@cfr This has been very productive, I think, but it’s way past my bedtime, so I’ll just say good night to everyone. Let’s keep in touch about the fixes to Polyglossia.

@cfr Sorry, just digged into Babel’s welsh.ldf again: it has these two lines:

\def\headtoname{At}%  % `at' on letters meaning `to ( a person)

                    % `to (a place)' is `i' in Welsh

So the original author was aware of the problem. Signing off now.

@PauloCereda <3 I knew you'd like it :)

Currently lamenting how bad Texinfo is.

@SeanAllred <3

@SeanAllred uh-oh.

@PauloCereda I thought Texinfo was kinda like TeX. Nope. :( No support for even command aliases in HTML export.

=> no chance for semantic markup

@SeanAllred It's awful indeed.

@ArthurReutenauer I'd recommend putting that comment back. And it must be in the context of 'anfon' (and similar verbs).

@SeanAllred Why did you ever think that?!

@cfr Well, it is a TeX format.

@SeanAllred Presumably, you'd never looked at the source...?

@cfr Oh I have :) change a few catcodes around and it's not that crazy.

@SeanAllred I don't seem to have it installed ;).

The TeX part of Texinfo is not terrible. It's not that great, but it's not terrible. The terrible part comes in where the HTML export doesn't support the entire language.

@cfr But you have info, right?

@SeanAllred Yes, yes. I must have it. I just don't have a command texinfo. I've got a package texinfo.

@cfr :) I did have to install it with my package manager

But then again, I run OS X – not the most up-to-date UNIX

@SeanAllred I assume that info is texinfo? I think I knew that once but had managed to forget. texinfo is part of both base and base-devel which means that packages are entitled to assume I have it installed without listing it as a depends.

@cfr info is a command to parse/navigate Info files. Texinfo is a format used to generate HTML, PDF, and Info files from the same source.

@SeanAllred Ah. OK. That makes sense. Confusing that info is part of texinfo.

I guess I never knew it. Maybe I thought I knew it and forgot it. Hmmm...

@cfr, @ArthurReutenauer The 'portable' document business is an important worry, and one of the drivers behind latexrelease. I'm imagining that for a new format we're going to work hard to make updates generally much more controlled. However, I do wonder if there is any way to do what some other languages do and have the installation have several versions available 'out of the box'. If nothing else, one could go with latex3-<package>-<date>.sty or similar ...

... and load the version depending on the date. That would then also link to using the MD5 sum to check the exact file version is the same.

@DavidCarlisle ^^^ Any good as a concept?

@cfr, @ArthurReutenauer, @DavidCarlisle, @WillRobertson I may have mentioned I'd been wondering about font interfaces. There's an attraction to having a single interface for pdfTeX/XeTeX/LuaTeX but when I thought about it there is too much difference between the two models. Unicode system fonts are very different beasts from 8-bit encoded ones.

@JosephWright Could you please finish these synonyms? Thanks :) tex.stackexchange.com/tags/table-of-contents/synonyms

@yo' Done

@JosephWright thanks

@JosephWright yes possibly, maintaining entire dated versions of old code in the input tree is in practice a lot easier (if expensive in theory in disk space and search time) than maintaining conditional code within the same package.

Time to go to São Paulo today, have a great day, friends! :)

@PauloCereda have a good journey!

@JosephWright Yeah, in practise the set of fonts that are actually available with both 8-bit and Unicode TeX is small enough that I also don't think it's worth it. However, Frank doesn't (entirely) like the fontspec interface so we'll need to have a fair few discussions about how we see font support in LaTeX evolving in the future. I don't know what he doesn't like about it, though, so I can't comment further :)

@JosephWright Of course portability is an important concern, but I was trying to point out that there is in fact very little we can influence from the macro layer, or even from the format. There’s just too much possible sources of breakage out there – again, my experience assisting others in trying to have a completely reproducible output (never had to do it myself, thankfully) is that they end up freezing absolutely everything anyway.

@WillRobertson Indeed: I suspect we all have a view

@ArthurReutenauer I do see that :-)

@ArthurReutenauer I guess cross-system stability was actually easier in the past, when with plain and very few font files etc. you were starting in much the same place anyway

@JosephWright By contrast, the statement “Your will still be able to compile your TeX document in a lakh of years is just unhelpful – far better to educate users how to make their documents (reasonably) portable for when they actually need it.

