@UlrikeFischer see my profile (me Karl Boris and Norbert so far:-)

@PhelypeOleinik good idea, I have other questions for that tag, eg

@DavidCarlisle I know :-) There's that one with two people also. It would be more popular than tikz-pgf ;-)

@PauloCereda You should tell something to Phelype

@egreg I hope @DavidCarlisle doesn't see that ;-)

@PhelypeOleinik \the sis

@DavidCarlisle If only it were in English, I could sneak that pun somewhere in the document

@DavidCarlisle -- I sent an inquiry to Karl and Boris. They jumped in and did something, but I haven't actually seen it yet.

@barbarabeeton Boris sent me an invite but there is nothing actually there, just the group set up:-)

@DavidCarlisle and @JosephWright -- Please take a look at this answer, my comment, and the poster's comment. I'm not competent to take on this argument, even though I know for sure that s/he's wrong. It needs a response from someone who's a LaTeX authority.

@barbarabeeton link?

@DavidCarlisle -- Oops! tex.stackexchange.com/a/517154

@barbarabeeton I left a couple of comments

@DavidCarlisle -- Thanks; that's good (and part of what I tried to say) But what about the recommendation to use unadulterated plain macros/primitives directly? That's the real danger there.

@barbarabeeton I could point to the comment thread from today under tex.stackexchange.com/q/517167/1090

@DavidCarlisle -- That certainly is relevant ...

@DavidCarlisle what is the problem with TU?

@UlrikeFischer the tests at the end of the format are going the wrong way so it's set up for OT1, not sure yet (@JosephWright)

@DavidCarlisle I'll take a look

@UlrikeFischer surely OT1 encoding has enough characters for anyone

@JosephWright it worked for me originally with the new format but i had a locally installed l3kernel when I deleted that and picked up the new l3bckend from texlive it failed

@DavidCarlisle Yes, I had tested things, there must be like you say another change somewhere

@DavidCarlisle This is what pre-testing is for

@DavidCarlisle what is with lualatex?

@DavidCarlisle, @UlrikeFischer It's not at the expl3 end: if I add an empty expl3.ltx to block any pre-loaded code, I get the same. Trying to track it down ...

@egreg ooh

@DavidCarlisle ooh

@PhelypeOleinik ooh

@UlrikeFischer lualatex-dev? that matches xelatex (I was just driving in)

@DavidCarlisle the curious thing is that texlive didn't update my lualatex-dev format, so it is over a week old and now fails too.

@UlrikeFischer Often if not always, building the formats fails when triggered by tlmgr update --all. Then, I run it manually.

@AlexG yes, that is not really a problem, in this case it is even good as it shows that the error was present earlier and I can wonder why it didn't show off.

@JosephWright @UlrikeFischer in fonttext.ltx

\show\Umathchar
\show\@undefined
\ifx\Umathchar\@undefined
\fontencoding{OT1}

> \Umathchar=undefined.
l.48 \show\Umathchar

how did that get undefined...

@JosephWright oops /usr/local/texlive/2019/texmf-dist/tex/latex-dev/base/fonttext.cfg

@DavidCarlisle and how did it get redefined? At the start of a latex document it is there.

@UlrikeFischer that the infomous config just used so that xelatex and lualatex use OT1 for the main test suite so that most \showbox show no diff...

three minutes on math in browsers

@DavidCarlisle it looks like a packaging error, if I install locally the fonttext.cfg is fine, l3build must have picked up the test version <--@JosephWright

@UlrikeFischer which is good as it means we can blame the packager?

@DavidCarlisle ooh

@DavidCarlisle six minutes would probably easier to understand ;-). Is there a good test page for mathmal support in browsers?

@DavidCarlisle ;-) we should probably get rid of this fonttext.cfg anyway. I looked at the other files in the support folder but I think nothing else can end up in a release too. All the others are "external" files.

@UlrikeFischer well we had the old test suite but that was several thousand pages you had to check by eye so even with some javascript support to let you say pass/fail and move to the next it was incredibly painful to run (I ran it several times:-) the web platform tests Brian mentioned have thousands of smaller unit tests and is fully automatic ...

... (basically the desired rendering specified in mathml or in exact css absolute positioning then the browser itself reports if the two versions give exact pixel equivalent rendering. But the version of chromium he was using is their internal build, we have a way to go to get that into the public chromium and then hopefully into chrome and edge

@DavidCarlisle it was more wondering if there is some page where I can look how good/bad mathml looks in my browser now?

