@JosephWright I updated the test logs on LaTeX for TL'21: colour stack and line wrapping stuff. There is one test though, m3opacity001, that gives some odd results with XeTeX. Can you take a look at that?

@DavidCarlisle Yeah, thats what I endd up doing -- matching \intercal was very difficult. So \raisebox{\depth}{\scalebox{0.75}{$\m@th#1\mathsf{CT}$}} seemed to work well.

@DavidCarlisle OK. But even if textwidth is the same, the page breaking can still be different, right?

@FaheemMitha yes obviously if \textheight is different or \parskip or \topskip or \finalhyphendemerits or \abovedisplayskip or ... is different, then page breaking will be different.

@DavidCarlisle \finalhyphendemerits?

@FaheemMitha discourages having a hyphen on the last-but-one line of a paragraph,

@DavidCarlisle Oh. That sounds exotic.

@FaheemMitha it is no more exotic than \textwidth (\hsize) as far as tex is concerned, both are consulted every paragraph to control the line breaking,

@DavidCarlisle OK.

Is it reasonable to manually change textwidth for one page, if I want the text to break differently there?

@FaheemMitha no (it is possible if you have a forced (\clearpage)` page break in the changed region but getting the timing right is tricky

@FaheemMitha the usual way is to use an enviornment like quote (or adjustwidth from the package of that name which is the same but wth options to control the indent on each side)]

@DavidCarlisle OK

@DavidCarlisle Thank you. I'll look at adjustwidth.

@PabloGonzálezL ooh

@PauloCereda \begin{longtable}{c}Breakfast\end{longtable}

Runaway argument?
{\tablename \ \thetable \ -- \textit {Continued from previous page} \\\ETC.
! File ended while scanning use of \multicolumn.
  oh no

@DavidCarlisle oh no

@PauloCereda must be user error

@DavidCarlisle ooh

@DavidCarlisle wow ;-)

@PhelypeOleinik Will do

@DavidCarlisle Did you mean adjustwidth from the changepage package? I can't find another adjustwidth.

@FaheemMitha oh probably yes.

@DavidCarlisle The adjustwidth environment doesn't appear to alter pagewidth, at least according to layout. It seems to push the layout display sideways, but all the variables that layout is listing remain the same.

@FaheemMitha no that is exactly the point it does not change \pagewidth or \textwidth it changes the width of the text block within that.

@DavidCarlisle Oh. I misunderstood, then. But isn't just changing \textwidth on the fly a reasonable thing to do?

Just including the geometry package changes the page parameters. It seems to me that violates the principle of least surprise.

@DavidCarlisle So, the bottom line is that the value of \textwidth does determine the value of the text width, as long as it's specified at the beginning of the document. But it's not so easy to change horses in mid-stream?

@FaheemMitha if you want to change mid-document you need to specify what you mean by that and the answer depends on the details and some are harder than others, do you want the page head and foot layout to change, do you want paragraphs that split over the page break to change their line width mid-paragraph, do you want floats that float in to the region with a different width have the new width, ...

@DavidCarlisle I just want a specific page to not break where it does. I want it to break later.

@FaheemMitha so why discuss \textwidth? \enlargethispage{20pt} sounds more useful. (if you want the page to be bigger) or if you want to squeeze the text into the normal page size, do that. You nkow this is the 10000th time you have asked a question without providing any test case or hint of your real problem.

@DavidCarlisle OK, I'll try that too.

@DavidCarlisle 10,000 sounds unlikely. And I can't very easily provide a test case, because it's embedded in the middle of other stuff.

So if I provided a test case it would likely be quite complicated, and nobody would be interested. That's happened before.

@FaheemMitha you can always supply a mwe and maybe 10000 was a over-exageration but I just looked and you have been asked dozens of times for a MWE in this chat for different questions.

@JosephWright -- There's not a page on the meta site for "community promotion ads -- 2021". The one still used for reference is 2020. I'm trying to get the TUG 2020 ad replaced by one for 2021, and it's not clear whether to replace the image there, or wait for a new page to replace 2020, to post a new ad. (Presumably, if/when the 2020 page is replaced, everything there will go away, and we'll have to go through the whole exercise again.