@JosephWright I guess so. I’m trying to find that recent post on the GUTenberg list by someone trying to have pdfLaTeX have the exact same output – she’s working for a small publishing company that uses TeX for typesetting and needs that kind of cross-portability – the origin of differences was absolutely baffling.

@ArthurReutenauer I guess I'm thinking the part I can address is 'You will still be able to compile your document' in the sense 'Control sequences will still be defined' not 'The output will be the same'

@JosephWright Yes indeed. That part is of course important.

@JosephWright The discussion on the GUTenberg list boiled down to the following: two people were working on the same document using identical installations. One one person’s machine the text was reflown because the hyphenation of one word was different (“cons-tante” as opposed to “con-stante”). Why was it different? Well, one person was using the French patterns that don’t have any breakpoints in the word (long story ... basically “con” is vulgar slang), and said \hyphenation{con-stante}.

@JosephWright The other one did not have the French patterns at all and was thus using hyphen.tex, that breaks “cons-tante”. I am obviously missing some parts of the story (why didn’t the second person inherit \hyphenation{con-stante} from the first one?) but the interesting bit is to me that hyphenation patterns are in a way hidden because they’re part of the format (hence even \listfiles doesn’t see them).

@JosephWright The current situation is not helped by TeX not reporting the list of hyphenation patterns anymore ... but it was getting silly with a list of 70+ languages and variants in recent installations, so Javier was right to replace it with a simple “hyphenation patterns for 78 languages loaded.”

@ArthurReutenauer it would be possible for the standard hyphenation patterns to have some cross-format version of \ProvidesFile so the format recorded a version number and filename for each language then you could have some simple macro that displayed the currently active hyphenation file using the csname made up using the current language number.

@DavidCarlisle I think Babel already has something in that direction, I’m looking for it right now (just for reporting what language / version is loaded, not for setting it).

@ArthurReutenauer you can use the option showlanguages. But it doesn't show versions. There is also an option debug but I don't know when it will kick in.

@UlrikeFischer Thanks, that’s what I was looking for.

@UlrikeFischer Well, at the moment there are no version numbers attached to pattern files, but for some languages there are different versions that are embedded in the file name. German may be the only one.

Hi all, what controls the placement of accents on characters in math mode? I read through the TeX book but couldn't find any variable that specifically controlled this part (though it was mentioned in the main chapter "More on math"). Bonus points if I can change the spacing for individual characters or change the spacing for capital/lower case seperately.

The otftotfm manual (and I believe fontinst too) mentions this too.

@1010011010 which means basically you don't have much control over a primitive \mathaccent, if the base is smaller than the x-height of the font, the accent is placed over the x-height, and if the base is bigger than the xheight the accent is raised by that difference.

Hmm, that explains why the lower case accented characters look nicer than the capital letters...

Good maen

@JosephWright Does it make much difference if the fonts available on the system are different? Even if the installation of TeX is entirely identical, you'll get different results if you have, say, different versions of Times - even if the document does not specify a font as a list of possibilities.

@cfr As I said, I'm thinking more of packages having the same APIs, rather than saying you'll get the same visual output. That said, some people are very unhappy with the idea of loading system fonts at all for the reason you outline.

@cfr If I send a document to someone else and they are going to typeset it, I assume that something will change. Otherwise I can just send them the PDF. What I want is that the entire document doesn't break badly :-)

@ChristianHupfer Hei

@1010011010 Hello

is there a standard question for "use latex not tex on latex files" ? it looks like it ought to be this but @JosephWright answered in a comment and the actual answer is about bibtex tex.stackexchange.com/questions/102886/…

@DavidCarlisle Same problem: I thought there should be a dupe :-)

@JosephWright the other one I found had luatex specific answers

@JosephWright this question is quite clear, perhaps we should just answer it:-)

@DavidCarlisle Go for it: the current one is very good so would make an excellent maseter

ah he was editing...

@JosephWright I never send anybody source as they would have no idea what to do with it. But I am unhappy with the idea of using system fonts for roughly this reason, where the person receiving the document is also me on another machine or me in a few years. I accept that the output may not be identical anyway. But I don't want it to look radically different. Generally, for example - unless I'm unlucky - I get the same number of pages and more-or-less the same page breaks, bad boxes...

