Time to go to SP, you guys behave. :) Have a great day, friends!

@PauloCereda I just caught a train with a delay of just 75 minutes. :)

@JosephWright See tex.stackexchange.com/questions/280692/… in relation to texdev.net/2012/08/26/exploring-chemfig-going-further

@JosephWright Using internals is not always a good idea. ;-)

@egreg When I wrote that post there wasn't a public interface available (is there now?)

\tracingonline1

\ifx\directlua\undefined\else
\directlua{
% callback.register('define_font',font.read_tfm)
% manual sec 4.2.2, simplified
 callback.register('define_font',function(name,size) local f=font.read_tfm(name,size) return f end)
}
\fi

\font\x=cmr12

\setbox0\hbox{\x va}
\showbox0

abc \box0

\bye

could someone without a totally hacked version of luatex compare the above with pdftex and luatex???

@JosephWright with pdftex or luatex without the callback I get a kern \kern-0.65277 but with the callback I get \kern-0.32639

@DavidCarlisle Give me a mo: I'll take a look

@DavidCarlisle See the same as you with stock TL'15

@JosephWright so what was happening was that in most cases the test were using preloaded fonts so the lua code was doing the right thing and returning the same name as pdftex, but for new fonts the lua interface "default font loader" isn't doing the same things as the actual default font loader. Another email to luatex list...

@DavidCarlisle Odd

@JosephWright yes although no-longer surprising

@DavidCarlisle Certainly

@DavidCarlisle Will be getting laptop back later on: plan to get everything this evening :-)

New bigger HD, not that I need the space (next time I get a laptop I'm going for a SSD)

@DavidCarlisle Nothing from HO yet: we may need to go with a plan B

@JosephWright yes I got SSD on this one although I wish I'd fought for a larger one, prices were coming down but I went from 500Mb real disk to 250GB SSD so it's faster (a lot faster when generating thousands of small files in a single run, which I do quite regularly) but if I keep cloning github I'll run out of space:-)

@JosephWright like the plan in luatex85.sty that defined ver@zzzz for lots of common packages :( ?

@DavidCarlisle I'm thinking of a 256Gb: my disk usage is below that at present including six TL versions, a full Windows VM and a coy of all of my CDs

@DavidCarlisle I don't actually have a plan B yet ;-)

@DavidCarlisle We may be forced to to that, yes

@DavidCarlisle Perhaps we could do something on pdftex.def

@JosephWright where is he, do you know, we could send Frank to camp outside his door

@JosephWright I could ask for it back, for example.

@DavidCarlisle Yes, that was my plan B in that area

@JosephWright apparently I have 29Gb free on this disk

@DavidCarlisle About 4 hours drive for Frank I think

@JosephWright on his bike?

@DavidCarlisle :-)

Hmm, Germany so perhaps less than 4 hours

@DavidCarlisle This is true

@DavidCarlisle You know something I don't?

@JosephWright oh frank doesn't drive

@DavidCarlisle OK, you did know something I didn't know :-)

@UlrikeFischer yes, which is why in a few hundred files in the latex test suite, just one hits the different kerns, and there the log is totally different (as you get overfull box messages) so spotting that the problem was kerns and only in non-preloaded fonts, took a while:-)

Have people seen the lasted c.t.t. post from JF?

@JosephWright I suppose I'd better look...

@DavidCarlisle Background is particularly good

Background
==========
1. Don Knuth finished version 1.0 of TeX in November 1983.
2. At about the same time the TeX Users Group (TUG) was established.
3. TeX became widely used for technical typesetting, by both authors and publishers.

@JosephWright typical post, just a list of pointless statements:-(

@DavidCarlisle You might very well think that. I couldn't possibly comment. ;-)

Hullo all

Another simple TikZ question: I've got some vertices of a regular n-gon. Now I would like to draw lines between those vertices. My code so far is

(copied from here)

That same page has some more elaborate answer that would probably enable me to do what I want but it uses "asymptote"

I'd like to avoid "heavy machinery" if possible.

For instance, if I were able to label each vertex and simply put \draw (vertex k) -- (vertex k+1) in my forall loop, I'd be very happy

This should probably be possible, but I don't know how to find out how.

@JosephWright None of my business. :P I just told where the problem is. :)

Any pointers as to where to find something like this would also be very welcome

@Danu (180:2cm) is just another coordinate, and I guess you know how to draw a line from one coordinate to the other. \draw (180:2cm) -- (0:2cm);

@Danu That said, did you try my answer to that question? (Which draws the polygon.)

@TorbjørnT. The subtlety is that I only want 8 out of 12 vertices (and faces) of the 12-gon.

Perhaps there is a better way to do that, though?

@TorbjørnT. I'm trying something like \draw (\a:2cm) -- (\a - 30:2cm) now, but it doesn't seem to work.

