@DavidCarlisle Nice use of \text_purify:n :)

@JosephWright see my "2020" added note?

@DavidCarlisle :)

@DavidCarlisle Yes, I was thinking about an answer just before you posted your one

@JosephWright would you have used purify as well or something different?

@DavidCarlisle I'd do the same as you: it's relatively slow, as a comment mentioned, but at the moment it's the best plan - if we need more performance in this area, we could provide a version that doesn't do \text_expand:n first

@PauloCereda enjoying breakfast the votes?

@DavidCarlisle why "by accident"? They are the same command aren't they?

@UlrikeFischer well if it was "by design" making the inputenc characters robust was a "breaking change" so it's just a choice of words.

@DavidCarlisle ;-) good politics

@UlrikeFischer it is more or less accidental that it expanded without error to something unusable for anything but that test

@UlrikeFischer like describing reversing your tax policy overnight as "listening" not "a U-turn" ?

@DavidCarlisle :)

@DavidCarlisle Oh, newspeak. Is it your new leaders?

@mickep bbc.co.uk/news/live/…

@UlrikeFischer The traditional position is one shouldn't \edef user input, so the fact this doesn't explode is a happy accident

@JosephWright well yes, but then it would have exploded in both parts in the same way ;-)

@Later ?

You see the same result?

@Later I get

I use TexPad on iPad, so that is a bug of the app?

Ok. Thank you.

@UlrikeFischer :)

@DavidCarlisle what a mess Brazil has become :)

@DavidCarlisle Ooh no :(

Friends, what's the default --interaction, is it errorstopmode?

@PauloCereda yes, all other don't stop in some cases.

@PauloCereda hard to tell if you have no errors

@UlrikeFischer Thanks :)

@DavidCarlisle ooh

@UlrikeFischer but last night's dinner would prove that test invalid

@DavidCarlisle oh no

\nodinnerfortheduck

@UlrikeFischer just 34441 to go:-)

@DavidCarlisle the main problem will be how to get the 1 ;-) What should we do with this \mag business? (this old discussion pull request looks complicated ;-))

@UlrikeFischer a lucky downvote (or in this case two downvotes) helps (see meta thread)

@DavidCarlisle you mean I should give a few answer for another format?

@UlrikeFischer or just answer vim is evil, and @Skillmon will downvote you

@UlrikeFischer (@JosephWright) on \mag probably need to get l3backend to follow graphics-def unless we want to learn enough Japanese to be confident that we can provide compatible behavour.

@DavidCarlisle I see it a bit differently ...

@JosephWright I'd think \let\mag\undefined would be fine for any supported documents, so what happens when it is used is not so much of a concern. We could try to really have supported use of \mag but I'm not sure I see any gain really.

@DavidCarlisle well if we change something we should change hyperref and graphics-def similarly. (Actually I now realize that we aren't completly in sync here. hyperref doesn't test for mag either.) But wouldn't it be better if their classes would set \stockwidth correctly?

@UlrikeFischer the trouble is we haven't defined correct. if mag is used should stockwidth be using true units or normal units, the answer depends on the driver interpretation of papersize special.

@DavidCarlisle sounds as if testing for mag is the easy way out

@UlrikeFischer that's the conclusion I came to years ago for graphics:-) It's a bit of an admission of failure, we could try again to get some consistent behaviour (there are fewer dvi drivers now) but...

@UlrikeFischer for plain it does

\def\magnification{\afterassignment\m@g\count@}
\def\m@g{\mag\count@
  \hsize6.5truein\vsize8.9truein\dimen\footins8truein}

so the media size is fixed (as it was physical paper) and \magnification just affects the font sizes mostly. But these days you may also want to simply scale the entire document to change its default size so use \mag in a way that does affect the media box. My recollection is the ptex classes do both depending on .. something

@DavidCarlisle then I think I should adapt hyperref to test for \mag too. And we should get geometry in line ;-).

What room can't a skeleton enter? The living room.

@UlrikeFischer @JosephWright basically we'd need to answer this github.com/latex3/graphics-def/pull/1#issuecomment-231579159

@DavidCarlisle I saw that, and after reading it I liked the plan to back out ;-)

I may never see another palindrome, but the number of questions on the main site should hit a nice one in a few hours: It's at 239,923 now.