@DavidCarlisle You pinged me?

@ChristianHupfer history was erased so no:-)

@DavidCarlisle :-P Surely you found something to frown, most likely on the pdftex error question?

@ChristianHupfer see the comments above, Joseph and I were looking for a dup.. :-)

@DavidCarlisle No chance -- that question has never been asked before ;-)

@cfr Indeed

@cfr It's a tricky one, as this is all about fonts and the idea of a 'TeX sandbox' is tricky to sustain when there are other fonts out there (some good paid-for ones from Adobe for example)

@cfr but you are British (I guess) so using frozen fonts shipped with texlive is feasible, but when you see people asking about trilingual chinese/cyrillic/greek documents and you know their system has fonts for those languages (as does mine, as I can see their example text in a browser) then the effort required to set up a sandboxed environment compared to three lines of fontspec to make it work using system fonts in xetex or luatex means that you have to really need that stability.

@DavidCarlisle Even then, good luck setting up Aldus Nova Pro :-)

@JosephWright personally I stick to cmr10

@JosephWright @DavidCarlisle sits at home and still casts lead letters ;-)

@JosephWright I didn't yet have the need for something like it but I can imagine cases where it could be useful

@ChristianHupfer I refer you to my highest ranked answer.

@DavidCarlisle About letters made out of lead? ;-)

@DavidCarlisle: One of your legendary one-line- comments answers ;-)

@JosephWright: Perhaps off-topic post

@cfr But, again, if you are the author of the document and you don’t want it to change fonts suddenly you should just stick to one font per specification; likewise, if you work on a document collaboratively, you should agree on that policy beforehand. I don’t really see a legitimate use case where the font list feature (were it to exist) would pose problem. Surely you don’t mean to say that no one should use it in their documents?

@ArthurReutenauer I think the issue is that on different systems a font with the same name might not have the same metrics/glyphs/...

@JosephWright I agree that is is one of the issues. But Clea seemed to have a specific problem with the font list feature, while acknowledging that even one single font may give different results between installations.

@ArthurReutenauer I guess its the same as the question of glyph fallback: that's what you see in many other programs, but in a typesetting context one can argue that either the font/glyph should be right or the output should fail

@DavidCarlisle Oh, yes. I know. It just annoys me when people talk as if there are no losses in moving to Xe/LuaTeX and only a dinosaur would still be using pdfTeX. I just don't think that's true. There are always pluses and minuses. For many people, the minuses are greatly outweighed by the pluses. But for others, that isn't true and a non-dinosaur for whom that isn't true may have good reason to stick to (pdf)TeX. That's really my only point.

@cfr Remember @DavidCarlisle (and I) are on a team with a stated position that includes continued support for pdfTeX :-) (That is somewhat complicated: there are lots of issues worrying me!)

@ArthurReutenauer I think, ideally, what would be requested would be a particular version of a particular font. If that's not available, typesetting should fail. (I am deliberately overstating my position here. But I think there is something to be said for this.)

@JosephWright Yes. And I do appreciate that ;).

@ArthurReutenauer I think the font list feature would exacerbate the problem. The problem exists anyway.

@cfr oh sure. As an English speaking mathematician I do stick to pdftex (and cmr10, mostly) but there are other use cases..

@cfr the problem I see it [font list] solving (which is a problem I see often here) is an example text gets posted using some script, Indic, or East Asian or whatever. I can't run the MWE as I don't have the font specified, my browser says the needed glyphs are in some (possibly proprietary) font installed in the windows system fonts folder, so I make an answer using that, but I don't know if anyone else has that font (may have come with Windows or Office or I may have downloaded it or)...

... with a list, I could specify the original font, any font my system says supports that script and the document will produce output as long as the receiver has one of them. When the alternative is no output at all, worrying about minor upgrades in font metrics seems to me to be a secondary concern for most people who are not actually working for an archived publication.

@cfr Wholeheartedly agreed! This is the actual problem, in my opinion – people claiming that the move to XeTeX or LuaTeX has no consequences. I think such people, when you hear them – it’s usually more of a vague rumour in my experience – should be taken to task and shown the many legitimate use cases for sticking to pdfTeX. That has been my point all along.

