@CarLaTeX Nice palindrome!

\documentclass{article}
%\usepackage{siunitx}
\begin{document}

\begin{tabular}{lp{6em}}
One & two three four five\\[1em]
Six & seven eight
\end{tabular}

\end{document}

^^ @JosephWright seems like a bug in siunitx?

@yo' what is the problem? I see no difference. Ah, I see it now.

@yo' \usepackage{array} - a known feature of this package and one reason Frank won't let me include it in the kernel

@JosephWright ah dammit, so an upstream bug :) (it's important to have someone to blame)

@yo' It's not a bug

@JosephWright ok, then I'm lost I think, but given I never use \\[...] I'm probably fine being lost :)

@yo' It's a deliberate difference in approach between Frank's code and Leslie's: it's covered in the array manual in section 1.1 ('The behavior of the \\ command')

> In the basic tabular implementation of LATEX the \\ command ending the rows of the tabular or array has a somewhat inconsistent behavior if its optional argument is used. The result then depends on the type of rightmost column and as remarked in Leslie Lamport’s LATEX manual [3] may not always produce the expected extra space.

@JosephWright ah ok thanks for the explanation. So in the end, we blame Leslie? :D

@yo' Yes, basically

@yo' Like I said, Frank doesn't want to force this fix on people as it changes design, but really the original behaviour is questionable

@JosephWright yeah, got it.

@yo' I'd rather just force this: as the old behaviour is wrong, I'd tell people to use rollback if they want dodgy tables

@JosephWright I would be with you on that one :-) (and we'd be happy to cover it in our anual TeX Live blog post ;-))

@AlexG (@UlrikeFischer) I've just released an update to expl3 to tighten up various things, including improved graphic inclusion. What I haven't done yet is cover SVG inclusion in dvisvgm, or other 'non-standard' graphics. That's next on my list - should just be a backend adjustment

@AlexG The key is I've fixed various backend bugs, as I've extended testing: I really hope I'm close to covering everything dvisvgm.def can do (other than SVG imagine inclusion, a I said)

@JosephWright Who decides what to fix and what not to fix?

(To me this sounds like something broken that just longs for a fix.)

@mickep It's documented behaviour in Lamport's book, so it's a feature in that sense

@mickep Well the team, but this is tabulars and long-standing, so we defer to Frank - he is after all the lead

Hm hm, ok.

@mickep Anything that changes spacing is basically a no-no for the LaTeX kernel - if it's at all conceivable users might rely on the behaviour, we have to keep it. Here, users might reasonably have 'adjusted' their tables on this basis

@JosephWright but I think in the tagged pdf project we can change such things. Perhaps we should start with loading array directly.

@JosephWright Yes, but two wrongs do not make a right... Don't you ever feel like "Ah, let's freeze what we have now, and let us go on and fix things."?

@UlrikeFischer Good point

@mickep Oh, of course: I am well-known on the team for pushing hard for fixing things - it's just that we have a different approach to ConTeXt

@JosephWright Oh, I did not mean to compare with context. I understand that the situations there is different. I was merely thinking that there must have happened a lot since latex first arrived that could be done differently, and that require people now to make "bad solutions" just because some old strange code is there (probably for a good reason by then). (Since you seem to agree, I stop here.)

@mickep It's more common that people use packages for stuff that now we'd like them not to: there, the team are proactively seeking to 'retire' packages, e.g. atbegshi

@JosephWright Yes. Looking at questions it is not so rare to see double dollars or eqnarray either...

@mickep Double dollars are interesting: formally, we could trap that using active $, but we are stuck with eqnarray

I wonder, is anyone aware of an environment using other characters than [a-zA-Z0-9] in their names? I'm thinking of some weird stuff like spaces, [_-@] etc.

@yo' by convention, only [a-zA-Z]. I've seen some internal environments use @ in their name, but I guess you're talking user-level?

@PhelypeOleinik well, if @ are used in .cls\.sty, we should include it really. (We don't have a separate grammar for these files in mind ATM.

@yo' Well, then @ is pretty common. I can find examples if you need. Also * like in tabular*.

Resisting the urge to join the German language site just to say how completely wrong an answer is w.r.t German syntax. :)

@PhelypeOleinik oh the star! How could we have forgotten the star!