@barbarabeeton That was a bit technical and about dvi, but as far as I understand everything is very limited. It is impressing how well working TeX could be with all the limitations.

@mickep -- Yes. At the time, nothing else came close. Moore's Law has changed the landscape. I didn't think earlier of HarfBuzz, about which I know very little other than it addresses font shaping, but it may be able to address the matter of character fitting in math.

The issue with the overleaf editor is now solved. As suggested by @DavidCarlisle, I contacted overleaf support. They asked me to join the beta program. They said "We're currently beta-testing a new version of the editor, and one thing we aim at improving is better handling of non-Latin scripts and unicode characters."

@barbarabeeton I know (almost) nothing about HarfBuzz, but can you explain what you mean by character fitting? Do you mean control by glyph rather than by bounding box? Just so we talk about the same thing? If so, I would be happy to see HarfBuzz in action doing this.

@PNDas yeh...

@PauloCereda vvv

There are three options "source", "source(legacy)", "Rich text". In the "source", there is no issue but in the last two the problem is still there. From now on I'll work with "source".

Thanks again.

@PNDas sure the "legacy" one is the old ace editor that really is hopeless for non latin scripts. They can't really fix that as it is only there to be the same as the old editor....

@mickep -- Yes, I do mean control by glyph shape (perhaps as well as by bounding box). You have now exhausted my knowledge of how HarfBuzz works. Something more I will have to dig into.

@PNDas (@JosephWright) I really should switch the editor at learnlatex....

@DavidCarlisle Sounds good to me

@PauloCereda -- The news isn't encouraging. Keep safe.

@barbarabeeton That's not really what HarfBuzz does though. Normally HB just uses the font layout tables and doesn't look at the actual glyph outlines. There is some support to actually parse the glyphs in recent versions, but IIRC even that doesn't do deeper analysis but just passes along the drawing instructions. (I would be happy to be proven wrong though)

@JosephWright At the tme I got codemirror 5 to work but cm6 was in beta and almost completely different and I failed to integrate it, but think cm6 is now released and documented, I should try again

@MarcelKrüger I imagine that looking at the glyphs could be slow.

@mickep who wants to look at glyphs anyway? \tracingall logs are the display technology of choice.

@DavidCarlisle How are you related to learnlatex ?

@DavidCarlisle Yes, but that is a secret.

@PNDas I wrote the code to integrate an editor and run latex

@MarcelKrüger -- Thanks for that information. But do you think it might be worth someone who really knows how it works (and has worked with it) looking into the math situation? If so, @mickep has pointed out some good starting examples, and I can probably come up with some more.

@barbarabeeton this? frederic-wang.fr/opentype-math-in-harfbuzz.html

@DavidCarlisle Ohh, interesting!

@DavidCarlisle psk, unlike Emacs, there is no evil in VIM.

@mickep Fred is the lead developer on mathml-in-chrome

@barbarabeeton Beside the performance issues mentioned by @mickep I guess that a lot of this would end up needing more subjective judgement than an automated solution can provide. Even if you have access to glyph shapes, how to you determine spacing from that? Minimal distance at equal heights? What if both characters have elements at similar but slightly different heights? Should that affect the spacing?

I suspect that it might be more helpful to add regular kerning also between math characters to allow font designers to specify directly between which characters the spacing should be adapted instead of leaving this to automatic determination.

@DavidCarlisle -- Definitely interesting, but it seems to deal mainly (only?) with extensible and composite glyphs, rather than with "adjacent" glyphs. (By the way, the "spell checker" for this page doesn't think that "glyphs" is a real word. But it doesn't complain about "glyph" singular.)

@MarcelKrüger Oh, I thought that was supported in the fonts, for math characters?

@MarcelKrüger -- I'm not competent to say how to address the situation, but I can say for sure that the way "A" and "W" behave as a subscript to the same base character is certainly quite noticeably different, and rather distracting to even a novice non-typographer.