\documentclass[border=2mm]{standalone}
\usepackage{tikz}
\usetikzlibrary{shapes.geometric}
\begin{document}

\begin{tikzpicture}
% create the node
\node[draw=none,minimum size=2cm,regular polygon,regular polygon sides=12] (a) {};

% draw a black dot in each vertex
\foreach \x in {1,2,...,8}
  \fill (a.corner \x) circle[radius=2pt];
\foreach [evaluate=\x as \y using int(\x+1)] \x in {1,2,...,7}
  \draw (a.corner \x) -- (a.corner \y);

\end{tikzpicture}
\end{document}

Oh, jk, it is working!

@TorbjørnT. Thanks for your reply. Everything is working as wanted now, though :)

@Danu See above for a quick modification of my code. (I misspoke above, of course it doesn't draw the sides, due to draw=none.)

Now, I'm trying to put arrows on my lines, but they're not appearing :(

The markings are not appearing :\

I guess my mistake must be in this part:

It's not giving me any errors, though... Just no arrows :\

@Danu You have only set up the decoration, not applied it. Use \draw[decoration={markings, mark=at position 0.5 with {\arrow{>}}},postaction={decorate}] ...

@TorbjørnT. Oh, damn that was stupid!

Thanks a lot :)

@DavidCarlisle I found the reason why only va is affected and posted it on the luatex list.

@DavidCarlisle Reading all of UnicodeData.txt in XeTeX/LuaTeX seems to take about three seconds: I'm wondering if this might actually be the best plan (once my laptop is mended I can time properly)

@UlrikeFischer yes thanks I had been wondering if I should dig out tftopl and look for patterns, I'm sure you are right. So another luatex bug found by the latex test suite then:-)

@DavidCarlisle But I have been wondering why va has a different kern to wa. And what the rules for grouping of lig-info is. It looks actually quite logical that a second setting overwrites the first.

@UlrikeFischer texdoc tftopl sort of documents it, I think the label command starts a new program anyway so you shouldn't need an explicit stop (but whatever the manual says a tfm reader that can't read the cm tfm files the same way TeX does has got problems:-)

I was just reading that:-)

@DavidCarlisle Sorry

@DavidCarlisle sourceforge.net/p/xetex/code/ci/master/tree/source/texk/web2c/… has a comment at the top that makes sense

@DavidCarlisle label doesn't start a complete new programm, in the section I copied k, v, and w all use the kern values. But looking at the other section I think that the pdftex handling is as intended. A label should use the first kern value for a char it encounters and ignore following settings.

@UlrikeFischer a ligature lookup always goes "forwards" so that's a list for v (just specifying the following letter in each case, then there is the list for w th efact that both lists have an a shouldn't make any difference

@DavidCarlisle \XeTeXmathcode = (7 1 #) for BMP letters, (0 1 #) for SMP

@JosephWright do you mean letters or all of the BMP?

@DavidCarlisle This applies to all letters

@DavidCarlisle Anything of class L<x> in Unicode

@DavidCarlisle Was checking my script, so back to the Perl original

@DavidCarlisle I'm leaning toward telling Karl we should just parse the original data directly, so am checking the logic. I need to comment my hacked-up loader.

@JosephWright I suppose so, making it var sort of implies that you think that most slots have a font that has the full range which may not be as true as it was in classic tex, but then making them all 0 implies you only have one font at all, so not clearly better

@JosephWright also logically why split on the BMP/SMP range? If they add a new alphabetic script then it's likely to go in a higher plane just for space reasons.

@JosephWright better to make all things of category Lx var unless they have mathclass A (from the MathClass.txt file currently in TR25 but planned to be added to UCD at some point)

@DavidCarlisle Ah

@JosephWright unicode.org/Public/math/revision-14

@DavidCarlisle Yes I'd found that :-)

@DavidCarlisle Short file, easy to parse

@DavidCarlisle Will add to my 'to do' list for this

@DavidCarlisle Good job someone is interested in this, hey ;-)

@DavidCarlisle XML files are not definitive (plus I'm not keen on trying to parse them)

@DavidCarlisle sourceforge.net/p/xetex/code/ci/…

@JosephWright that one is definitive for HTML and MathML ('cause I say so:-)

@DavidCarlisle Well OK but there is a health warning on the Unicode website about the XML versions

@DavidCarlisle I want to see if I can pick up the two special cases for \sffcode based on their Unicode class: I think I can :-) Then I'll look at the maths stuff!

@JosephWright not their XML, mine (which is older and better!): github.com/w3c/xml-entities

@DavidCarlisle I'm thinking a series of loaders, one for each 'area' with the intention they can be loaded v. early by any L3 format (before the renaming) so that everything can be shared with plain/2e: sensible?

@DavidCarlisle Ah

@DavidCarlisle You want to parse the XML in TeX for me? ;-)

@JosephWright yes but 2007 was a while ago, the math alphabets were probably the on;y relevant characters in plane 1 at that point, but now they are adding new human language scripts in whatever plane has space.

@DavidCarlisle I hope Chris, Karl, ... appreciate the effort going in here :-)

@JosephWright I have a parser to hand...

@DavidCarlisle :-)

@JosephWright yes

@DavidCarlisle Yes, makes sense. The Perl script has not had a lot of TLC in recent years.