@yo' Does + count as weird? \begin{+pmatrix}...\end{+pmatrix}

@samcarter_waiting_for_siparty is that a thing?!

@yo' from tabularray

oh ok, this snowballs quickly :)

@yo' with lim n->inf number of packages, you always find something :)

@samcarter_waiting_for_siparty yeah. So far we found out that \newenvironment{a~b}{}{} is not a good idea, all other things work, including \newenvironment{a_b$c d}{}{}

@yo' In mathtools you have \newenvironment{MT_gathered_env}, and I bet there is a template somewhere that uses some oriental script in the environment name.

@yo' \edef~{\string~}\newenvironment{a~b}{}{} and now it works :)

or just \newenvironment{a\string~b}{}{}

@PhelypeOleinik wasn't there something like an Indian (?) package with macros and environment in their native script?

@samcarter_waiting_for_siparty I don't remember what language that was, but there certainly is

For the sake of sanity, we decided that non-ASCII is not a thing w.r.t. csnames and envnames

@yo' :thumbsup: just make everybody write in English :)

@yo' I don't know what you are doing, but really anything that expands to characters is game :) Is that for syntax highlighting?

Then comes ABNTeX and creates \begin{sumário} (@PauloCereda)

@PhelypeOleinik syntax highlighting and code check (linting)

@yo' My suggestion would be to take anything between a { and the next }. You can have all kinds of weird stuff like environments in macros with \begin{#1}, or \begin{\@envname}, and so on... If someone sticks something there that isn't a character, it's not your fault :)

@PhelypeOleinik well, \begin{\@firstofone{\a_b:}} is valid in expl context, and contains } :)

@yo' Well, yes, balanced braces then :)

Yeah, we'll talk about it in more detail tomorrow :)

Haha, I didn't see this answer coming. tex.stackexchange.com/a/640411/52406 :)

@JosephWright But it is not "trapped" now, right? Why stuck with eqnarray? Why not just remove it? Is it used by some "sane" macro?

(upvoted the answer of course, for the joy of it)

@mickep users that send things with eqnarray to arXiv, for example

@PhelypeOleinik Yes, but wouldn't it better to give an error on that. That way people will be forced not to use it.

@mickep For some, it would, but some people would just come angrily shouting that "LaTeX developers don't care about their users"

@PhelypeOleinik But as soon somebody uses it (1251 hits on the main site) people show up very quickly and say that it shouldn't be used, and that there are replacements (amsmath, mathtools, ...). Isn't that more like caring for the users? To point them to good solutions.

@mickep $ is just catcode-3 in LaTeX, so we'd need to adjust to catcode-13 to gain control: not risk free

@mickep Remember we still have a mode in LaTeX to process 2.09 docs, which predate 1994(ish) - there's a long tail

@mickep Yes, I agree, but some people just like to not hear arguments :)

@JosephWright One could let eqnarray live in 2.09 without it staying in 2epsilon?

@PhelypeOleinik Oh, so this is even sensitive among the developers?

@JosephWright I see.

@mickep Yes, but what I mean, and @PhelypeOleinik is getting at too, is that there are millions of existing LaTeX documents, and we can't break a large number of them - eqnarray is there is the docs from before the team took over, so it's part of standard LaTeX. The absolute best we could do is add a warning to it

@mickep No, I mean users. For example, we were shouted at (this was a weird case though) for including xparse in the kernel, which cannot possibly change a document in any way

@mickep We had the whole LaTeX3 idea to free us from this, but that has it's own issues (essentially: no users will ever pick it up), so we had a change of plan

@mickep Any change at all to the LaTeX kernel is emotional for some people: they'd rather it was truly frozen, and users made changes only using packages. But that doesn't exactly scale brilliantly

@JosephWright Isn't it better to give an error (maybe with an instruction how to enable it)?

@mickep We can not break lots of existing old documents, only to educate the writers of newer documents.

@PhelypeOleinik Shouted at by who? Users?

@mickep Then you are back breaking documents - the run stops and people complain

@mickep Some users, ones known to us ...

@JosephWright Well, their old installation will work... :)

@mickep one user ;-)

@mickep Specific ones :)

Something is rotten in the...