@UlrikeFischer most new tests should not use it but historically it has been pretty useful, especially for luatex, as without that xetex and luatex would need specific tlg in far more cases and so we would probably never have spotted the dozens of luatex issues that we reported and got fixed where the behaviour of luatex was unintentionally different. Now i's more stable it's true that it mostly means we are not testing the public format (except when we make that format public..:-)

@UlrikeFischer yes let me dig out the old mozilla torture test

@UlrikeFischer developer.mozilla.org/en-US/docs/Mozilla/MathML_Project/…

@DavidCarlisle ooh comfy chair

@DavidCarlisle nice (or not so depending on the browser). The right-to-left stuff is interesting too.

@UlrikeFischer firefox has been able to render that page since it was called netscape, it is taking longer than we expected for other browsers to catch up,,,

@AkiraKakuto thanks for the mail:)

@DavidCarlisle there is probably not so much pressure, if everyone uses mathjax.

@DavidCarlisle really?!

@PauloCereda yes

@DavidCarlisle wow, that's impressive!

@UlrikeFischer yes that's the good and bad part of mathjax, it let the browser makers off the hook to some extent, but requiring a million lines of javascript to render a fraction is a failure of the web platform really and it still impacts on usability of math on the web: if you want interactivity you have to cope with the fact that when you click on something the elements being rendered are not a simple fraction but dozens of javascript elements inserted and positioned via css

@UlrikeFischer, @DavidCarlisle I see where the issue is. For format-building I have to copy the extra file. On the master branch, the last test fun leaves fonttext.cfg unchanged, so all works normally. But on develop, we have a typesetting building a format, and that was zapping the file. So I need to acount for that.

@DavidCarlisle yes, I quite agree that a native rendering is better. On other news: sourceforge.net/p/koma-script/activity/…

@UlrikeFischer let's say that isn't a universally shared view:-)

@UlrikeFischer oh:-)

@DavidCarlisle I got a mail, will forward.

@DavidCarlisle, @UlrikeFischer I have a fixed upload off to CTAN

@JosephWright thanks

@JosephWright now let's look how many people use the dev format on unicode engines ;-)

@UlrikeFischer ooh a poll

So, why is this return value non-zero? So, consider this Lua code.

path = "./trig2.tex"
assert(os.execute(string.format([[latexmk -pdflua -shell-escape %s &> %s.log]], path, path)) == 0)

I get the error:

lua5.3 foo.lua
lua5.3: foo.lua:3: assertion failed!
stack traceback:
        [C]: in function 'assert'
        foo.lua:3: in main chunk
        [C]: in ?
faheem@orwell:~/personal/business/smyt$ Latexmk: This is Latexmk, John Collins, 25 October 2018, version: 4.61.
Latexmk: All targets (trig2.pdf) are up-to-date

@FaheemMitha What's the exit code the command returns?

Does anyone know where this return value is coming from?

@PauloCereda The underlying command?

@FaheemMitha Yes.

And for me, os.execute returns a 3-tuple.

@PauloCereda Oh.

latexmk -pdflua -shell-escape trig2.tex &> trig2.tex.log

echo $?

0

So I assume the return value is 0.

Also, I would expect the TeX code running this to exit on assert failure, but it doesn't.

@UlrikeFischer I meant real people, you (and I) don't count.

@DavidCarlisle ;-(

@UlrikeFischer oh

@DavidCarlisle oh no

I guess the problem is that I'm doing:

assert(os.execute(string.format([[latexmk -pdflua -shell-escape %s &> %s.log]], path, path)) == 0)

Probably I should just have:

assert(os.execute(string.format([[latexmk -pdflua -shell-escape %s &> %s.log]], path, path)))

Not sure why I have the following construction. Probably someone provided it as example, and I blindly copied it. Cargo cult programming.

Is the latter version correct, then?

Since

print(os.execute(string.format([[latexmk -pdflua -shell-escape %s &> %s.log]], path, path)))

returns true, exit, 0

@FaheemMitha if os.execute returns a boolean value in the first place, you should change your equality check to a boolean check and not an integer one.

Or unpack the tuple and get the corresponding exit code, then do an integer check.

If two people on opposite sides of the world simultaneously drop a slice of bread, Earth briefly becomes a sandwich.

@PauloCereda I see I was copying example code from Joseph.

I.e.

I'm not sure exactly what that discussion was about. I guess I need to review. Sigh.

@PauloCereda In any case, according to the documentation for assert, it checks for the first value being true. So explicitly checking for the value seems redundant.

@FaheemMitha agreed. :)

Not sure what Joseph had in mind, but it's possible I was misunderstanding him.

I tried os.execute with a non-existent command, and the first argument returns nil.

Not false, as one might expect.

I get weird errors with this code:

documentclass[12pt]{article}

\begin{document}
\directlua{
  path = "./trig2.tex"
  assert(os.execute(string.format([[latexmk -pdflua -shell-escape %s &> %s.log]], path, path)) == 0)
}
\end{document}

I assume it's unhappy with the % signs.

Is there any way to fix it, or do I need to put the Lua code in a separate file?