@DavidCarlisle Everything automated and documented, of course

Men with diggers arrived in village (my spies tell me....)

@DavidCarlisle Just spotted that if not writing everything back to disk I don't need to convert numbers to hex: makes the code a lot faster!

@JosephWright would it be too bad to suggest \@empty as the Unicodedata.txt version comand?

@DavidCarlisle ?

@DavidCarlisle You mean \expandafter\def\csname ver@Unicodedata.txt\endcsname{}?

@JosephWright much more sensible than my suggestion!

@DavidCarlisle Ah, now seen @ArthurReutenauer's mail

@JosephWright oh I guessed you'd seen it between your above two comments, seems you got the context without seeing it:-)

@UlrikeFischer looks like you read it right, thanks:-) I'll have to to read it again more slowly one evening:-) so suppose that's two more passes in the test suite once that gets to texlive svn and I rebuild everything....

@DavidCarlisle Hm. Not quite sure if I got the meaning. If I did understand Hans correctly, you were right to suspect a luatex font loader bug (and I while I wasn't sure at the start I do agree now too). But the larger negative kerning of the va still remains curious.

@UlrikeFischer I misread the tftopl output, so you were correct that the issue was two kerns being specified with tex using the first and lua using the second. The choice of kerns in the cm font was specified on stone tablets and not for mortals to query, even if they are illogical they could only be fixed in lm or some such cm is not going to change its metrics ever.

@DavidCarlisle

@UlrikeFischer exactly!

@DavidCarlisle github.com/latex3/unicode-data: thoughts welcome!

@DavidCarlisle Probably I should add a version to the file itself ...

@UlrikeFischer The one from Mel Brooks' movie is missing (where Moses drops one of the tables, so reducing the commandments from 15 to 10).

@JosephWright "aht" <<. Not bad, only 146 lines. Possibly you should do an ifeof test on stream 0 before opening it, just in case

@JosephWright and perhaps \closein although not sure that makes any difference after you have read to eof.

@wget @ isn't a letter there (but use the questions on site not external links)

@wget error message clearly will show that it only read as far as \key not \key@ifundefined

Indeed. Thanks @DavidCarlisle.

@wget you are missing a gazillion % from ends of lines, as well:-)

Laptop is back!

@DavidCarlisle Ah, good point

@DavidCarlisle Give me a couple of hours to restore everything: outside the Apple shop at the mo making sure the fix is OK!

@DavidCarlisle It is in the L3 repo, you could change it ;-)

@DavidCarlisle Hmm, yes, could do that

Couldn't have done it without you, TeX - LaTeX! <3

TIKZ I assume?

Yes, all TikZ.

Some Unix advice from anyone? What's the best way to copy an entire directory tree preserving ownership, etc.? On Windows I'd used xcopy. (Context: restoring TeX Live from my external HD so the directories are owned by root.)

Ooh, new JabRef icons

oh we're back?

@JosephWright cp -r -p (r for recurse down p for preserve attributes)

@DavidCarlisle Ah

@JosephWright or the old way was tar and untar

@DavidCarlisle So if I'm in /usr/local/texlive and I want to copy [backup disk]/usr/local/texlive/2014 I do cp -r -p [path]/2014/ .?

@JosephWright yes it "tries" to preserve ownership and other attributes

@DavidCarlisle OK, will do that once the current bacth of copying is done

I did MactTeX 2015 'fresh' as it includes GhostScript, etc.

@Danu Hehe... tori ... a great plural :P PS: Looks good :D

@BMWurm But...that's the correct plural, isn't it? :P

@Danu it's one of the correct plurals I think...

In any case, it's the historically correct one :)

@Danu historically correct is also "thou shalt not" :)

@yo' Thou shalt not maketh fun of my justification!

@Danu I think that you've just used the 2nd person singular past tense in place of the infinitive :)

Thou shalleth nothe judhtehaml;iodfjlkfaj;lajkfsn;kajdsgj'woe mia

@Danu It certainly is the best one :P although even the OED allows for toruses (oxforddictionaries.com/definition/english/torus -- which just looks wrong in my eyes)

@BMWurm -- you could always fudge the situation by referring to "toroidal objects". (my first reaction when i saw "tori" was an image of "torii", which is quite a different object.)

@barbarabeeton which made me question what the plural of "torii" would be... (it's torii btw), strange coincidence though, both look like food... "tori" looks like donuts, and "torii" looks like $\Pi$ ;)

@BMWurm -- ^^^^^

@barbarabeeton A nice joke ;-) I fear, I can't use it for my German students, since they probably don't know the meaning of 'circumference' ;-)

@barbarabeeton hi I was just about to mail you! (mail coming up tr25 complaints as usual:-)

@barbarabeeton funny :-) btw, slightly related:

@BMWurm toriis I would say. There is no reason to treat it differently than other Japanese words ending in i, such as Kawasakis (motorbikes).

@barbarabeeton Congratulations btw, you make an important contribution to the community!

The ideal xmas gift for @PauloCereda and @DavidCarlisle: