@barbarabeeton The command \endgrag in the .ltx you sent to me gives Undefined control sequence, what should I add to compile without errors? Thank you!

@CarLaTeX I guess \endgraf without knowing anything about context...

@mickep You're right, with \endgraf it works, since it's put at the end a footnote, it is probably that one, thank you!

do we know anyone who knows about thesis requirements for this place? \instituicao{Escola de Engenharia de S\~ao Carlos, Universidade de S\~ao Paulo}

@DavidCarlisle @PauloCereda?

@DavidCarlisle I know the place and some rules. :)

@PauloCereda but have you ever seen a draft thesis?

@DavidCarlisle They are our competitors actually, as I am in the Engineering School in the main campus of USP. :)

@DavidCarlisle No idea. Do you have the link for the question?

@PauloCereda it is latest question :

@DavidCarlisle Hm judging by the document name, it's not a Masters/PhD thesis, but a bachelors.

@DavidCarlisle: ah they are using abntex2.

@PauloCereda no ducks, so clearly a failed thesis

@DavidCarlisle exactly!

How can one check, whether a specific key was already cited in biblatex (outside of biblatex's cite commands)? Something like \ifciteseen but with the ability to check a specific key.

@PauloCereda This chapter design is fantastic!

@Skillmon Can you give an example how the key was cited before?

@barbarabeeton on the proper centering of commas:

Oh, two palindromes, cheers!

:39779826 \documentclass{article}
\usepackage[utf8]{inputenc}
\usepackage[citetracker=true]{biblatex}
\addbibresource{biblatex-examples.bib}

\begin{document}


\makeatletter
\blx@ifentryseen@global{doody}{yes}{no}

\cite{doody}

\blx@ifentryseen@global{doody}{yes}{no}
    \end{document}

@DavidCarlisle Congratulation to the palindrome!

^^^ @Skillmon

@samcarter 404,404 would have been better, but someone unkindly accepted one of my answers when I was 10 short of that:-)

@DavidCarlisle Outrageous!

@samcarter <3

@DavidCarlisle Congrats!

@mickep yes next challenge is to get 6 silver and 3 bronze badges and 1,100 points to get all numbers pallindrome:-)

@DavidCarlisle That would indeed be impressing.

@mickep I don't suppose "it works for me, so I don't care" is suitable to be posted as an answer?

@DavidCarlisle:

mickep@hurf:~$ mf
This is METAFONT, Version 2.7182818 (TeX Live 2016/Debian) (preloaded base=mf)
**\relax

*drawdot(35,70);

*showit;

*bye
Transcript written on mfput.log.
mickep@hurf:~$ mf
This is METAFONT, Version 2.7182818 (TeX Live 2016/Debian) (preloaded base=mf)
**\relax

*drawdot(35,70);

*showit;

*bye
Transcript written on mfput.log.

The first run showed the dot, the second did not...

@UlrikeFischer Thank you very much! In the documentation I only found \ifentryseen and the internet wasn't kind to me either.

@DavidCarlisle I saw your comment. Maybe something is fixed in TeX live 2017, or just broken in ubuntu.

@mickep everyone should use windows (@egreg will confirm)

@DavidCarlisle Yes! Let us get into that "war"! :D I think I leave this for now, I just cannot understand why it works sometimes and not others. I somehow get the feeling that Knuth is joking with us...

@DavidCarlisle no

@samcarter Ulrike answer works for me.

@Skillmon But be aware that it uses internal commands, so there is no garanty that the command names doesn't change in the feature ...

@mickep I would guess it's not Knuth's code but some buffer not being flushed in the X display code of the system specific display parts of the mf executable?

@DavidCarlisle That is probably correct (even though I don't understand exactly what you say).

@UlrikeFischer Maybe I should add a feature request. I thought I'm searching for the wrong keywords in the documentation, because I was sure that this is an already implemented macro (on the user side)

A nice article on the implementation details of how the TeX program internally stores tokens: overleaf.com/blog/522-what-is-a-tex-token

@ShreevatsaR Hmm, but really shouldn't need this detail to use any well-designed LaTeX code (Overleaf after all is LaTeX not plain): even for plain one doesn't have to mess with the non-accessible hash table code, etc.

Yes no one should need this detail if they only want to be a user :-) But I thought some may be interested here in the internals of the program

or just as an insight into how "tight" programs were written in the 70s

@DavidCarlisle Mass just ended, I can't reply how I should

Probably someone who would like to read the TeX program (texdoc tex or the raw tex.web) would benefit from the pictures…

@ShreevatsaR Oh sure, wasn't a criticism from that POV

@JosephWright How would you translate \ifx\relax#1\relax to expl3? \tl_if_empty:nTF{#1} is probably okay in my case, but it doesn't give the same result if #1 is \relax.

@UlrikeFischer Depends on what the meaning is ;)

@UlrikeFischer You either use a low-level meaning test, or you do a higher level test but that means deciding on whether you want to know 'empty' or 'defined'

@egreg can't you just ask for forgiveness next week? (my understanding of these things is rather slight)

@TeXnician chat.stackexchange.com/transcript/message/39780956#39780956

@DavidCarlisle That's rich, coming from an X window user on Windows.

@TeXnician :)

@TeXnician well you need a reasonable window system these days.

Does anyone know about the status of development of unicode-math? The github was updated just some days ago, but there are a lot of "issues".

@mickep Will fixes stuff: not sure what more can be said

@barbarabeeton I read in your new AMS style guide that you plan to change to the stix fonts in the future. I assume that this means stix two? Do you plan to use unicode-math?

@JosephWright I see.

@mickep unicode-math is practically speaking the only way to use the font from latex currently

@DavidCarlisle 'Do it all yourself'?

@mickep actually more of an issue than unicode-math is the state of the engines, both luatex and xetex have had relatively large issues with opentype math font setting (which is in its infancy and rather under-specified) so the underlying engine support is not very stable between releases.

@DavidCarlisle OK, the set of people who might try that is very small

@DavidCarlisle Well, the only way to get something worse than Windows' default is trying to compile Wayland with experimental DirectX bindings (one had to develop first) using Cygwin, so X seems reasonable ;)

@DavidCarlisle Oh, I see. But that sounds like good news to me. Correct me if I'm wrong, but at least luatex seems to be developed quickly.

@mickep yes although it's not uncommon that it's one step forward and two back

@DavidCarlisle :)

@DavidCarlisle Oh, is that so :/

@TeXnician I saw your comment on the Metafont question, did you also get the "random success" using linux, or did it fail all the time?

@mickep actually that's a bit unfair, but compared with the engine stability we are used to with tex, it's relatively unstable (and intentionally so) and that causes some headaches for publishers like the ams who need a stable workflow.

@mickep Actually I got random success (Arch Linux, TL17). I don't have the machine on now, but I'll test it again in the very near future.

@DavidCarlisle I'd imagine that only the AMS want source stability in this way: for everyone else, the PDFs will be archival so they just stick to a fixed binary for a production run

@JosephWright yes but (for example) it was a pain while testing the stix fonts as the number of people who could tell if disappearing delimiters and accents was an issue with the fonts or with the tex engine or with the macros was rather limited. (and you do the same tests 6 months later and get different results....)

Btw, this is how the binomial coefficients set in latin modern look like in current ConTeXt:

@DavidCarlisle or the fontloader version ...

@UlrikeFischer yes:-)

But, in next version one will have some options to set to make them look better :)

@mickep I re-implemented the ams binomial coefficient macros for amsmath when used for lua or xetex last year, so hopefully the latex ones look OK:-)

@DavidCarlisle :)

@mickep @DavidCarlisle There was a few days ago a question on the french list with binom and the xits-fonts where the output was quite horrible (but the source of the problem are imho wrong math contants).

@DavidCarlisle I suppose so. I suspect this is a side-effect of the "fences" that were implemented some time ago. I think the problem is that Hans is using cambria (THE opentype math font) for testing.

@UlrikeFischer I asked on the ConTeXt list, do you remember if the other person used ConTeXt or latex? I have never seen latex fail on these things (but then I have not really used unicode-math/open type fonts with latex yet).

@mickep @UlrikeFischer that's part of the issue, the documentation for the opentype math font parameters from Microsoft is sketchy at best, and they don't have a natural mapping to classic tex parameters so font designers (and typestetting engine maintainers) have to try to reverse engineer what cambria/Word are doing and then creatively interpret that in a tex setting

@mickep @DavidCarlisle \documentclass{article}
\usepackage{amsmath,unicode-math}

\setmathfont{XITS Math}

\begin{document}

Texte : $\binom{n}{k}$
\end{document}

@UlrikeFischer oh great.

@DavidCarlisle yes, looks good ;-). But as I said imho the font has a too large textstyle ` \Umathstackdenomdown` value.

@DavidCarlisle That is a bit sad.

@UlrikeFischer, @DavidCarlisle The equivalent result from latest ConTeXt...

(except the french spacing before the colon) :D

@UlrikeFischer unicode-math probably needs (if it hasn't already) a "fix parameters in known fonts" configuration block...

@DavidCarlisle Yes, imho context has configuration files for fonts, where you can add needed patches. It is not very difficult to add them but it would be good to have a more official interface. (With xelatex the output is not so problematic, perhaps harfbuzz has some correction).

@UlrikeFischer FWIW:

\documentclass{article}
\usepackage{amsmath,unicode-math}
\setmainfont{STIX Two Text}
\setmathfont{STIX Two Math}
\begin{document}
Texte : $\binom{n}{k}$
\end{document}

@HaraldHanche-Olsen Indeed, the stix two is so much better than the first version/xits. Also, the text font is much nicer in my opinion.

@mickep well they did spend 10 years on the update, you'd hope they make something better,

@DavidCarlisle You got a point.

I am currently putting together a conference proceedings volume. Using STIX 2, and after hitting a few snags (and getting some great help from @UlrikeFischer), I'd say the project is going quite well. I am happy with it.

@HaraldHanche-Olsen This is the output ConTeXt gives with stix two. There is a difference in the spacing to the left of the second parenthesis.

@mickep Indeed. The context one is better. Thanks for pointing that out.

@mickep @HaraldHanche-Olsen hmm I'd better check that...

@DavidCarlisle the same extra space seems to be present in the xits examples above.

Maybe someone removed a % too much somewhere ;)

@mickep :)

@mickep Yeah, it's just harder to notice, given the other horrors in that one.

@mickep no it's \nulldelimiterspace=0pt (or rather, it isn't that...) There are some corrections for \nulldelimiterspace but set it to 10pt and the left ( stays where it is and the right ) goes off on its own...

@DavidCarlisle Nice! Quickly found.

@mickep not so quickly fixed:-)

@DavidCarlisle The evening is saved :D

@DavidCarlisle Why does it happen only with unicode math?

@UlrikeFischer because I messed up the luatex-specific definition of genfrac, I would guess

@DavidCarlisle (Is this one of the two steps backwards?) hide

@DavidCarlisle perhaps it is lualatex-math faults, it patches genfrac.

@UlrikeFischer @mickep @HaraldHanche-Olsen and today's question is: how much correction for nulldelimiterspace gets applied on closing c delimiters:

\nulldelimiterspace\z@
\ifx c#1\kern-\nulldelimiterspace\fi

@barbarabeeton you might want to follow the above conversation....

@DavidCarlisle ?

@JosephWright amsmath.dtx lines 682-683

@JosephWright and dont run svn blame on that file

@DavidCarlisle :)

@DavidCarlisle I wondered about this already ...

@UlrikeFischer does the above definition look better?

@DavidCarlisle The 10pt version looks good to me :)

@mickep I could flag that as abusive comment :(

@DavidCarlisle Please do! But to be serious, the suggested version looks good!

@CarLaTeX -- oh, dear, bad typing. not \endgrag, but \endgraf ("end paragraph"). sorry. (i see that @mickep has already come up with the correct spelling. i've spent much of my day trimming a hedge. a large hedge -- about 6 meters long and 2 meters high. that's why i haven't been here to respond to questions.)

@DavidCarlisle -- showoff.

@DavidCarlisle make much mor sense ...

@barbarabeeton that was before \binom spoiled my day :(

@DavidCarlisle I'm sorry it spoiled your day. :(

@mickep -- re stix, yes, stix two would be a much better choice than stix one. unfortunately, we still have to figure out a way to do automatic checking of individual graphics, and until we can do that, a transition is on hold. also, xetex is the only possibility; luatex explicitly does not promise reproducibility, and that is essential in the ams production environment.

@barbarabeeton OK! I saw in the web page that elsevier, ams, IEEE, ... are backing up stix, so I guess we will meet it much more in the future.

@JosephWright -- source stability is not enough. we need the resulting output to be identical at least 6 months to several years later, since the production cycle for some books is quite long.

@barbarabeeton I'm just working on the final version. I'll send to you soon, and if you give me your blessing I'll send it to the official address :):):)

@DavidCarlisle -- am following. thanks. sigh. (but am leaving in ten minutes, so will be awol for the "answer the unanswered. i'm stiff and sore from too much hedge trimming.)

@barbarabeeton Sure: I doubt any publishers update very often. But that's only dependent on the final production machines for output PDFs. Other than maths I can't imagine many publishers wanting to be able to reprocess without re-setting for different editions or the like.

@JosephWright -- it's the production cycle that is our hangup. if a book takes three years to complete, the copyediting is often done early on, and author adjustments come in slowly. you don't want to do the full copyediting more than once, only the changes. and since math copyoediting involves checking in-line math for overfull lines, any unanticipated change means starting over. too costly. absolute reproducibility is mandatory.

@barbarabeeton I spent my day touring in Tuscany, ending it with a gorgeous fiorentina (in the sense of the steak).

@barbarabeeton Again, I suspect maths-specific: in my area the copy-editing is last, at the galley stage

@egreg -- that's a nice way to spend a day. glad you enjoyed it.

@JosephWright -- probably true about math-specific. but as soon as we get a manuscript, it goes directly to "galleys". (editor-mandated author changes have already been applied. later changes are from copyediting questions.)

@barbarabeeton Indeed: ours come out as Word files (if we see the edit trace) so are not related to the setting (which is done separately from editorial change)

@barbarabeeton Anyway, my point was that outside of maths I can't imagine may publishers wanting to take sources and re-create PDFs without the text being entirely revised

@JosephWright more amsmath fun:

  Check failed with difference files
  - ../../build/test/amsldoc-1.etex.diff
  - ../../build/test/amsldoc-1.luatex.diff
  - ../../build/test/amsldoc-1.xetex.diff
  - ../../build/test/amsldoc-2.etex.diff
  - ../../build/test/amsldoc-2.luatex.diff
  - ../../build/test/amsldoc-2.xetex.diff
  - ../../build/test/amsldoc-3.etex.diff
  - ../../build/test/amsldoc-3.luatex.diff
  - ../../build/test/amsldoc-3.xetex.diff
  - ../../build/test/amsldoc-4.etex.diff
  - ../../build/test/amsldoc-4.luatex.diff
  - ../../build/test/amsldoc-4.xetex.diff

@DavidCarlisle Oopsie

have we changed the filename normalisation in l3build?

../../build/test/amsldoc-1.etex.diff
::::::::::::::
*** ../../build/test/amsldoc-1.tlg      2017-09-02 21:59:59.213714500 +0100
--- ../../build/test/amsldoc-1.etex.log 2017-09-02 22:00:00.160331000 +0100
***************
*** 1004,1007 ****
  ...\glue 0.0 plus 1.0fil
  ...\OT1/cmr/m/n/10 1
  ...\glue 0.0 plus 1.0fil
! (.aux)
--- 1004,1007 ----
  ...\glue 0.0 plus 1.0fil
  ...\OT1/cmr/m/n/10 1
  ...\glue 0.0 plus 1.0fil
! (amsldoc-1.aux)

@DavidCarlisle That: yes

@DavidCarlisle Normalising out the test name itself made issues elsewhere

@JosephWright I'll sort out those first and see what's left:-)

@JosephWright I thought I remembered something along those lines

Let's start with a very unclear question:

@egreg Closed

OT (not reproducible, see comments)

Unclear (lack of MWE)

If I want to test the width which something produces, is \setbox0\hbox{#1} or \setbox0\hbox{\begin{tabular}{@{}c@{}}#1\end{tabular}} better suited?

Is this even on-topic? Moving fonts between systems seems off topic.

@TorbjørnT. Nope

A vote here for closing: tex.stackexchange.com/a/389548/4427

A vote here for closing: tex.stackexchange.com/a/389549/4427

If you please, of course.

@egreg Done

@JosephWright Thanks

@DavidCarlisle You might know someone involved in maintaining amsmath: tex.stackexchange.com/questions/141287/…

Unclear. Good night.

@TorbjørnT. oh I think that may be in one of @barbarabeeton's lists of amsmath features, it's not taking account of the list indent, I'll see what can be done...

@TorbjørnT. -- i can't check this now, but will do so when i get to a real computer. one known cause of thie failure to report an overfull line might be the fact that align* (and indeed all amsmath display structures assume that the full text width is available, and ignore the fact that the input is within a list. this is am ams style default, and i will grumble loudly if there is any attempt to change it except as (possibly) an option.

@barbarabeeton simpler test file here:

\documentclass{article}

\usepackage{amsmath}


\begin{document}



\noindent X\dotfill X

        \begin{enumerate}

\item
\noindent X\dotfill X

        \item
            \begin{align*}
                abcdefghijklmnopqrstuvwxyzabcdefghijklmnopqrstuvwxyzabcdefghijklmn
            \end{align*}
        \end{enumerate}

\end{document}

@barbarabeeton it doesn't report an overfull box as it's less than (full) display width but then it does indent by the list indent so sticks into the margin, so something isn't right it should either warn its overfull or center on the page ignoring the list indent.

@DavidCarlisle -- please also take note ... if this isn't properly documented in amsldoc, that should be done. (i'm ignoring the fact that align shouldn't be used on a 1-line display.)

@barbarabeeton that's just an artefact of the example you could add more lines with same behaviour.

@DavidCarlisle -- it should not indent by the list indent. as noted, i will grumble loudly. but i will test this next time i get to a real computer.

@barbarabeeton well it does (and I didn't change it:-)

@DavidCarlisle -- and as i said, about ignoring the list indent, one line or multiple lines, shouldn't matter.

@DavidCarlisle -- okay. i will try to track the history.

@barbarabeeton better example:

\documentclass{article}

\usepackage{amsmath}


\begin{document}



\noindent X\dotfill X

        \begin{enumerate}

\item
\noindent X\dotfill X

        \item
aaa
            \begin{align*}
a&=bcdefghijklmnopqrstuvwxyzabcdefghijklmnopqrstuvwxyzabcdefghijkl\\
1&=2
            \end{align*}
        \end{enumerate}

aaa
            \begin{align*}
a&=bcdefghijklmnopqrstuvwxyzabcdefghijklmnopqrstuvwxyzabcdefghijkl\\
1&=2
            \end{align*}

\end{document}

@DavidCarlisle -- thanks. (but couldn't you at least say 1 & \neq 2?)

@barbarabeeton don't be so conventional

@barbarabeeton but good news for you: it only happens with article not amsmart

@DavidCarlisle -- ah, thanks for checking that. one less thing to investigate at this end. i think this does need to be documented, and will make sure that's on the list.

@barbarabeeton too late to debug now but clearly what's happening is that some parts of the measuring are assuming the ams layout with full width displays which don't apply in other classes so it should be able to narrow that down and make it warn of the overflow without affecting the ams classes behaviour

@DavidCarlisle -- would it be reasonable to test whether an ams class is in use? again, if done that way, user documentation needed.

@barbarabeeton on the \binom \nulldelimiterspace that;s a straight bug in code that I added last year so I checked in a fix (as above) to our svn, probably I need to coordinate with you on a ctan release

@DavidCarlisle -- okay. thanks. i will try to pay attention to this tomorrow.

@barbarabeeton hopefully not (as other classes may do the same) I just need to check (when I'm awake) what settings amsart is using that are affecting this and then make sure amsmath uses the right length (so at a guess without looking at the code) if amsart arranges that \displaywidth is always \textwidth then if amsmath changes to test against \displaywidth rather than \textwidth it will make no difference in amsart but with article it will see that the alignment is too wide, or something...

@DavidCarlisle -- sounds good. i still think that documentation of the behavior should be considered.

@barbarabeeton yes while debugging I'll see what the behaviour actually is, then try to document it:-)

@DavidCarlisle -- i'm reworking the "short math guide". do you have any suggesitons for that?

@barbarabeeton not at the moment:-) I did however make some notes on that other stuff, I'll send an email now...