@mickep We (team) have to be very mindful of vocal users who might raise issues that then lead to concerns about the continued stability and usability of LaTeX - that's a central part of the offering, and we worry about it a lot

@JosephWright So, freezing whatever there is now, and move along. :P

(I know that there are latexers who agree with me on this point... Even ones inhere)

@mickep That was the LaTeX3 plan, like I said - it really didn't look viable

@JosephWright Why do you think it would not be used?

(Even large publishers (Springer/Elsevier) uses eqnarray it seems, or let authors do. It is a shame.)

@mickep There are several 1000s LaTeX packages at present, and even with a significantly expanded kernel, we could only hope to cover the functionality of a small subset. It's likely such a move would not carry packages with it, so they'd all be 'killed'

@mickep The 2.09 -> 2e move was touch and go, and it was a lot less severe and with a much smaller corpus of existing documents to worry about

@mickep To a first approximation, we'd have no users - but I'd say that's also true for ConTeXt

@mickep If you need users to opt in, most won't, so wouldn't get any gain from the work - hence the change of focus. For example, the tagging project is all about allowing existing sources to produce tagged PDFs without having to edit, at least too much

@JosephWright Well, I think people would adapt packages after a while to a new version. And the current version is still there. (I understand that it is not a small step, but staying with old solutions of things forever sound like a worse option.)

@mickep They might adapt, it's true, and that was the plan for a long time, but at least to those of us on the LaTeX team, it really doesn't look viable

@JosephWright In context tagging is just working perfectly well. \setuptagging[state=start]. If someone writes a module they will probably work hard to break it as well.

@mickep Oh, I know that :)

@mickep In LuaTeX it's relatively easy (as ConTeXt also has the internal data structures)

@JosephWright So, looking 20 years ahead, you think that there will still be eqnarray and strange spacing in tabulars with \[1ex] depending on the type of cell (was it like that?)

(It is easy for me to say things like this since I'm neither a coder or have any responsability against users...)

@mickep Probably yes on eqnarray as it's in Lamport's book; @UlrikeFischer and I might manage to convince Frank about the array business, particularly if longer-term plans come off and we can move away from \halign internally

@mickep A lot of the current work is refactoring 'from the inside'. For example, we are likely to make all active chars \protected, which makes some changes a lot easier to do. This is stepwise but we are moving toward the right outcomes

@JosephWright Sounds good! (And trust me, I am also a latex user, and really appreciate all the (unpaid!) work everyone is doing!)

@mickep One thing you might notice, if you look carefully, is that the way we describe things in LaTeX News is not always quite what happens in the code - we are slowly making things \protected, for example, but not exactly advertising it

web.mit.edu/jemorris/humor/500-miles

@JosephWright Thank you, Joseph! I will try it when it is available via tlmgr.

@samcarter_waiting_for_siparty Haha, good one!

@JosephWright I have not seen it. To me it sounds like protecting some stuff might have large consequences and not being backwards compatible, but I will not complain. :P

@mickep We check quite a lot - Frank complained at me about something last week here :)

@JosephWright So, is it a too bad simplification to say that you (and some others?) try to push forward, but Frank (and some others?) are holding back?

@mickep :) I think I'm going to abstain from complaining about debugging for at least my next 12 bugs. This shows how much worse things could be :)

@mickep no that isn't true.

@mickep Er, that's one rather extreme reading - I'd say more that Frank has more experience than me :)

@samcarter_waiting_for_siparty Well, it shows that upgrading can be dangerous (the opposite of what I try to push here...) :)

@JosephWright OK, I asked since I only heard his name regarding keeping things so far.

@mickep yes it can :) There are two types of programs: the ones for which you look forward to and update because all the new toys and features to play with and the ones where the message "a new update is available" just results in a groan "Oh, no, don't break it again" (for the later case, I'm looking at you, firefox!)

@mickep Frank is working pretty hard on new ideas, it's just that they likely don't show up for end users: things like paragraph hooks are a big change and are needed for tagging (we can't just require LuaTeX/LMTX), a new marks mechanism, etc.

@JosephWright Good to hear. (Freezing would allow for moving on to luatex, but I'll stop the nagging. I'm just curious, since I see more or less only pros with moving on.)

@mickep There are a lot of specialist fonts only available in classical formats, so T1 or similar encodings; those don't work properly with LuaTeX ...

