PC for a while. Just ran a tlmgr update (full system) and it fails to rebuild the formats because pdftexcmds is missing

tlmgr: skipping forcibly removed package: pdftexcmds

@MarcelKrüger I'm wondering about luaotfload update. I'm inclined not to make an update before the texlive freeze. What do you mean?

@UlrikeFischer define "expected":-)

@UlrikeFischer actually yes I think it is, once you start to specify the bits you need to specify them all, this works: \includegraphics[type=pdf,ext=.pdf,read=.pdf]{example-image}

@DavidCarlisle "we only forgot to document it ..." ;-)

@UlrikeFischer no one reads the documentation anyway

@DavidCarlisle The documentation only says that you need type if you use ext. It doesn't mention also read.

@UlrikeFischer haven't checked but I suspect that the default read file is .bb there so you don't need read= only need it if you want it to work.....

@DavidCarlisle hm, the logical argument is a bit to complicated for me ;-). But I will close latex-project.org/cgi-bin/ltxbugs2html?pr=graphics/4042, saying that it works now but that one needs read too if one want to use type and ext.

@UlrikeFischer the whole split doesn't really make a lot of sense in the modern era where the tex engine (or a script allowed by restricted shell escape) can read the image file directly.

@UlrikeFischer I agree. The luahbtex version in TeXLive is too old to support the new stuff anyway.

@MarcelKrüger good. Seeing the callbacks in the color code - I guess we should tackle the "unique" callback problem at some time too?

@UlrikeFischer You mean for mlist_to_hlist?

@MarcelKrüger yes.

@Skillmonlikestopanswers.xyz me too

@JosephWright I think the syntax of .cs_set:Np in the documentation misses a \marg{value}... No, none of them allow you to specify a weird argument syntax. Which one of those in the list do you think is ambiguous?

@Skillmonlikestopanswers.xyz qdefault

@JosephWright default does <key-code>{<default>}, while qdefault does \exponce{<key-code>{<default>}. There is also edefault which does \expanded{<key-code>{<default>}}.

@DavidCarlisle I was trying the test suite, and I get a failure for foo bar.eps. epstopdf complains about too many arguments. Does this work for you?

\documentclass{article}
\usepackage{graphicx}

\begin{document}
\includegraphics{foo bar.eps}
\end{document}

@DavidCarlisle and lualatex can not trigger the epstopdf conversion of foo€€€.eps. \includegraphics{foo€€€.eps} works only if I first run pdflatex.

@UlrikeFischer I adapted luaotfload to use the new post_mlist_to_hlist_filter callback, so now I thing we set no exclusive callbacks (in the color code) anymore.

@DavidCarlisle @UlrikeFischer I just noticed that the latest version of luacolor breaks PGF shading: PGF uses \saveboxresource with the resources keyword, while the code only expects a number.

@MarcelKrüger oh, I thought you had used it because the new callback isn't there yet not that it was an oversight ;-).

@MarcelKrüger ups.

@MarcelKrüger well it's OK, because we can blame you:-)

@MarcelKrüger I guess we can still use the token library to check for the keyword and put things back in the right order?

@DavidCarlisle I'm not a ho-tex member, so I can't be blamed for that. :)

@DavidCarlisle Yes, I'm working on a fix.

@DavidCarlisle I thought Marcel changes the code, you upload and I get the blame.

@MarcelKrüger do you want to be a member (save making a PR)

@UlrikeFischer actually the ctan announce message just blames Heiko for everything (we probably should change that one day)

@DavidCarlisle If that's not too much work it would be great, but I also can just send a PR.

@MarcelKrüger you should have an invite from gh

@DavidCarlisle Thanks. Does luacolor has to be compatible with old engine versions?