I got rid of the %s, so the code runs. But the assert gets lost.

So, if one runs

\documentclass[12pt]{article}

\begin{document}
\directlua{
  assert(os.execute(string.format([[latexmk -pdflua -shell-escape trig2.tex &> trig2.tex.log]])) == 0)
}
\end{document}

then the assert failure doesn't show. But if one runs the Lua code directly, one sees the assert failure. What's with that?

@FaheemMitha Use the luacode package, perhaps?

Also, it hangs, and doesn't return.

Funny, if I instead do:

\directlua{
  assert(os.execute(string.format([[ foobar ]])) == 0)
}

then the assert error does show up in the LaTeX compilation. Weird.

Maybe I'll write a question.

@HaraldHanche-Olsen I guess. But that doesn't seem to get very good press. Apparently sort of a bandaid, and not very reliable. Do you use it?

The big news for the pre-release of LaTeX2e 2020-02-01 is pre-loading expl3. That means that things like fontspec load a lot faster: very noticeable for XeLaTeX and LuaLaTex users. #TeXLaTeX twitter.com/texdevnet/status/1197127659001651201

@FaheemMitha It's more that it hides some stuff that is best handled by loading Lua code in a separate file

@yo' ooh

"Telling a programmer there's already a library to do X is like telling a songwriter there's already a song about love." - Pete Cordell

Thanks @JosephWright for the big news!

@yo' No problem

What version of Lua is LuaTeX 1.10 using, again?

@FaheemMitha 5.3

@AndréC try texdoc -l amsldoc to get all the results in a list.

@AndréC deeply sorry then, I am out of ideas... :(

@PauloCereda I think I should report this to Christian Shenk:-(

@PauloCereda I just reported this error to Christian Shenk: github.com/MiKTeX/miktex-packaging/issues/144

@JosephWright Ok. Thank you.

@AndréC CTAN has two extra versions that are probably not up to date (judging from the dates inside them): Italian and vietnamese.

Hi everybody, I just cannot remember the name of the xetex counterpart of the command line utility dvitype. I want to look into a *.xdv file.

Found it. It's dviasm that can be used here.

@FaheemMitha I use luacode only rarely, mostly when I have too little code to want to bother with an external file, yet with more special characters than I feel up to dealing with.

@daleif I downloaded the English version of the CTAN, on MikTeX, there are only the Italian version available!

@yo' ooh, is that the “secret plan” people have been talking about?

@AndréC that might depend on how you installed your miktex.

@HaraldHanche-Olsen This was just for demonstration purposes.

@FaheemMitha Then using luacode seems to me a good choice.

@HaraldHanche-Olsen That's the one

@AndréC try texdoc amsmath, then you get a web page with lots of options, including the english docu.

@UlrikeFischer This is what I did first before doing as indicated by Paulo Cereda

@AndréC that's not what you are saying above. There you write texdoc amsldoc and not texdoc amsmath.

@HaraldHanche-Olsen ooh spoilers

@AndréC did you see that the name is different?? amsldoc gives the italian version for me too, but with amsmath I get the standard web page.

@UlrikeFischer On CTAN, the link to the amsmath documentation is amsldoc.pdf: ctan.org/pkg/amsmath

@AndréC yes but the miktex documentation finder (mthelp (texdoc is an alias of mthelp on miktex) goes by package, it looks for an html page.

@UlrikeFischer I don't like the typography of html pages, I always read the documentation in pdf.

@AndréC the html page a link collection to the various pdf. Why don't you try it??

@UlrikeFischer Because my browser reads pdf pages and I didn't bother to set this browser to prevent it. I'm going to be forced to do it. Argh :-(

@AndréC forse è perché lo sostengo e il mio italiano è molto meglio di quello di @greg

@DavidCarlisle Il est dommage que vous ne parliez pas français mieux que Bernard. Auquel cas, je n'aurai rien dit...

@AndréC My French ought to be better than my Italian (since I studied it at school) but....

@AndréC well yes, that is why I normally let texlive search a documentation - I don't like it either if a pdf is opened in a browser ;-)

@UlrikeFischer Great, do you know where I can find the doc of amsmath in French?

@AndréC I don't think there is one, but you could ask the french group.

@DavidCarlisle I certainly wouldn't use “molto meglio”, which is grammatically incorrect in that context. :-P

@UlrikeFischer I was joking in the slightest hope that David would translate this doc into French.

@DavidCarlisle And “sostenere” is not the correct translation for “to support” in that context.

@DavidCarlisle “Parlo inglese molto meglio di David Carlisle” is correct as well as “il mio inglese è molto migliore di quello di David Carlisle”.

@DavidCarlisle “Meglio” is an adverb, “migliore” is an adjective.

