@PauloCereda

@AlanMunn -- makes one wonder where hawaiian would be found. far away from russian and czech, i predict. and how about welsh?

@barbarabeeton Well Hawaiian would pattern with Japanese I would think. It's a pretty strongly CV language. I'm assuming Russian and Czech both pattern with Polish.

@AlanMunn -- to return to our earlier discussion about pronunciation of "...ease" vs. "...eaze", contrast "ease" and "please" with 'cease". "crease" and "lease". and i found a "-y" word where it's attached to a verb, not a noun: "clingy".

@AlanMunn -- that was my guess too, when i was at the tug meeting at bachotek last spring, i found that polish cognates to russian often enough to not sound too unfamiliar to me. (the meanings may, of course, be enough different to get me in trouble, but from a sound point of view, i wasn't totally baffled. czech is another matter, however; i would still like to get some coaching from @yo'.)

@barbarabeeton It occurs to me that it may be a false etymological connection on my part, since 'house' is Germanic, but 'please', 'grease', 'cease' are borrowed from Old French. And their /s/ /z/ distinction comes from there: 'please' < 'plaisir' (z) but 'grease' < 'graisse' (s).

@barbarabeeton Yes, but here we're talking about pure phonological properties, which at this level of granularity are probably the same.

@AlanMunn -- very true. i ignored etymology. there is probably a similar dichotomy with "-ing" words, both verbs and nouns.

@AlanMunn -- does one assume that this is a barking duck?

@barbarabeeton I think this is a sign written by a German.

@AlanMunn -- or "validated" by a typical automated spell checker.

@barbarabeeton To clarify: since German devoices word final stops, 'dog' and 'duck' would sound pretty similar, and Germans often spell English voiced stops as voiceless ones. And the vowels are very similar too, since ʌ is pretty back.

@AlanMunn -- well this at least partially explains why frank mittelbach complains that my german accent is execrable,

@barbarabeeton :)

@StefanKottwitz @PauloCereda texdoc.net is outdated. Please update to TL2017.

Early quack!

@HenriMenke oh

@AlanMunn LOL

@AlanMunn: I was wondering if a typical duck would enjoy Metallica. :)

Must dash, shaky bus time! You guys behave. :)

@barbarabeeton Russian and Polish sound very similar as they moreorless use the same system of vowels and stresses. The vocabulary differences could probably give you some tough (and even embarrassing) times.

@PauloCereda Yes, Sir!

@PauloCereda Shaky bus time already? It must be the middle of the night!

idownvotedbecau.se

^^^ This one is quite useful.

@HenriMenke it needs idownvotedbecau.se/therearenoducks

@DavidCarlisle Just open an issue github.com/WillSullivan/IDownvotedBecause/issues/new

@HenriMenke I need to preserve my real reputation:-)

@HaraldHanche-Olsen Very early in the morning. :)

@Mico please undelete your answer with the lua patterns. I think I found a good one which does the trick:

function myreplace ( s )
  s = unicode.utf8.gsub ( s, "E", "Ẹ" )
  s = unicode.utf8.gsub ( s, "e", "ẹ" )
  s = unicode.utf8.gsub ( s , "(\\[ẹẸ%a]+)" ,
        function ( x )
           x = unicode.utf8.gsub ( x , "ẹ" , "e" )
           x = unicode.utf8.gsub ( x , "Ẹ" , "E" )
           return ( x )
        end )
  tex.sprint ( s )
end

@PauloCereda: Another Must - Have - Duck for you?

@ChristianHupfer Risikoduck?

@egreg Perhaps ... or James Bond -- The World Is Not Enough - Duck ;-)

This is a tough question: tex.stackexchange.com/q/403297/2693.

@AlanMunn You know someone is in trouble when you read lost LaTeX source.

@AlanMunn yep. \tikzmark comes to my mind: mark a position in each line, mark the left stop, and then combine the x and y coordinates of them. Finally, enumerate the lines somehow. But this means digging deep into the package source code.

@yo' Right, except there's no concept of 'line' in the package code...

@AlanMunn yeah, right.

What are the consequences of grouping a single (active by utf8x) character (which doesn't form ligatures) outside of math-mode? Are there any?

Hi

I think @DavidCarlisle have produced a tool to check backward compatibility basing of tracing. Do you remember the name/where I could found data?

@Maïeul more Joseph than me these days (in its current form) but I think you mean l3build

@Skillmon no kerns either?

@Skillmon why utf8x rather than utf8 ?

@DavidCarlisle thanks. That is. I am not sur to understand well how it test regression on box output

but in any case, it work only one page by one page

and I need to test multiple pages output

so I would stay in my technical to export the pdf in png, and then check the png

@Maïeul the usual mode compares whatever boxes you show in the log whole pages (if you use \showoutput) or more commonly specific boxes shown with \showbox, it also has a mode to compare pdfs (but that's a bit more fragile)

ho

because in my case, the position of page breaking is cruccial

oh, it check binary pdf file

is it possible?

@Maïeul yes sure so if you use \showoutput then all line and page breaking is logged

ok

@Maïeul it's actually easier to compare 2 PDFs as images. You only have to count for possibly different anti-aliasing.

yo: that is what I do

but It could be very slow

@Maïeul yes it uses the (newish, texlive 2016 or later) epoch settings so that tex makes reproducible pdf without the build time being internally used in identifier then it can check that the pdf is bitwise identical.

oh

Nice

@Maïeul but \showbox logging is more understandable and useful in most cases.

it works for all engines?

