@DavidCarlisle you mean \whatingwhatwhats ? No, only \smurfingsmurfysmurfs :)

has anyone used a format where it is set to 1.

@DavidCarlisle well, I've used 0 many times, but not by \the\savinghyphcodes

@yo' I'll take that as a "no" then :(

Where does it come from? It doesn't seem to be in the TeX reference

@yo' etex

@DavidCarlisle ah ok, makes sense then, the reference I use (and love) doesn't include eTeX :(

@PauloCereda btw, Chuck Norris can also decide the halting problem. Because he can decide everything.

@DavidCarlisle Just did: see mail

@DavidCarlisle Direct e-mails from Chris?

@yo' He actually can't decide. For him, all TM's just halt. :)

@JosephWright the list is bouncing his IP address

@DavidCarlisle Ah

@PauloCereda that's a decision, too

@DavidCarlisle It does work

@yo' for the TM's, not his decision. :)

@JosephWright so why don't we use it instead of documenting that changing encoding mid paragraph is broken...

well, his decision is final

@DavidCarlisle Probably 'stability' :-)

@DavidCarlisle Haven't tested LuaTeX

@DavidCarlisle > The motivation behind the ε-TEX extension \savinghyphcodes was that hyphenation heavily depended on font encodings. This is no longer true in LUATEX, and the corresponding primitive is ignored pending complete removal. The future semantics of \uppercase and \lowercase are still under consideration, no changes have taken place yet.

@DavidCarlisle Works for pdfTeX, XeTeX, fails (as expected) for LuaTeX

@JosephWright chris and I just asked same question, does it allow per language codes in same para. I suppose I should write a test case...

@JosephWright I think I read that once:-)

@DavidCarlisle I'm testing now: suggestion for a good second language to use (is #2 no-hyphens?)

@DavidCarlisle Quick test suggests all is well: I added \language2 at the end of my quick test par and still got the hyphenation point for English even with lccodes changed

@JosephWright How to deal with questions like tex.stackexchange.com/questions/174477/…

@Johannes_B Borderline but I've closed OT (bug report)

@JosephWright I have feeling we need to sort this out before TL2016;-)

@Johannes_B I was going to zap it, but the system didn't allow me (@Joseph was faster)

@JosephWright @yo' Thanks :-)

@JosephWright Should be reported to the package, not the class.

@DavidCarlisle True

@DavidCarlisle Possibly I need a more complex test as English letters as 'sensible' lccode in most other languages: need one where it changes I guess

@Johannes_B Changed

@JosephWright Thanks

@JosephWright mixture of Czech and Russian? :D

@yo' I was going to ask you (for a slightly different reason) for a czech example using letters from the characters that are in T1 but not in latin1, eg compare this german example although by coincidence its almost readable (compare pdflatex with xelatex or lualatex)

\documentclass{article}

\usepackage[utf8]{inputenc}
\usepackage[T1]{fontenc}

\begin{document}

[straße]

\end{document}

Guys, what's a quick way to disallow a math sequence to be broken?

@PauloCereda {}

The opposite of \allowbreak. :)

@DavidCarlisle Oh thanks!

@PauloCereda That also sets the glue at natural size (like \hbox) otherwise set \binoppenalty and \relpenalty to 10000

@DavidCarlisle Thanks, it works like a charm. :)

@DavidCarlisle Indeed: need a good test for the language question (tricky: I guess a change of input encoding will be required)

@DavidCarlisle V současné době je LaTeX rozšiřován týmem LaTeX3.

@DavidCarlisle Unicode test of course no good as lccode values do not change

is that enough?

oh you want it in ISO-8859-2, right?

@yo' no utf8 input os fine

@DavidCarlisle ok :)

@yo' Looks OK (or at least does what @DavidCarlisle expects!)

@yo' let's see if it messes up...

@JosephWright well, it doesn't contain capital letters

@yo' Try it with XeLaTeX!

@DavidCarlisle It does

Čeleď (latinsky familia) je základní taxonomická kategorie hierarchické klasifikace organismů tvořená příbuznými rody.

^^ here you have a nice capital accented letter :)

@JosephWright which one?

I was wondering if one can guess if this drawing was made in tikz? Here it is

@yo' so typeset this with xelatex:

\documentclass{article}

\usepackage[utf8]{inputenc}
\usepackage[T1]{fontenc}

\begin{document}

[straße]

[V současné době je LaTeX rozšiřován týmem LaTeX3.]

\end{document}

@Nasser Could be: without more info we can't really tell

@DavidCarlisle I spot two problems :D

Here is the pdf file cds.caltech.edu/~murray/books/MLS/pdf/mls94-complete.pdf

I think all the drawings in the book looks like Tickz but I am not sure.

@Nasser Ask the authors: could be PSTricks, Asymptote, ...

@yo' the main problem is that it doesn't stop and scream at you, but just silently permutes and drops your letters

@DavidCarlisle which is considered bad, right? :)

@JosephWright professors never reply to email :) so no point of trying

needed to say, if I remove both packages, it still doesn't work. Is that correct?