If I want to do a redirect to a file from an os.execute command in Lua, I was wondering if I should use something else instead. Perhaps os.popen? E.g.

os.execute(string.format([[latexmk -pdflua -shell-escape x.tex &> x.tex.log]])

And this redirect doesn't actually work. Does anyone know why?

Hi everyone again

Which environment is best suited to write a subtraction of hours like this?

Not yet, I'm connected from my phone ...

@PabloGonzálezL ok, no pb.

Here is what I get with Tikz's matrices:

@AndréC I'd go with array and s columns from siunitx

@egreg : insignificant details

@AndréC looks like a standard tabular, unless you want to ask David for a special tabulartime package ;-).

@UlrikeFischer sounds great, let's ask David for a new package :)

@UlrikeFischer Oh yes, a special tabulartime that would allow you to simply write the duration calculations (addition and subtraction would be a real pleasure). I used TikZ so I could add arrows. I don't know if this will be clear to the students. I'll show them tomorrow, they'll tell me. What do you think about that?

:52631925

\documentclass{article}

\begin{document}

\def\zz#1#2#3#4{%
\begin{tabular}{rrr}
&#1h&#2mn\\
$-$&#3&#4mn\\[2pt]
\hline
&\the\numexpr#1-#3\ifnum#2<#4 -1\fi\relax h&
\the\numexpr#2 \ifnum#2<#4 +60\fi - #4\relax mn
\end{tabular}}


\zz{7}{50}{4}{15}


\end{document}

@DavidCarlisle there is the hour missing in the second line.

@UlrikeFischer exercise for the reader

@DavidCarlisle, @UlrikeFischer Had to do PR2 ...

@DavidCarlisle in the french usenet group someone was searching for a package and described it as "En gros ça se présente comme un code source complètement abscond, mais quand on le compile ça sort une comptine en anglais.".

@JosephWright I saw it. But we should really change the date ...

@UlrikeFischer You fancy the palindrome?

@UlrikeFischer :)

@JosephWright yes ;-).

@UlrikeFischer why would write such a thing?

I just finished my exercise :-)

@JosephWright been out since I got back (something at M's school but I need to tweak test2e a bit more 265 test failures for my first guess (actually I'm confused about some of them, need to check what they did before....)

@DavidCarlisle Cool there is no rush

@AndréC ohh fancy spacing as well as an h:-)

@DavidCarlisle I need to work out how to fix the deploy keys ...

@JosephWright did removing the quotes not work?

@DavidCarlisle Nope

@JosephWright I could get the travis cli tool to generate the yml fragment rather than just the encrypted key (or at least I could if @UlrikeFischer reminded me of the incantation:-)

@DavidCarlisle something like that, yes ...

@DavidCarlisle travis setup releases

@UlrikeFischer Ok I'll look at the manual, need to tell it to use the machine user key presumably

@DavidCarlisle probably, I didn't have one when I tried this.

@DavidCarlisle Is it possible for you to make it an answer to my question, even if this code is not completely finalized (it is a good starting point for me)?

Quack

@PauloCereda Quack

@AndréC was there a question? OK I might do later or feel free to self answer

@JosephWright ooh a duck

Yes, here, an answer from you will be very useful.

@JosephWright machine user hasn't got access latex3 repo gives admin rights to l3 team (that the machine user is in) but the latex2e repo doesn't as far as i can see, I'll make them the same (thetravis ci tool wouldn't make take the machine user token for setup releases)

@DavidCarlisle I'm going to try force-pushing the tag, as this is purely a deployment change

@DavidCarlisle Working :)

@DavidCarlisle Right, shouldn't have to build LaTeX2e again on my own machine :)

@JosephWright OK good, so do same on master?

@DavidCarlisle Plan is to use this for master from the 2020-02-01 release: we've already seen that the deployment itself is non-trivial to get right, so I'd rather not try to add it as a 'hotfix' to master

@JosephWright sure I was thinking that was the next release from master, but I suppose there may be more PL releases.

@DavidCarlisle Quite: we should stick to the idea that we only move major stuff after it's been tested in PRs

@JosephWright I'm too tired for this tonight:-) In the existing setup for testfiles-TU how come TU fonts get used, not the special format with OT1 fonts? that is why is /home/latex2e/base/testfiles-TU/tu-tlb4500.tlg showing TU/lmr

@DavidCarlisle if I understood Joseph description today, then there is a second round of "- We unpack files - We run the TU tests without fonttext.cfg change"

@UlrikeFischer yes but what stops it being used (given that it is in the support directory) but I've given up for the day

@UlrikeFischer @JosephWright oh I see fonttext.cfg isn't handled via the usual suppfiles variable but by

  if dest ~= typesetdir and
    (not options["config"] or options["config"][1] ~= "config-TU") then
    cp("fonttext.cfg",supportdir,unpackdir)
  end