@DavidCarlisle: Can this be a consequence of restricted write18 :

@cfr I don’t consider this a particularly extreme position, and I think the only thing stopping such a feature to exist is the lack of people to implement it. But if someone feels like working on that it would make sense – as an optional feature, at any rate – and would be in line with the way LaTeX worked in the 8-bit world (ConTeXt is another matter entirely).

@JosephWright Yes, of course, but again if a font list feature was added nobody would be forced to use it.

@DavidCarlisle Oooh: babel has been updated :-)

@ArthurReutenauer Also true

@ArthurReutenauer There are lots of complexities!

@cfr For sure, but I still think making people aware of the problem should still be a priority – to sum up, the technical issue is in my opinion that in nowadays’ TeX world, it is not longer true that a document can be relied on to yield the same output on different installations. I don’t think that the feature we’re discussing (that is still very hypothetical) would compound the current situation so terribly.

Can any of the current available mods (@JosephWright ;-)) summon Nicola Talbot? It's about a small issue with \newacronym in this question

@cfr It's all tricky: one has to wonder about people writing in languages other than English in the future (e.g. do we port inputenc to LaTeX3 when there are UTF-8 engines for other languages)

@DavidCarlisle Which could be the primary use case we would like to solve, actually – helping wizards on Stack Exchange :-)

@ChristianHupfer @NicolaTalbot See above /tex.stackexchange.com/questions/258636/…

@JosephWright Thank you. Would mentioning cricket have helped too? ;-)

@ArthurReutenauer of course!

@DavidCarlisle I can see we will have lots of fun when we do get to trying to fully work out a new form :-)

@cfr Meanwhile I’ve done the changes we discussed to Polygossia’s Welsh gloss file. If you want to check github.com/reutenauer/polyglossia/blob/master/tex/… – the code for \today should be pretty self-explanatory. I’ve already tested it on today’s date and will make a more comprehensive test document.

@JosephWright I really hope so. I have UTF8 input all over the place and would be at a serious loss if everything relying on it were to break in latex3. This is true even for the majority of my documents which are primarily - or even entirely - in English. And there are many languages other than English which are well-supported by T1. The same considerations which might make an English-only user want to stick with pdfTeX apply to other well-supported languages, surely?

@cfr UTF-8 is a given, the issue becomes more how one deals properly with the 'back end' (storing UTF-8 data in an 8-bit engine)

@cfr If we restrict ourselves to coverage in T1 then I think such things are reasonable, it's the full range covered by LaTeX2e/inputenc/fontenc that worries me

@cfr Note that ConTeXt have made a break here: LuaTeX-only

I'm officially on holiday. ;-)

@cfr As I've said, the team are well-aware of these issues and at present nothing is decided

@cfr I agree with you that LaTeX 3 should probably have inputenc or something equivalent, but you also don’t have to switch to LaTeX 3 and could go on with LaTeX 2ε for these documents; much like you don’t have to give up on pdfTeX even if there are native UTF-8 engines.

@ArthurReutenauer I think if we do provide anything like inputenc for a new format it's only going to cover UTF-8 input, not all those other encodings, and then probably only for the T1 range. However, that's a long way off: LaTeX2e is with us for the long-haul.

@JosephWright :

! LaTeX Error: Missing \begin{document}.

See the LaTeX manual or LaTeX Companion for explanation.
Type  H <return>  for immediate help.
 ...

l.30 \RequireLuaModule { lualatex-math } [
                                         2013/08/03 ]

@DavidCarlisle I suspected we would need that optional arg

oops :-) if you are reshuffling ltluatex can you add an optional argument so Will's code works:( (was trying Claudio's example with luatex...)

@DavidCarlisle I will do when I get home

@ArthurReutenauer Looks good to me;). Thank you!

@JosephWright That would make sense, of course.

@cfr You’re welcome! I’ll see about making a formal option if I have time tonight.

@ArthurReutenauer Mainly a case of porting some code, although I'd have to address for example case changing properly (doable if rather painful)

@ArthurReutenauer You mean for the o...? That wasn't in the file I just checked. Do note that this is the equivalent of providing of... in English. I'm not saying it is not nice to have. But it is not essential, in my opinion. I'm quite keen on you plan for getting this into Babel as you may imagine. I'm also wondering about datetime/datetime2.... Hmmm.... Would be nice to at least get the Breton out of the Welsh language files ;).