@yo' In a way

@yo' We are trying to get to grips with what is 'right' here

@JosephWright ok, so XeLaTeX without any package doesn't actually support unicode...

@yo' With pdfTeX it doesn't work without the two packages, so failing there is fine in all cases

@JosephWright yep, that I agree. (I only don't like it failing silently, but that's another story)

@yo' Sort-of: we've never changed the default font so it's CM in OT1 encoding

@yo' well perhaps but so far the only plan to make it better that doesn't have problems was posted here yesterday:

@yo' @DavidCarlisle is trying to find a way to have a pdfTeX ->XeTeX/LuaTeX switch work with no changes at all to output if the document doesn't change (basically)

@DavidCarlisle in other words, to realize: Ah, user wanted to use T1 and utf8 because of these two packages being tthere, let's give that to him. ?

@yo' Broadly

@yo' yes but the problem is that (as set up in TL2014, and probably TL2015) it isn't possible to give it. You can set up a T1 encoded font but hyphenation was set up at format time assuming unicode so even if you get th eright letters hyphenation is messed up

@DavidCarlisle is that problem even with [utf8]{inputenc}?

@yo' That's the easier part to solve as it's a no-op with XeTeX

@yo' It is a problem with every package, it is not currently fixable

@JosephWright now the two of you seem to contradict :P

@yo' No we are in agreement just abbreviating differently:-)

@yo' No, we've been working hard on the detail and are on the same page

@DavidCarlisle Quite

@JosephWright ah ok

Friends, any suggestions on rephrasing this sentence: Table~\ref{tab:performance} presents an overview of how each approach behaved during the performance tests.

@yo' inputenc deals with getting chars into LaTeX, fontenc with getting them out. With XeTeX the 'in' part is easier than for pdfTeX but that doesn't directly address the 'out'.

@yo' but basic fact is that in tl2014 if you use classic 8bit fonts with xelatex or lualatex you will get incorrect hyphenation (whatever packages you load)

@JosephWright ok get it I think.

@DavidCarlisle font package wrappers?... No that sounds too bad

@yo' You'd have to recode the fonts to use Unicode slots: XeTeX (like pdfTeX) 'bakes in' hyphenation at format-building time

@yo' we could have a wrapper if we had any idea what to put in it:-)

@JosephWright well, that was my thought (recode the font)

@yo' Back to some extent with 'is this sensible': Latin Modern is available in Unicode form and is trivially-easy to load with XeTeX

Of course, the metrics are different from CM so that's not an option, ...

@yo' yes that works but then you haven't got a T1 encoded font anymore, just as not using cm or ec and using latin modern, but as Joseph just said lm isn't an exact drop in replacement

There is though CM Unicode

Ah, still not Unicode encoded!

@JosephWright provide Unicode-encoded CM? :)

@yo' Actually, trying to understand quite what it is set up as!

@cfr I meant your reputation in tex.stackexchange.com/questions/224771/…

Help? :)

@JosephWright ok

@PauloCereda what with?

@yo' Testing I think it is Unicode :-)

@PauloCereda We tested the performance of each approach; the overview of the tests is in Table~4.

$ find -name t1\*.fd |wc -l
308

@DavidCarlisle not a small number

that's a lot of fonts to re-encode, then there's t2 x2 ...

@yo' Thank you! You got close, it's table 3. :)

@DavidCarlisle provide a tool for re-encoding on-the-fly? Similar to mktexpk or how's it called?

After all, we speak about one code table per encoding, no?

@PauloCereda you're welcome. It's not a perfect sentence, but I'd say it's fine

@yo' possibly, but the 8bit font might be type1 or metafont (or both) or .. and the alternative of just saying "it doesn't work, use a unicode font" has some appeal, then if anyone actually needs their 8bit klingon font re-encoded to unicode private use slots they'll just have to do it

@DavidCarlisle well, I don't object :)

Coould we close tex.stackexchange.com/questions/166259/… as a dupe of tex.stackexchange.com/questions/1092/…?

@Johannes_B not that one I think

@yo' What to do with a queation like that?

@Johannes_B I mean, the dupe doesn't seem to be the right dupe

I'd simply zap it as unclear. It seems to me that the user wants to instill a new pacakge on and old TL. You can't teach an old dog new tricks, they say.

@yo' I joined you :-)

@PauloCereda @egreg @JosephWright Are any of you using homebrew? I seem to have screwed something up, and I want to remove everything and reinstall. I thought I was doing it correctly, but I still get errors in the initial install.

I've just got a rare example of a submission whose author obviously at least saw the class manual :)

@yo' reject it. (S)He clearly didn't spend time doing the research and spent too long worrying about the tex.

@AlanMunn I've got homebrew installed: don't you just remove all the stuff it adds in /usr/local?

@JosephWright Yes, I thought I did that, but I think I missed the /usr/local/share/doc and that was causing the problems. I'm trying to delete that now and will try again to install.