@MarcelKrüger I cleared out all the code for 0.3 syntax for \directlua at the last release, also the pre 0.85 code when pdftex names are primitive, so basically no you can assume luatex 1.x at least , I think unless you mean something very new

@JosephWright Hi, please I have one thing (actually Overleaf-related this time): this question tex.stackexchange.com/questions/458865/… is the top TeX.SE hit if one googles for "overleaf arxiv", however, it's not very useful as it explicitly deals with "document with errors". Would it be fine to make the question a bit more generic, so that we could provide an "official" answer to it?

@yo' There is nothing to stop you posting an answer, provided you say somewhere that you have affiliation with Overleaf

@JosephWright yep I know, but I would prefer to rephrase the question to make it slightly more generic, which I already see as potentially controversial, so I better ask...

@yo' I think it's fine

@JosephWright ok thanks a lot!

@yo' Provided the core idea of 'just send the PDF' stays

@JosephWright yeah, don't worry, we want to make it more useful to users :-) (I just don't like that the question focuses on the errors, which are not the most crucial thing. The most crucial is getting the biblatex/biber version right when it's in use in the document)

@yo' Fine, cool

@JosephWright (it won't happen now, we are doing some hopefully useful changes to the Overleaf interface soon)

@MarcelKrüger the ho-tex files have the version number in 100 places, but l3build tag auto should bump the version by one and update everywhere needed

hi all, anyone know of a package that creates a wrapped bibliography? as in each entry follows on the same line from the last, perhaps separated by some kind of dot or large point. I don't have any MWE so am hesitant to post as a question on the main site

@astronat This is similar: tex.stackexchange.com/questions/209571

@DavidCarlisle I pushed the new version to GitHub. It was a bit more complicated because I triggered a LuaTeX bug in the progress but it should work now. Also I changed the tag auto script to avoid changing older log entries.

@MarcelKrüger a known bug or a new one?

@MarcelKrüger thanks (did it change old entries? I'll look at your diff:-)

@UlrikeFischer I remembered seeing it before, but it was never reported. Now I have a fix, so I'll send it to Luigi. Then it might even make it into 1.12.

@MarcelKrüger your comment says the segfault is from returning many tokens to tex, but here aren't you just returning the keywords to the \saveboxresource command?

My original code scanned the token list arguments to token.scan_toks in order to get exactly the same scanning behavior as the primitive. Afterwards, the tokens had to be returned to TeX (using token.put_next because tex.write can't handle token arguments in older versions). But the expanded resources token list happened to be quite long, triggering a segfault. The new code avoids this by making sure that the token lists are never read by Lua.

@MarcelKrüger yes I guessed later you meant the general text bits were long, how long is long, thousands of bytes or do even reasonable sizes segfault? anyway I'll package up and send it to ctan, thanks?

Therefore (almost) no tokens have to be returned to TeX, but it leads to different scanning behavior: E.g. the resources and attributes have to delimited by explicit braces, while the primitive would be happy even if a } is replaced by \csname egroup\endcsname or similar. Also instead of relativly clean Lua code it became a more complicated combination of TeX with quite a lot of \directlua calls inside.

@MarcelKrüger yes was just following that path just now:-)

@DavidCarlisle The error appears starting at ~150 tokens, so it was triggered e.g. by PGF shading.

@MarcelKrüger it's surprisingly easy to segfault luatex:-)

@MarcelKrüger your change to the auto tag is for cases where the existing file date is inconsistent and uses both / and - ? as I just guarded Version with oldv but you used the regex version of oldv ?

@DavidCarlisle The old version never actually matched if - was used because - has a special meaning in Lua patterns (it's like * but matches the shortest possible instead of longest possible match) That can be fixed by either replacing - with %- to escape it or by using it inside of a character class.. For consistency I used [-/].

@MarcelKrüger ahhh (i knew about - in theory but clearly not at the time I did that tag code, and oddly enough all of Heiko's originals use / dates:-) Hmm I'd better copy that to the other repos (I did wonder about having a shared l3build setup for all the ho-tex repositories but having split them up it seems best to keep them sepaarte but now there is a certain amount of copied setup....