@mickep LuaTeX deliberately isn't 100% compatible with pdfTeX/TeX90, for example in line breaking, so it would alter lines in typeset material, and that's a real issue

@JosephWright Indeed, as a start, they would only work in the frozen version. :)

@mickep So what happens if you want to re-set them as part of a collected edition or similar? OK, you could do some old-LaTeX -> PDF -> include as a PDF in new LaTeX, but most people would prefer to run all of their sources with one engine

@JosephWright No, old documents will continue to work as intended in the frozen version. Isn't it true that if one writes a book in LaTeX one keeps the installed version to be sure to have the intended output?

@mickep freezing wouldn't change anything. You still would have to consider how to communicate changes and get people or packages authors to use them and how to do more changes without breaking existing documents.

@JosephWright Well, if you re-set them, you re-set them. Doesn't that usually mean you have a new/slightly different layout anyways?

@mickep To be 100% sure, yes, but that's largely due to non-team packages (or fonts): the team code should work with any source for at least the past 25 years

@mickep No, that's rather the point for some use cases (collected editions of mathematics, for example)

@mickep You might want the main typeblock to stay the same, just alter the page numbers, headers, etc., as part of a collected volume

@JosephWright OK, so that mean one could never move to luatex/xetex, since it will potentially change the output slightly. That is sad, I think.

Any German speakers around? I have a quick question.

@mickep It means we could never move only to LuaTeX; for new documents for many users, there's no downside - it all depends on the use case

@mickep For work in western European languages, pdfTeX is always going to win on performance, which is something one has to bear in mind also

@samcarter_waiting_for_siparty He's right, but I wonder why he picked SO - it's a problem generally, and there are lots of media companies who I'd say are a lot more obvious

@AlanMunn Wie kann ich helfen?

@JosephWright Well, but coding things for different engines is a down side, right? I remember the other day when you were going through the graphics file name parsing...

@mickep Most of the time it's not too bad: I do the backend stuff, which yes is a little more complex, but then even with LuaTeX-only you need to think about dvisvgm, dvips, etc.

@JosephWright I'm not sure pdftex is faster than lua(meta)tex. Perhaps for "hello world" but I have a feeling that lua(meta)tex scales well. But it is difficult to compare..

@JosephWright dvi could officially be dead. :)

@mickep I'm pretty sure it's faster: with 8-bit internals and much more limited font structures, there's a lot less to do

@samcarter_waiting_for_siparty In the German interrogative pronouns (wen/wem etc.) are there any forms that are the same but correspond to different cases/genders etc. I.e. same pronunciation but one is e.g. dative and the other accusative (or nominative)?

@mickep Defo not!

@JosephWright luckily there are addons to automatically remove such banners :)

@mickep It's a much faster output format than PDF, and you can post-process it to get lots of end-user docs from a single TeX run - for high-throughput systems that's important

@AlanMunn oh sorry, that's too theoretical for me.

@JosephWright Haha, OK.

@JosephWright I tried on a document with some math and metapost figures and so on, and I get 20-25 pages/s. Not too bad...

@JosephWright Btw, thanks for the patience answering my questions. :)

@mickep Once again, you don't have users complaining :)

@mickep Well, trying some relatively simple document with some math and structure in pdfTeX I get around 400 pages/s. And that's with an unoptimized debugging build of pdfTeX...

@MarcelKrüger Wow! I never heard of those speeds before!

I always have this problem with texstudio. Can I post question on it on main board? If I have a little larger file than normal, it hangs opening it. I have now a Latex file which is only 100,000 lines. After waiting 5 minutes, it crashes. I have very fast PC (one of the fastest) with 128 GB RAM. And using latest version of texstudio. Crash screen show below. Windows 10

So now I use notepad++ to open my large Latex files, since I am afraid to use TeXStudio, else it will crash and I could lose any unsaved work. I guess no one tested texstudio on Large latex files before releasing it.

@Nasser 100k lines it too much for a lot of editors

@mickep I guess having a rather fast system helps, but based on many attempts to get people to switch to LuaLaTeX I get the impression that the speed difference hits a rather critical window: For many typical math papers pdfTeX is on many systems just fast enough to feel more or less instant. Even relatively small slowdowns make the compilation time much more noticeable.

@Nasser Maybe ask on their bug tracker? The developers mights be better equipped to debug the problem?

