Hey, not sure if this is a bad place to ask this, but I'm trying to update my tex distribution (the mattex package) but it seems to be absent from all the mirrors on ctan I've tried so far

anyone have an idea where I should look first on this?

Well, look next

and I found it on a different mirror

weird

@DavidCarlisle I guess I will

People are strange: latex-community.org/forum/viewtopic.php?p=96840#p96840

Amazing how only the making of a minimal working example can solve/find the cause of a problem. I was just about to ask something here (without a MWE) and realized you guys'd kick my butt not like it, then I made it and discovered the problem... :D

@egreg oh no

@DavidCarlisle ouch :)

@Johannes_B Just saw that. Indeed, strange People inhabit this world... o.O

@Johannes_B People are probably not very fluent in english and they are strangers.

@GuilhermeZ.Santos do you think this question should be tagged tikz-pgf ?

@DavidCarlisle It could be... Though the OP did not request for such solution. Still, I'm disapointed as picture mode could not handle this simple task...

@GuilhermeZ.Santos picture mode is too regular and professional looking to fake hand drawn work.

@DavidCarlisle Ohh I see. You don't want to shame picture mode with such horrible examples of usage.

@DavidCarlisle \protected\def all the things!

@DavidCarlisle I really am tempted ...

@DavidCarlisle The issue basically makes the point that restricting ourselves to TeX90 is increasingly hard to justify as real issues can be solved quickly by using e-TeX

@DavidCarlisle: "~" should not work, as it cancels the shell expansion.

[paulo@cambridge ~] $ cd ~
[paulo@cambridge ~] $ cd "~"
bash: cd: ~: No such file or directory

@PauloCereda ah of course my message was nonsense bash expanded the ~ (@UlrikeFischer)

@DavidCarlisle <3

@PauloCereda my emails (unlike my code) sometimes has bugs (or even typos)

@DavidCarlisle /gasp

@DavidCarlisle I thought so, it is similar to the "short tilde " expansion

@JosephWright I was tempted but it doesn't work:-)

@JosephWright except I was thinking of the - case (recent abnt class fix) if you protected def ~ to the nonbreakspace it doesn't expand, but it seems to terminate the filename (unlike protected def with _ I used the other day. I should try again (I did build a format with protected ~ but it didn't work as I expected (may have been issues with bash expansion though)

@PauloCereda it should work in tex though as kpsearch should then see the literal path starting from ~ and do its own tilde lookup

@DavidCarlisle oh I see.

@PauloCereda see comments in texmf.cnf:

 ~ expands
% to %USERPROFILE% on Windows, $HOME otherwise.

@DavidCarlisle Interesting!

@DavidCarlisle Mail re. e-TeX positive

@DavidCarlisle When used in \inputit doesn't seem to end the file name. pdflatex \protected\def~{blblb}\input{test~tilde} gives as error message test~tilde.tex not found. which is better then the error from pdflatex \input{test~tilde} but still not what one would want.

@WillHunting 'ello!

@DavidCarlisle (and others) L3 bundle sent to CTAN

@UlrikeFischer but stopping ~ expanding should work at the start of the argument if the engine's kpse lookup treats it as your hone directory (but I'm not sure that's working)

@JosephWright ah yes hadn't seen it. all good.

I LOL'ed. :)

@JosephWright the downside of \protected... @UlrikeFischer's \protected\def~{blblb}\input{test~tilde}

@DavidCarlisle I wouldn't call it a downside -- after all it doesn't make the situation worse. \input{test~tilde} didn't work anyway.

@UlrikeFischer yes but \newcommand\foo{zzz}input{test\foo} works as input testzzz but the \protected version \NewDocumentCommand\foo{}{zzz} doesn't as \foo won't expand. so if latex were to move (like expl3) to all user-level definitions are protected then mostly everything is better, except for the things that are not.

@DavidCarlisle Ah. You meant this. Yes there are cases where you don't want a robust command.

@UlrikeFischer with the latex mechanism \protect can be locally \empty or \noexpand or \string or \noexpand\protect\noexpand (my personal favourite:-) or whatever's needed to get different effects, but etex protection, once it is applied, is harder to ignore

@DavidCarlisle why are you so reluctant of using packages ^^^^? Haha

@GuilhermeZ.Santos why load a package for a for-loop if there is one already in the format?