*************************************************
* lualatex-math warning: "wrong-meaning"
*
* I've expected \new@mathgroup to have the meaning
* macro:->\alloc@ 8\mathgroup \chardef \sixt@@n ,
* but it has the meaning
* macro:->\e@alloc \mathgroup \chardef {\count 18}\m@ne \e@mathgroup@top .
*************************************************
)

@DavidCarlisle Oops

@JosephWright always expect the unexpected..

It's time to update the TeX installation on my Mac. It appears that TeXLive 2015 has been released since I last updated. Should I just download the binary and reinstall or is there a better way?

@Gregory Download MacTeX: tug.org/mactex

@ArthurReutenauer I checked the ordinal abbreviations against a list I have in my office from the Gloywi course I took with the Centre for Welsh for Adults here. (Gloywi = Polishing) It is possible that somebody may tell you at some point that something doesn't match what they use. The problem is that there are many ways to construct numbers using the old, traditional system. But these abbreviations are the standard ones. (We have 2 number systems in Welsh and the older allows... creativity.)

@DavidCarlisle Did you see the irony of why this is happening?

@egreg Thanks. That's what I figured.

@JosephWright I didn't dare look:-) (I'm on holiday)

\__lltxmath_patch:NNnnn \new@mathgroup \cs_set_nopar:Npn { } {
  \alloc@ 8 \mathgroup \chardef \sixt@@n
} {
  \alloc@ 8 \mathgroup \chardef \c_two_hundred_fifty_six
}

@DavidCarlisle ^^^

ah.

@DavidCarlisle Also

\luatexbase@ensure@primitive { Ustack }

Did we cover that?

@JosephWright yes but I can't type (sorry)

@DavidCarlisle Is this some sort of zen?

\let\luatexbase@ensure@primitive\@gobble

@JosephWright ^^^

@DavidCarlisle Ah :-)

oh it fixed itself by magic

@JosephWright one could imagine other implementations than gobble but I took the view that prefixing or not was a global issue to be handled elsewhere.

@DavidCarlisle 'magic'

@DavidCarlisle Sounds good

is this a known feature too?

@DavidCarlisle This one is new on me: I'll have to take a look

@DavidCarlisle I take it this stuff only breaks with our dev code?

:23211513 not sure hang on...

@JosephWright seems so, need to tweak emulation a bit then:-) (how come the newmathgroup test didn't hit anyway, that's a change in the 2015 kernel... (no need to answer, I'll look in a bit:-)

@DavidCarlisle That's what I'd expect: we've not actually changed the namespace business in the wild

@DavidCarlisle Ah, I see what happens

@DavidCarlisle This is @WillRobertson's fault

\cs_new_protected_nopar:Nn \__um_new_cramped_style:N
  { \cs_new_eq:Nc #1 { luatex \cs_to_str:N #1 } }
\__um_new_cramped_style:N \crampeddisplaystyle

@DavidCarlisle Will defines the 'natural' names of the primitives but that would be best done using

\directlua{tex.enableprimitives("",
  {
    "crampeddisplaystyle",
    "crampedtextstyle",
    "crampedscriptstyle",
    "crampedscriptscriptstyle"
  })
}

oh so prefixing gone....

@DavidCarlisle Well we thought it was probably better to drop that long-term

@DavidCarlisle As this is @WillRobertson's code there should be no issue: I'll put in a pull request later on

@DavidCarlisle We aren't committed to that yet

@egreg After \tracingall of 168 Mb, I've found the unicode-math bug :-)

@JosephWright no, but I mean that's why it failed

@JosephWright oh I see newfam test in lualatex-base works as luatexbase loads etex.sty so puts the old version back....

s/lualatex-base/lualatex-math/

Who can tell me what the cm-super variant of cmex10 and cmsy10 are? The manual isn't very clear on the relation between the old names and the new

@1010011010 texdoc cmsuper lists all the names but doesn't say there are any math fonts?

@david @paulo ^^^^ ;-)

@Johannes_B expected behaviour

@DavidCarlisle I had the problem with ed the one time i tried to use it ;-)

@DavidCarlisle But cmex10 isn't a bitmap font, right?