@JosephWright I did not know this. It opens instantly using notepad++.

@Nasser (also, did you try with syntax highlighting switched off?)

@samcarter_waiting_for_siparty Yes. I basically turn off all these highlighting stuff first thing I do when I install it.

@Nasser Notepad++ is a 'programmers editor', mean for very large files - TeXstudio is not. I can't imagine having a useful TeX source that long (siunitx v2 is too long at 18k lines)

The file is only 2 MB in size.

Could someone please if they have texstudio, just try to open the file to see if they also get hang and crash? Make sure to save your work first. I just need to know if it is my system (it is new PC) or not. Please find the file in this folder (safe, my own page) 12000.org/tmp/large_latex_file it is called report.tex

@Nasser Works fine on TXS 4.2.3 on macOS 11.6

@Nasser I can arrange to have TeXstudio

@Nasser Opens fine in TeXworks here (I use the dev version, which fixes some large file issues) - which platform are you on? I'll download TeXstudio to test, but I want ot be sure I pick the right one ;)

@Nasser IMHO it's not exactly related to the size itself, but to all motions editor components do while rendering text (syntax highlighting, lint, intellisense, etc).

I have 4.2.2 on windows 10

@PauloCereda Fair: TeXworks used to have issues with syntax highlighting and larger files, for example

@Nasser Works for me

@MarcelKrüger In our department people are more like not moving because they dont know that luatex exists. But then again some of them probably (still) use eqnarray and stuff as well, so ...

@JosephWright :)

I just did an experiment. I had about 20 files also open at the time in TexStudio. So I closed them all. Closed T.S. also. Opened T.S. again and opened same file. Now it opened instantly ! So I think it is because I had many other files loaded in the editor at the time?

@JosephWright Ok, thanks for trying.

@JosephWright One thing you did not answer was if there was any feedback on the ontarget-math. Maybe you missed the question, or that is just too sensitive stuff(?)...

@mickep There were some positive thoughts, @egreg said he though the syntax was horrible ;)

@mickep It's nice stuff, but the question is whether it will be available to LaTeX users, as that needs an engine that we can work with

@JosephWright Oh, syntax of what? :)

@JosephWright Well, it will change spacing in current documents, so from what I read above, probably noone will take it up. :(

(thanks!)

@JosephWright I managed to make it hang again. Could you please if you have a minute try this on your system since you have the file? I simply opened it. It opens OK., Then I selected all text inside it (starting from after \being{document} to end of file) and clicked the DEL key to delete this selected large amount of text. Now it hangs. been like this for 10 minutes. This editor is not very robust. It takes this long to delete text? Why?

@mickep That you'd have to ask him about

@Nasser Deleting the whole document body takes a bit longer than usual, around 3 seconds for me.

@mickep Ah, don't think I'm objecting :) The question is always how to manage change. For example, on the array example, I'd just say load the package and adjust the input. And as @UlrikeFischer says, we now have a marker for 'updated docs'

@MarcelKrüger Why do you try to get people to switch to LuaLaTeX?

@samcarter_waiting_for_siparty This is really strange. Why it took almost 12 minutes for me? which version are you using? I am using 4.2.2 Do you have the "structure" left panel open at same time? This might have an effect. (view->Show->Side panel)

@Nasser No, I had structure closed. Let my try again

@AlanMunn Not sure I completely get the question. But do you refer to the case of the answer? E.g. “Wohin willst Du?” could be answered with dative or accusative.

@mickep Please don't get me wrong: these discussions are really useful

@FaheemMitha Full Unicode support, system fonts, build-in programming language, callbacks to lots of TeX processes, etc.

@Nasser This might be the trick. I'm seeing the beach ball of death (mac equivalent to the sand clock) for over a minute now.

@JosephWright Ah. Callbacks to lots of TeX processes, such as...?

@FaheemMitha Linebreaking is an obvious one

@mickep What are you referring to?

@JosephWright I will. @egreg, do you care to elaborate about what was the syntax you did not like with the stuff in ontarget-math?

@JosephWright Hmm. I'm actually not sure what a "TeX process" is.

@samcarter_waiting_for_siparty good. sorry I did not mention you need the structure panel open. I think this is also what caused the earlier crash when I had 20 other files open.