@DavidCarlisle Code simplicity I belive. =P

@GuilhermeZ.Santos well not really, the for-each macro you used is not much diferent than \@for just doesn't require a @, the simplicity in your answer comes from the point I made in the comment under mine, it is much easier to just make a vertical list of href and stick it to the side of the first line text, doing loops inside the table with changing first column is more complicated (but was what was asked for:-)

@DavidCarlisle True. I assumed, since the OP used centering inside flushleft, that as long as the output was as he wanted, everything would be ok.

@GuilhermeZ.Santos who cares what the OP wants, idea is just to write some interesting code isn't it?

@DavidCarlisle: ^^ my talk had no value. :)

@GuilhermeZ.Santos re your comment about _ do you mean the MWE as posted or are you using a class with abnt?

@DavidCarlisle LOL, abntex2'd probably fix that bug of yours.

@GuilhermeZ.Santos abnt class doesn't support _ in filenames:

@DavidCarlisle Is it not gb4e (indirectly loaded by linguex)? If not then the guy who answered apparently dos not know much about TeX... ;)

@GuilhermeZ.Santos oh yes so it was, well I blame ABNT it probably mandates the use of linguistics instead of mathematics, or something,

@DavidCarlisle ooh

@PauloCereda found it: Regra ABNT 555: Você deve estudar lingüística e ser tão fluente em línguas como aquele cara Tex

@DavidCarlisle ooh your Google Fu is really impressive!

@DavidCarlisle I'm with @UlrikeFischer: not a downside, just tightening up on semantics

@DavidCarlisle We've discussed this :) There are things that are commands and things that are (basically) 'text shortcuts'

@JosephWright yes I know and it almost makes sense until you switch (as I guess we will effectively one day) and someone says "I've got 40 years of collected \newcommands which are commands and which are text shortcuts...?" I don't think (currently) there is anything much that can be done other than document the different behaviour.

@DavidCarlisle yes, ABNT sucks. But abntex2 is awesome, makes it easy to comply with suckin' ass unusual ABNT rules..

@DavidCarlisle @egreg Who should I blame here

\documentclass{book}
\usepackage[greek,ngerman]{babel}
\usepackage{varioref}
\usepackage[LGR,T1]{fontenc}
\usepackage{microtype}
\usepackage{hyperref}
% \makeatletter \def\@roman#1{\romannumeral #1}\makeatother %restore latex command works
\begin{document}
\pagenumbering{roman} %grr
\chapter{label}\label{chap:A}
blub\footnote{blub \vref{chap:A} blub}
\end{document}

! Argument of \MT@res@a has an extra }.

varioref, hyperref, microtype, greek.ldf are all candidates ...

@UlrikeFischer babel-greek, I guess

@egreg But what it should do instead to avoid that roman page numbers are printed in greek if the shipout happens in a greek context?

@UlrikeFischer the official answer is that it should use an encoding that has latin letters in the latin slots, but as that doesn't really work there has never been an official Tx encoding for greek so LGR it is and things break and have to be fudged as well as can be managed.

@UlrikeFischer See this

\documentclass{book}
\usepackage[greek,ngerman]{babel}
\usepackage{varioref}
\usepackage[LGR,T1]{fontenc}
\usepackage{microtype}
\usepackage{hyperref}

\makeatletter
\AtBeginDocument{%
  \protected\def\ensureascii{%
    \ifx\protect\@typeset@protect
      \fontencoding{T1}\selectfont
    \else
      \expandafter\@firstofone
    \fi
  }%
\pdfstringdefDisableCommands{\let\ensureascii\@firstofone}%
}

\begin{document}
\pagenumbering{roman} %grr
\chapter{label}\label{chap:A}
blub\footnote{blub \vref{chap:A} blub}

@egreg No wonder that I didn't get anywhere -- I was looking at the texlive 2015 greek.ldf (which uses \textlatin and not \ensureascii) while running texlive 2016. Sometimes too many tex systems have drawbacks ... This seems to work fine, I will sent it Javier.

@DavidCarlisle Indeed: not sure we can/should change \newcommand in the absence of arguments (\newcommand\foo[2][]{... is different)

@DavidCarlisle BTW, you said you'd got an example recently where the expansion behaviour of a 'fragile' command inside \edef was used 'by design'. Don't remember where, do you?

Karl's asked me to write a TUGboat about xparse: teach me to talk about stuff!

@PauloCereda You should write something about templates!