Because the extensible brackets show as normally when I compile them (?)

e.g. they are not pixelated when I zoom in?

@1010011010 there are bitmap and type1 versions of it same as for all the original metafont-defined tex fonts.

@JosephWright catcode changes :(

@DavidCarlisle Where?

@DavidCarlisle Did you see the latest 2e bug report?

@JosephWright unicode-math:-)

@DavidCarlisle Oh, that

@JosephWright ... oh, what???

@DavidCarlisle Oh, don't worry, it's not an issue (or not at least one for us)

> While writing the editor does not always display the correct character.

@JosephWright I guess I could reply.

@DavidCarlisle If you remember how to make Gnats work :-)

@JosephWright yes

@DavidCarlisle Speaking of editors, did Frank mention to you he's bought a new laptop?

@JosephWright windows 10?

@DavidCarlisle Not quite :-)

@DavidCarlisle MacBook

@JosephWright ah

@JosephWright The curious thing is that it happens only after \setmathfont.

@egreg I'm not sure where else the data table gets read

@egreg I was going to suggest a fix but I think it probably needs @WillRobertson to look at it properly

@JosephWright I think it's related to the problem with LuaLaTeX and category codes

@egreg Did you try the demo with XeTeX ;-)

@DavidCarlisle I think in terms of the Lua code we are about at the point where we need to branch the 2e SVN so we can commit it, and to agree on what goes to CTAN now (with Élie of course)

@JosephWright I was thinking exactly same yesterday, where do we want to branch? just base?

@DavidCarlisle Can we do that?

@JosephWright svn is very flexible with branching, not like git:-)

@DavidCarlisle I only know the 'normal' SVN layout, which we don't have so I'm out of my depth

@DavidCarlisle Aside: does Javier use the babel directory or somewhere else? He's released an update but there have been no checkins

@JosephWright it may be easier to branch at latex2e-public so then the existing l3build scripts would all work in the branch i think

@JosephWright I think it moved out

@JosephWright I could do it but only sporadically online until tomorrow evening

@JosephWright although I'd just do cd /home/l3svn; svn copy latex2e-public latex2e-20150101

@DavidCarlisle What does that do?

How much of the Latex skills one has learned are transferable to Context? I've looked at Context a little, but I find I have to re-learn many things all over again, in terms of document structure and all the tricks and commands used in Latex.

@JosephWright makes a latex2e-20150101 copy lo latex2e-public (which needs committing if you do it that way, might be better to do it server side svn copy ^/latex2e-public ^/latex2e-20150101 then it's done at the server and you don't need to commit (but you need to check it out if you want a copy locally)

@Nasser not a lot apart from basic understanding of tex macro processing, document structure and design aims are rather different

@Nasser -- probably very little of your latex expertise can be transferred to context. context was designed to solve problems different from those that latex addresses. the only thing that seems to carry over, as i interpret it, is the concept of structuring. (caveat -- i'm not a context user, but i've attended quite a few talks by hans hagen on the subject, and edited all the articles he has written for tugboat.)

@JosephWright technically none of this was written by me :-) this is phst's part of the code. But I'll fix up ASAP -- there are a couple other fixes that I want to release now.

@WillRobertson :-)

@WillRobertson Does my pull request make sense?

@JosephWright got the checkin thanks

@JosephWright just flicking over the stack thing, I wonder if setrangecatcode ought to be \@setrangecatcode everything else in the package has an @ and it's targetted at package use. Don't know, wondered now that's the only top level command in the file.

@JosephWright Seems to! Can't merge right now but hopefully this arvo. The @ thing is annoying ... Might just remove it from the list of symbols.

@WillRobertson should you be changing catcodes of any symbol though?

@DavidCarlisle I didn't really think I was! Need to check the code but suspect this is an unwanted side-effect of the workaround for that LuaTeX "feature" in which only letter-catcode symbols receive kerning / italic correction. (Can't remember which while I'm on my feet.)

@DavidCarlisle Hmmm as soon as I write that I don't think I'm right. I'm not sitting on a computer at the moment so without consulting my code I shouldn't guess where the problem is :-)

@JosephWright Spam/rubbish: tex.stackexchange.com/a/258743/16967