@egreg The work Hans is doing on extending math mode parameters

And I thought regular TeX did fine with linebreaking on its own.

@JosephWright But from what I understood from our discussion today, how to move on at all if nothing is allowed to break?

@JosephWright Yes, and hopefully something good will come out in the end :)

@Nasser might be. I usually don't have the structure open and never had a problem with many files. At one point, I opened several hundred files and texstudio was doing fine.

@mickep We spend a lot of time looking for ways to square that :) For example, rollback covers the worse problems, but we also test a lot to make change opt-in or dependent on clearly 'new' features (like \DocumentMetadata)

@JosephWright I think it is good that you are in the team. :)

@mickep Oh! The way parameters are specified, really complex and unfriendly (but that's very common with Hans and ConTeXt).

@egreg Maybe it could be done better. But this is the kind of thing that is done once (if you install, have a look at math-ini.mkxl). (Not sure what parameter you have in mind, exactly.)

@mickep -- Someone mentioned \eqnarray. I think the copyeditors at AMS wouldn't notice this being used unless one of its characteristic bugs (overprinting equation numbers) was present. So it may be present in lots of AMS journal article files, which are archival. Stability is urgent in a production environment.

@barbarabeeton Indeed, so a frozen latex version could be used for that. :)

@samcarter_waiting_for_siparty did the delete text finish now on your end? if so, how long did it take? thanks.

@mickep -- Actually AMS stores with each article or book the version of software and any packages. It's frequent enough that something is resurrected for a "collected works" that this is essential. So it is possible to always use a version that works. But it does require an enormous archive. There's also an effort to support things like pandoc, which only recognizes $$, but none of the amsmath environments; very poor support.

@barbarabeeton Oh, see see. Good. Regarding eqnarray, isn't also the spacing (horizontally) just wrong?

@mickep Horrible

@mickep -- Oh, I think spacing may not be great, but AMS now runs on "editorial lite". So some things that I learned are no longer observed. General degradation. Nit-picking is no longer valued.

@egreg @egreg, so why not just remove the macro? :)

@barbarabeeton Doesn't even ams care anymore? :(

hey

@oneofvalts Hi

how do i paste not so long source code?

@mickep Why should we inflict an error on people only because they choose to use a not so good command? There are lots of other bad stuff around in various packages. I don't want to spent my time defending such a change.

@UlrikeFischer Well, if we care for good typesetting (don't we?), removing macros that produce bad result could not be a bad thing. But I realise I'm alone here in that view.

\psset{
lightsrc=viewpoint,
viewpoint=30 40 30 rtp2xyz,
Decran=30,
solidmemory
}
\begin{figure}[H]
\centering
\begin{pspicture*}(-4,-4)(4,4)
\psSolid[object=dodecahedron,
a=3,
action=draw*,
lightintensity=1.5,
fillcolor=orange!80]%
\end{pspicture*}

i have this, a dodecahedron. i want to draw a rotational symmetry axis of it. the one that passes through the above and bottom surfaces of it.

@mickep No, but if you were to remove everything bad (which may also include some degree of subjectivity), CTAN would lose a lot of its packages

@mickep well I also care for good wine but I don't go round and remove bad wine from the market.

@UlrikeFischer But when we see somebody drinking bad wine, we scream to them "Avoid bad wine!" :) (tug.org/pracjourn/2006-4/madsen/madsen.pdf, good points, so not complaining about that)

As long as it is there, it will probably be used. The publishers don't care apparently (so we find it in their published articles and books, and our eyes will cry every time...)

@mickep sure tell them not to do it. But I see no reason to force the light on them by making it an error.

though, i don't understand how the solid is positioned, therefore the center of it. i have tried the line that passes (0,0,-3) & (0,0,3) but it doesn't work.

@UlrikeFischer Well, another option would be to fix it so that the output is actually "correct". But I understand from the discussion here that such a thing would probably be even worse. :)

One could even have a flag somewhere use-old-broken-eqnarray that loads the old behavior.

like this

i eventually want to draw other rotational symmetry axis too.

can anyone help me with this? anybody familiar with pst-solides3d

@mickep Too many legacy document use it.

@mickep -- Not as much as some of us old-timers might wish. Cost is one factor.

@PhelypeOleinik But eqnarray is not part of a package...