@FaheemMitha -- One thing that hasn't been mentioned yet ... since the narrower text width causes the material to break to a second page, I'm assuming that it does already start a new page. So you might try changing the type size to something smaller, say 10pt. With a 12pt base, that is probably \small. That way, you wouldn't have to monkey with margins. The line breaks probably won't be the same as the original, but with luck it will fit on one page.

@UlrikeFischer thanks for showing up in the GitLab thingy. I just took a very quick look at the code, most changes in that merge request were regex-based changes (everything with \newcommand for instance), all I did by hand was the change of the internals of \textipa and the fix for the \@ifstar using commands. And \textipa is defined using \newcommand in tipa as well (that I knew already from the question on TA).

@Skillmon yes I know, I only thought you would understand faster my point.

@barbarabeeton Thanks for the suggestion. The thing is that the identical text in a different one page document doesn't break. So I feel it shouldn't break here either. It's one of those times when I'm reminded of how little I know of what TeX is doing behind the scenes.

@barbarabeeton And yes, the text in question starts a new page in the document in question. Which isn't one page. To summarize Doc A is a one page document. Doc B > 1 page. One of the pages in Doc B has exactly the same content as Doc A.

But the page in Doc B breaks across two pages, whereas Doc A stays contained to one page.

I guess I could make a question out of it, but I was hoping it was sufficiently obvious to figure out myself.

260,000+ new cases in India either today or yesterday. And still rising, apparently.

Meanwhile our so-called PM is campaigning in West Bengal, and drawing huge crowds. I guess he wants more cases.

@UlrikeFischer still grateful that you took a look :) Thank you very much!

@FaheemMitha if you change the page parameters and the text doesn't fit, it doesn't fit ;-). I think that David Carlisle suggestions are ok --- you can use \enlargethispage and/or \adjustwidth with negative margins, or reduce the font size. There are not much more possibilities, and changing the page margin/textwidth/etc middle page is basically impossible. You can do it for a full page, but in my opinion it is still not optimal, documents with shifting page geometry are... ugly ;-).

@Rmano It seems that KOMA-script's letter and article classes use somewhat different page settings. Go figure. I was hoping to get some insight into how page breaking works, but for that probably I need to read the TexBook.

@FaheemMitha Well, I think it's logical --- I normally want a much narrower text size in a letter than in an article (unless two-columns). Anyway, I am sure I have seen a very nice article around here about how TeX breaks text into lines and paragraphs and pages but I'm unable to find it again...

@Rmano Well, I imagine it's all in the TeX Book anyway. They really should teach it in schools. It would be so much useful than the stuff they actually make you learn.

In summary, school sucks, TeX rules.

@FaheemMitha well, I wouldn't be so hard. You can understand TeX algorithm only if you know basic math and a bit of programming... ;-) (I'm a prof, in the end, understand I have to defend myself ;-P)

@Rmano School as in, where they send the children. Not college.

@FaheemMitha yes --- I know. I think that those professors have a much more difficult (and critically important) task than I have. And I still have fond memories of a handful of teachers in my basic years that really changed my life.

@Rmano Did they tell you about TeX? :-)

@FaheemMitha Well, in that you're right!

@Rmano What are you a professor of?

@FaheemMitha electronics (especially electronic instrumentation)

@Rmano Oh.

@Rmano After poking at this a bit more, it appears the real issue is simply that textheight set for the KOMA-script letter class is much much shorter than the corresponding article class.

592 pts for letter and 762 pts for article, at least if layout is to be believed.

This strikes me as discriminatory.

I'll test with an empty page, just to make sure I'm make sure I'm not messing anything up.

@FaheemMitha why would you expect them to be the same?

@FaheemMitha no: scrlttr2 has textheight 635pt article 598 so less

%\documentclass[a4paper]{scrlttr2}
\documentclass[a4paper]{article}
\showthe\textheight
\begin{document}

\end{document}

@PauloCereda youtube.com/watch?v=u0hp-alaSAU ^^

Is there any reasonable way to inject code into at the beginning and end of a footnote without redefining it, which other packages (e.g. biblatex) do too? It would be great if the kernel provided a boolean.