@JosephWright I should forward you the reply I sent to @barbarabeeton and Karl when they asked me for something the other day, perhaps you could re-use parts:-)

@JosephWright oh my

@JosephWright no, trying to remember (I think it was more "by accident" than "by design" but it was a question where something failed because it had become e-tex protected)

@DavidCarlisle Did if feature 'BXXXXX off'?

@JosephWright used slightly more words but in essence...

@DavidCarlisle tex.stackexchange.com/questions/333743/…, perhaps?

@JosephWright ah could have been

@JosephWright you've got mail (also about \protect(ed) as it happens:-)

@DavidCarlisle That one isn't 'using' some useful form of expansion of a macro that should be protected, so I'm not going to count it as a problem ;-)

@DavidCarlisle I was thinking if we do want to use \protected to start with things like \newcommand itself, \newcount, etc., which definitely fail inside an \edef or \write

@JosephWright yes we probably should

@DavidCarlisle Context of that forwarded mail would be nice ...

texhax thread but another mail coming up

@DavidCarlisle Let's do one release with just the test for e-TeX and see who (else) complains

@DavidCarlisle Ah, that thread from TeXhax

@JosephWright yes. If we start using \protected I'd be tempted to \let it to \@empty at least within latex.ltx but perhaps always just in case. (I know we have removed other compatibility branches but ....)

@DavidCarlisle Like I say, let's first see what else comes up from a simple test!

@DavidCarlisle Shame they didn't ask me: I'd have gone for an article which came down to 'Don't mess about, use e-TeX' :)

@JosephWright I could ping barbara now and she'll take that as volunteering to write a second TB article.

@DavidCarlisle Oh, go on then

@DavidCarlisle Just testing my l3build updates against the LaTeX2e set: looks fine but we probably need to get the rest of the team to think about \hjcode values in LuaTeX

@JosephWright yes I must read that section again my understanding of exactly when they need to be set and when you can get away with lccodes comes and goes (the latter at this exact moment)

@DavidCarlisle I'll have to check on what is happening with \savinghyphencodes and babel, true

@egreg It doesn't work if the label is on an arabic number page (in the mainmatter):

\documentclass{book}

\usepackage[greek,ngerman]{babel}
\usepackage{varioref}

\usepackage[LGR,T1]{fontenc}
\usepackage{microtype}

\usepackage{hyperref}

\makeatletter
\AtBeginDocument{%
  \protected\def\ensureascii{%
    \ifx\protect\@typeset@protect
      \fontencoding{T1}\selectfont
    \else
      \expandafter\@firstofone
    \fi
  }%
\pdfstringdefDisableCommands{\let\ensureascii\@firstofone}%
}

\begin{document}
\pagenumbering{roman} %grr

blub\footnote{blub \vref{chap:A} blub}

\newpage
\pagenumbering{arabic}

! Undefined control sequence.
\is@pos@num@ #1\@nil ->\ifx \@nil #1\@nil

@UlrikeFischer Blame Frank and his varioref

@egreg Why do you used \protected? In the babel.def it is a simple \def/\edef, and without it is seems to work fine @DavidCarlisle @JosephWright

@UlrikeFischer You get nothing in the .aux file, which means that when you use \pageref{foo} in a Greek context you get rubbish.

Does pgfmath have an e constant? (Hard to search for...)

@AlanMunn page 928 of the manual says the function e is recognized

\pgfmathe returns the value 2.718281828.

@StrongBad Great, thanks.

It also looks like exp(1) should work

@UlrikeFischer This is the most probable cause :-)

@egreg Well I get unexpected features with \protected too:

\documentclass{book}
\usepackage[greek,ngerman]{babel}
\usepackage[LGR,T1]{fontenc}

\makeatletter
\AtBeginDocument{%
  \protected\def\ensureascii{%
    \ifx\protect\@typeset@protect
      \fontencoding{T1}\selectfont
    \else
      \expandafter\@firstofone
    \fi
  }%
\pdfstringdefDisableCommands{\let\ensureascii\@firstofone}%
}
\makeatother
\begin{document}

\selectlanguage{greek}abc \roman{page} hey where is my greek?

\end{document}

(Sorry forgot to remove the hyperref code ...