@egreg In a frozen version, it could be enabled. But why keep it? (It has been wrong for so long...). Well, apparently I'm alone with that idea.

@mickep I know. What I meant is more like: there is so much bad code around, that eqnarray isn't nearly as bad (even though it's in the kernel, so people tend to use it much more)

@PhelypeOleinik That is of course sad, but a different problem. But since eqnarray is what comes by default, one could get a feeling that it is suggested to use it. At least new users could. And they do.

@mickep -- Minority, but not alone. (Maybe a mild warning that it would be better to use an environment from amsmath or mathtools.)

@barbarabeeton I don't know exactly what they cut down on. But it is not that it is for example cheaper to use eqnarray than the amsmath macos :)

@mickep Yes, that's half of the problem: new documents using eqnarray. But we can't remove it from new documents without affecting the other half of the problem (if that was possible, it would probably have already been done)

davidcarlisle/eqnarray-fixed

@mickep and @PhelypeOleinik -- Two recent places that might be checked to see how they deal with \eqnarray: learn-latex and @StefanKottwitz's 2nd edition of a beginner's manual. (I don't have the book, or time to check learh-latex.) But I'd really like to know.

@barbarabeeton learnlatex.org doesn't mention eqnarray at all. I don't have a copy of Stefan's book, but I'd bet he either didn't mention, or wrote against eqnarray. The same is probably true for TLC3

@barbarabeeton No results found for eqnarray site:www.learnlatex.org. <- if that is what you meant by learn-latex.

@mickep -- Yep, that's what I meant. Good beginning.

@PhelypeOleinik -- Yes, I hope it's omitted from TLC3, and I hope someone can check Stefan's book.

@mickep As @barbarabeeton says, a warning might be workable

@barbarabeeton It's not there, I'm pretty sure (I reviewed it for the publisher)

@barbarabeeton Definitely not on learnlatex.org

@JosephWright -- Thanks.

@mickep Like I said, it's in the Lamport book, and we can't really change that

@mickep I often push for more drastic actions, don't worry - I am keen to properly tidy up, it's just I also have users

@mickep For example, I dropped various things in siunitx, both in the v1 -> v2 and v2 -> v3 changes, and got moderate heat as a result

@JosephWright yes, better than nothing.

@JosephWright Well, siunitx seems to, aside of easybook, be one of the more progressive packages. ;-)

@PhelypeOleinik ooh

@JosephWright Moderate heat? From angry users? People are funny.

@Nasser I did not let it run

@TeXnician Thanks. Yes, that's the idea. Are there any of the personal interrogative pronouns that have this behaviour? (i.e. some version of 'who' inflected for gender/case/number)?

@mickep People are creatures of habit. "I've always done it this way, why should I change now?"

@AlanMunn Yes, progress sometimes hurts. But I guess that most users have moved along with @JosephWright to v3.

@mickep For siunitx, the change for complex numbers has been the most 'interesting'

@AlanMunn No, with the personal ones I don't know of pronouns with this behavior. I'm pretty sure wer, wessen, wem and wen all require a certain case.

@TeXnician Ok thanks. That's what I thought, but I wanted to make sure.

@AlanMunn Maybe for numerus it's less clear because they all target singular and plural. Gender as well, they can be used for all genders.

ooh Das Wort

@DavidCarlisle -- You've got mail; a TUGboat request.

@mickep Some people were not happy that \SI no longer allows complex form: it's a breaking change but I have good reason

@AlexG I think I have SVG inclusion working in dev: I'm not clear why dvisvgm.def doesn't use dvisvgm:img but uses a more complex route

@DavidCarlisle I realised I'd missed the multi-dot graphics business - oops :)

@DavidCarlisle Luckily your test files are helping out

@JosephWright do we have something in l3build to run tests with make4ht or htlatex?

@UlrikeFischer Not at the moment: would you like me to think about it?

@JosephWright well I'm changing stuff in hyperref and I'm flying a bit blind here.

@UlrikeFischer OK, on my to-do list for tomorrow

@JosephWright it doesn't need to be something complicated. Simply make4ht file and then log checking. It would at least prevent that one break it completly.

@UlrikeFischer Sure, I'll just play about a bit

@UlrikeFischer I'll check it out in siunitx I guess