@AlanMunn I guess you mean the kernel providing a hook?

@StefanKottwitz Id be happy with just a boolean \infootnotetrue

@StefanKottwitz But yes, if the kernel provided a hook that would be good too.

@AlanMunn Kind of low-level, but \@makefntext defined by the class?

@AlanMunn no, but it is something we will need to for tagging. But footnotes are not trivial as lots of packages mess around here.

@UlrikeFischer Yes, that's the problem I'm trying to resolve so that I'm not also doing that.

@AlanMunn what do you want to do? Where do you need the hooks?

@UlrikeFischer In gb4e if you use the {exe} environment inside a footnote, it saves the current counter and starts a new one and then pops the saved value when the footnote ends. I'm trying to emulate this with enumitem and I think I could do this with its before and after hooks without redefining the footnote if I could tell I was inside one.

\documentclass{article}

\def\zz{\ifnum\currentgrouptype=11 FOOTNOTE\else not footnote\fi}%

\begin{document}

\section{aaa \zz}
aaaa \zz  \footnote{bbb \zz}
\end{document}

@AlanMunn not foolproof but perhaps could form the basis of a test ^^

@DavidCarlisle what is 11 for a type? (And where is a list of the numbers?)

@AlanMunn but fails for color, would need an aftergroup there, hmm

@UlrikeFischer insert (texdoc etex page 7) but with color (or latex-dev) the color groups are enabled so you have to get past a begingroup group first (14)

@UlrikeFischer Fix for that siunitx-v3.0.0-beta issue coming up

@UlrikeFischer I've gone for a different fix; I realised two things. First, taken from collcell, that I need a group to restore \cr. Second, and more complex, I need to worry about the fact that \protected@edef isn't always safe

@JosephWright my large tabular compiles again ...

@DavidCarlisle I don't think that will work, since by the time i'm in the environment I'm not in that group any more.

@AlanMunn yes I had in mind an aftergroup tower to work up the group levels to see if 11 was there, which would work, but by the time you have found the group type you have probably aftergrouped yourself away from whatever you need to do. However \insert might be a safe palce to hook as fewer packages mess with that. \let\realinsert\insert \def\insert{\afterassignment\infootnotetre\realinsert{` or some such

\documentclass{article}
\newif\ifinfootnote
\let\realinsert\insert \def\insert{\afterassignment\infootnotetrue\realinsert}

\def\zz{\ifinfootnote FOOTNOTE\else not footnote\fi}%
\usepackage{color}
\begin{document}

\section{aaa \zz}
aaaa \zz  \footnote{bbb \begin{bfseries}\zz\end{bfseries}  and \zz}
\end{document}

@AlanMunn plan b ^

@DavidCarlisle first try with @PhelypeOleinik's hooks seems to work fine. The main question is if someone redefines \@makefntext and so removes the hook again:

@UlrikeFischer yes but as you commented originally, everyone redefines that, but there couuld be an insert hook similar to the shipout ones but the cmd hook done at begin document on @makefntext is perhaps safe enough....?

@DavidCarlisle perhaps. One need to check a bit the usage. I'm not sure if \@makefntext is redefined so often, the mark is more of a problem.

@DavidCarlisle but beside this we probably should also add some hooks at a "safe" place.

I just got a mail from a french with "Conclusion: IL FAUT TOUJOURS BIEN LIRE LA DOCUMENTATION ! :-)"

@UlrikeFischer As long as the redefinition happens before \begin{document} it should be fine

> Founder of Adobe Dies at Age 81

@PabloGonzálezL ooh

@UlrikeFischer probably a mis-translation and they intended to have a negation in there

> Microsoft Edge Adds a 'Kids Mode' Limiting Browsing to 70 Web Sites

@JosephWright @DavidCarlisle ^^ should we add LL to this list? ^^

@PauloCereda no, the editor has an option for vim-like keybindings so it's not safe for minors.

@DavidCarlisle quack <3

@PauloCereda supper

@DavidCarlisle oh no

@UlrikeFischer Any faster than with v2?