@UlrikeFischer That's easy: define \ensureascii with an argument

  \protected\def\ensureascii#1{%
    \ifx\protect\@typeset@protect
      \expandafter\@firstoftwo
    \else
      \expandafter\@secondoftwo
    \fi
    {{\fontencoding{T1}\selectfont#1}}%
    {#1}%
  }%

is this the easiest way to compile a latex doc to an image on windows? tex.stackexchange.com/questions/11866/…

Basically I just wanted to compile an equation to an image so that it could be imported into word

@baxx I use IrfanView but there are other approaches

@JosephWright I think I found a minor bug in biblatex: tex.stackexchange.com/questions/334839/…

@egreg Ah yes. I think I will go for the definition without \protected. In case that a \pageref in a greek context is needed one can use \ensureascii{\pageref{label}}.

@UlrikeFischer This simply means that you don't want to use \ensureascii in the (re)definition of \@roman.

@UlrikeFischer I don't think this is what babel-greek wants.

@egreg No, if I remove \ensureascii from \@roman then I can get page number in greek if the shipout is when greek is active.

@UlrikeFischer The purpose of \ensureascii is to make sure references to a \label that refers to something defined in terms of \roman will show a “real” roman number. So you're defeating this. The user cannot know whether a \ref should be encased in \ensureascii.

@UlrikeFischer I know it's a tough problem.

@egreg Well I want code that compiles. If this means that I have to be a bit careful when using \pageref in greek context, I can live with it. Imho it is anyway not such a good idea to store only a formated number in a label. A saner solution would be to store the number itself too, so that package like varioref can use.

@StrongBad IMHO, too minor to ever be of any interest... but I quite don't know as I've never used biblatex

@UlrikeFischer This works:

\documentclass{book}

\usepackage[LGR,T1]{fontenc}
\usepackage{microtype}
\usepackage[greek,ngerman]{babel}

\usepackage{varioref}
\usepackage{hyperref}

\makeatletter
\AtBeginDocument{%
  \protected\def\ensureascii#1{%
    \ifx\protect\@typeset@protect
      \expandafter\@firstoftwo
    \else
      \expandafter\@secondoftwo
    \fi
    {{\fontencoding{T1}\selectfont#1}}%
    {#1}%
  }%
\pdfstringdefDisableCommands{\let\ensureascii\@firstofone}%
}
\makeatother

\begin{document}
\pagenumbering{roman} %grr

@egreg ;-) compiles also in the large document. But I will try later to understand why this doesn't explode ...

@JosephWright I'm confused:-)

if protected active characters expand when looking for filenames, how come this works

@DavidCarlisle With a test.tex and a t_st.tex file in the current directory, the following document inputs twice test.tex

\documentclass{article}

\protected\def\es{es}
\catcode`_=\active
\protected\def_{e}

\begin{document}
\input{t\es t}
\input{t_st}
\end{document}

@DavidCarlisle doesn't work for me:

@egreg yes so it seems, so what the heck is the \protected doing in the answer I linked to, the mwe there seemed to work with that change is it just making the recursive use \gb@ifnextchar^{\automath@two_ safe and then working by accident:-) (too tired tonight for this)

@UlrikeFischer nor me, nor egreg, there seems to be some evidence building that my suggestion doesn't work\:-)

@DavidCarlisle ! LaTeX Error: File monog_pretextuais.tex' not found.`

@egreg yes that's what I got for the mwe in the question but if \protected\def_ added to automath doesn't stop _ expanding why does it do that

@DavidCarlisle The error message comes from \write, but the file name is built with the standard TeX rules

@egreg ofor .... sake it's our own trick code in \input ?

@DavidCarlisle I suppose \protected should be fixed to not expand also during expansion for the primitive \input

@egreg that did occur to me as one way out, so i could claim that my suggestion was correct barring an etex/xetex/pdftex/luatex bug in not respecting \protected here

@DavidCarlisle Unfortunately \ifincsname doesn't help here.

@egreg tried that:-) also tried inserting romannumeral tricks but they also expand protected commands

@DavidCarlisle Unless you change \input to do some \csname tricks.

@egreg if changing input could simply define ~ via classic latex \protect and get \input to set \protect to \string and \edef the filename before using the real input then only works in latex and only if you use braces

@DavidCarlisle The simplest thing is to warn users of gb4e that file names with _ are not allowed for \input when \automath is active.

@egreg yes odd thing is OP accepted my answer even though it's wrong. I suppose you are well used to that by now.

@DavidCarlisle My answers, contrary to yours, are never wrong. :P

@egreg, @AlanMunn ^^

@egreg Or simply to precede \input statements with \noautomath.