@AlanMunn I don't know how to remove it, sorry.

@JosephWright It seems that homebrew doesn't play nicely with things you've installed manually in /usr/local? (Judging from what happens when I do brew doctor)

@AlanMunn It does get a bit tetchy

@AlanMunn I had to set some group codes to keep it happy (/usr/local/share/doc stuff from GhostScript)

@JosephWright Hmm. It's not quite clear to me what I should do with the output of the doctor diagnosis then. Typically I've just installed stuff myself into /usr/local.

@AlanMunn What's it saying?

@JosephWright E.g.

Warning: Unbrewed .la files were found in /usr/local/lib.
If you didn't put them there on purpose they could cause problems when
building Homebrew formulae, and may need to be deleted.

Unexpected .la files:
    /usr/local/lib/libaspell.la
    /usr/local/lib/liblzma.la
    /usr/local/lib/libpspell.la

Same for .pc files, and similar for static libraries.

And a much more extensive list of unbrewed header files. (I think mainly from the same installed stuff.)

And I don't like the suggestion of changing the ownership of /usr/local/share/man to my user. This seems so wrong.

I think this is related to their fetish against using sudo, but if you have a multi-user system, making /usr/local/... belong to a single user is crazy.

@AlanMunn I'm not 100% happy with it but went with it as I couldn't work out how to work around it!

@AlanMunn I think for a multi-user system they seem to recommend installing outside of the usual location

@AlanMunn They've got it wrong about installing programs more generally of course: you do need root permissions to install all sorts of stuff

@JosephWright Right. It seems very much biased to just a single user's perspective. I guess I'll press on and hope for the best. It's a bit annoying though. But some software has a lot of dependencies and so it helps to have a package manager for installation.

Do we have a generic add entry to list question to dupe this one?

@AlanMunn Much the same as my view: it's annoying but no more

@JosephWright But now everything I want to install I get ` No available formula`

@AlanMunn Eek

I'm trying to install some python stuff.

@JosephWright So half of what I need apparently isn't packaged for homebrew although it used to be? (Since the documentation I'm reading recommends using it.)

@AlanMunn Possibly

@AlanMunn My homebrew is currently 'happy': name a keg to test

@JosephWright pyparsing or python-dateutil

Palladium:unpacked joseph$ brew install pyparsing
Error: No available formula for pyparsing
Searching formulae...
Searching taps...

:-(

@AlanMunn Install using pip and the system Python install?

@Johannes_B It is rather a precise target...;)

@cfr Oh no ...

@JosephWright More things to learn... :(

@AlanMunn You can put /usr/local under the control of a single, dedicated user account without sudo rights. But I don't see that would solve this kind of problem.

@AlanMunn Well the Mac does come with Python, so the Homebrew approach I'd expect is to use that rather than install a second one

@cfr I think that is the homebrew advice for these cases

@cfr Thanks. I've managed to solve the permissions problems. The problem now is that things that supposedly are available under homebrew don't seem to be (any more?).

@JosephWright I see. So perhaps the document I'm reading is before Mac had an up to date python or something.

@AlanMunn Could be

@JosephWright How do I find out about pip then?

@David that's some nice code here :D tex.stackexchange.com/questions/224910/…

@JosephWright No worries. I figured that out. Maybe I don't need homebrew at all.

@AlanMunn Could be

@AlanMunn I only had it to try to build TeXworks, but there are some recent checkins that break the build :-(

@yo' didn't dare look, I just searched for "sc" :-)

@DavidCarlisle lol

@DavidCarlisle Interesting discoveries re. babel

@JosephWright I was looking how to hook in a double loading and saw ` \L 00B5 039C 00B5` which looked oddly familiar...

@DavidCarlisle :-)

@DavidCarlisle As you'd expect, I've been checking that ltunicode.ltx gives the correct results for all of these (it does)

@DavidCarlisle Interestingly, lualatex.ini leaves "00F7 alone

I couldn't work out why it was there since as it comments xetex.ini does that already, but I'm not sure of the timing or of the timing of when it uses \xetexinputencoding

@DavidCarlisle No, don't get it either

@JosephWright we need to get the test release out then add Javier to the growing list of people to ping....

@DavidCarlisle Yes

@DavidCarlisle Except we're only doing what is already in the .ini files, which come before babel, so we should be OK!

@JosephWright yes I think we are Ok but if babel's going to overwrite it anyway it doesn't matter what we do (I know it doesn't overwrite everything but..)

@DavidCarlisle True

@DavidCarlisle See mail

@JosephWright thx

@DavidCarlisle The 'release' ltfinal sets codes twice, once before patterns and then again after them to allow for any 'leakage' of changes. So babel is covering the patterns part but not the final document stage (hence needing the .ini). Hopefully updated ltfinal deals with both areas.

@JosephWright almost sounds reasonable put like that...

Good job, people. :)

@PauloCereda ooh

@PauloCereda excellent lion, best of the series!