@barbarabeeton For subscripts OpenType math has staircase kerns which address exactly this without having to actually look at glyph outlines.

Of course if a specific font actually uses them (correctly) is a different question...

Good evening everyone, I'm writing a report in a book format \documentclass[12pt,a4paper,oneside]{book}, I would like to pause, stretching the lines to fit the page (i.e. tentatively). Is this feasible? See the following picture

Thank you for your invaluable time .

Thank you @DavidCarlisle for also asking for a minimal example...

@MarcelKrüger -- So I need to study the OpenType features as well. I do think staircase kerns could be helpful in that situation.

@MarcelKrüger They are not so much used in many of the fonts, as far as I remember from looking at the fonts lately.

(Lucida and Cambria has it, perhaps also garamond-math, but I don't remember how it is for the gyre fonts for example)

@CroCo pause? do you mean like beamer \pause ?

@DavidCarlisle I would guess it is about \flushbottom and whatever the other one is called.

@mickep I thought it was about \pagebreak vs \newpage. Btw. the other one is \raggedbottom.

@UlrikeFischer I tagged a luaotfload release and plan to upload it if nothing unexpected comes up soon.

@MarcelKrüger good.

@MarcelKrüger :)

Also firstaid-dev released :)

@JosephWright Any opinion on merging github.com/latex3/latex2e/pull/909?

@MarcelKrüger Looks good to me

@JosephWright I think I need to remove local ctanupload = ctanupload or "ask" from l3build-upload as otherwise this local one masks the global one set in build.lua/l3build-variables

@DavidCarlisle Ah, yes

@JosephWright I was talking nonsense the other day about my form string branch that was merged feb time.

@JosephWright I can do it, as I'm there?

@DavidCarlisle Please

Quick bash question: I want to use sed -i '.bak' '$d' filename.tex in a bash script. How do I escape things so it works?

@AlanMunn oh sorry $d is sed not a bash variable:-)

@DavidCarlisle Right, that's the issue. How do I get that $ to be interpreted by sed not bash?

@AlanMunn \$ I guess, or \\$ or \\\\$ until it works

@DavidCarlisle That doesn't seem to help.

sed -i.bak -e "\$d" filename.tex seems to work

@AlanMunn file gets one line shorter each time

@AlanMunn or sed -i.bak -e '$d' filename.tex is same

@DavidCarlisle Are you trying these within a script or just from the command line?

@AlanMunn a bash command line, but could try a script?

@DavidCarlisle My problem is when I use what works in the command line in a script it fails. That's the issue

@DavidCarlisle The command I originally posted works fine on the command line

#!/bin/bash
sed -i.bak -e '$d' filename.tex

I saved that as amsed then chmod a+x amsed; ./amsed did the right thing

@AlanMunn do you specify bash with #! ?

@DavidCarlisle Yes. My actual line is sed -i.bak -e '$d' $oname where $oname is a constructed filename.

@AlanMunn what error?

@DavidCarlisle It just hangs (it's a TeXShop engine file, which is just a bash script run by TeXShop). So maybe it's something else entirely. But it was working until I added the sed line...

@AlanMunn head -n -1 filename > newfile ?

@DavidCarlisle Doesn't work on the Mac...

@AlanMunn install windows on the mac and cygwin on windows then you'll be fine

@DavidCarlisle :D

@DavidCarlisle Doh. The filename I was testing on had a typo, and so the output file that I was looking at was not the one being changed. So the command worked, but I was looking in the wrong place for the output...

@AlanMunn ah

@DavidCarlisle At least I'm saved from having to install windows.

@AlanMunn keep that as a backup plan

@DavidCarlisle I had to delete my Windows container (which I never used) when I updated the OS this summer, so I doubt it will ever return. :)

@DavidCarlisle Thanks for the help though.

@AlanMunn I suspect the famous rubber duck would have been as useful

Any TikZ people who have a more automatic way of doing the placement here? @CarLaTeX @Rmano

@AlanMunn too late now, I'll look at it tomorrow. Maybe some h/voffset interaction?

@AlanMunn Too late for me, too. I'll look at it tomorrow if @Rmano doesn't arrive first.