@Maïeul yes (all ones in texlive: etex,pdftex,xetex,ptex,luatex etc)

nice

because if I could avoid to produce image diff, it could be faster

I will look on this

@Maïeul \showbox output (the traditional regression testing we've used for latex releases since the 1980s) has the advantage that if it fails it is easy to see what the difference is. binary comparison if they fail just really tell you that something changed, not what or why.

I understand

@DavidCarlisle I know about this question, but does the L3 team follow any standard?

or do you guys just all use your own style

a last question about l3build: it run one time the file, or it run how many time it need to get stable file (like latexmk)?

@StrongBad for 2e code pretty much our own style, within the confines of doc package markup. for l3 code we currently "discussing" how to evolve the current l3doc into something more usable....

@Maïeul it runs a fixed number of times settable in the configuration for the test (twice by default as I recall)

ok!

I will look on it a day.

But for now i have really to work (but latex is more funny than my job)

@DavidCarlisle so the Verna 2011 TUG paper on best practices has been slow to be adopted/modified/updated.

@StrongBad not sure I'd ever seen it:-) It was just a talk giving some personal suggestion as far as I can see from the abstract?

@DavidCarlisle checkruns=1

hey @Maïeul, you maintain reledmac, right?

yes

or I try ;)

would you have interest in code for two column paragraph footnotes?

is not you whick ask for this some month ago?

which>who

don't think so

but i did work it out, and saw that in theory you were working on it too

now, I am not working on it

someone have open a branch

and work on it

but it see some problem

So if you have a a "out of the box" code, I take it ;-)

github.com/maieul/ledmac/issues/655

i would need to know how you handle splitting between pages

we cant

that is a fundamental problem with the mechanism use for paragraphed footnote: a footnote can't be splitted

in my package it can

Ï tried some technical, but that didn't work

how?

really?

you was able to produce paragraphed footntoe which splitt in two page in the middle of a footnote?

yeah, in the output routine I measure how much space is left on the page, and if it won't fit, it splits the note

I'll post a picture of it, give me a sec

what is your package=

?

github.com/rabbiveesh/dfn-para

but it won't necessarily work with more than one set of footnotes

a yes

that is maybe the problem

i mean, the only issue is splitting between pages

but that's the next thing I'm gonna work on

I think the problem is that the "normal" vsplitting of footnotes made by TeX is not possible as we have a great great hbox, and so we can do splitting only in the output routine

but at this stade, is difficule

I don't remember exactly why

But I did a branch

I ask for user test

and it was wrong with real case use

very nice

the problem with splitting in the output routine is that you then have to reflow the box

exactly

at the last page

half of what i worked on is ways to transparently reflow boxes

sorry

I really must work

but we coudl continy be email

just thought you'd be interested

yes, of course

but after that, it need to be adapted to all the specificitys of reledmac

which are not eassy

I would like to get one year to work fully on reledmac

I must go

thansk a lot

hey @DavidCarlisle, are there packages out that provide macros for reflowing boxes?

guess not...

@AGoldMan basically you can't

You can by unvboxing and unhboxing, so long as you control the input somewhat

@AGoldMan well not really, you can't reconstruct hyphenated words, if there are any rules or colour or writes etc you can't dissasemble the box at all, what does work is to initially save the text as a macro or as a single hbox and then reflow it in different ways.

@DavidCarlisle if you only allow hyphenation the last time you reflow it, the hyphen problem doesn't come up. And you only have to unhbox each line, so the presence of writes won't matter

color is something I have to work on

@AGoldMan if there are any write nodes in the vertical list you can't access the lines before that via \lastbox, but if you are in a position to change the hyphenation in advance you can avoid the problem and save the whole thing as a single hbox (Frank has some code somewhere that does more saving each paragraph full width but still using vmode to save any explicit vspace, display math, etc. But either way you need to save something re-usable not just a normal typeset box.

@DavidCarlisle I'm talking about using an hbox or vbox for reflowing

i was just wondering if someone else did work like this before, or if I have to write everything myself

@AGoldMan sure but what I mean is you need to grab the original text in some kind of structure designed for reflowing, you can't take a normal typeset vbox and reflow it in any reasonable number of cases.

@AGoldMan I'll give some pointers to latex3 code in the general area (if I can find it:-)

@yo' I figured out a semi-automatic way.

@DavidCarlisle it's for an answer of mine (this one). The used character is defined in utf8x not in utf8. The pdflatex would be more robust if I'd replace the character with {\cA(Ẹ)} because otherwise stuff like \textbf e would throw an error, but I don't know whether this would probably break too many other things.

@Skillmon using utf8x is a rather invasive way to avoid adding a couple of extra utf8 character definitions:-) do you never have any inter-letter kerns between your accented letters and the following letter?

@DavidCarlisle honestly I didn't know the correct definitions and was too lazy to look them up. Don't know whether OP is concerned about inter word kerning. I think I'll leave the answer as is (since the OP was referring to the usage of XeLaTeX anyway...)

\usepackage[utf8]{inputenc}
\DeclareUnicodeCharacter{1EB8}{\d{E}}
\DeclareUnicodeCharacter{1EB9}{\d{e}}

@Skillmon ^^ :-)

@AlanMunn cool!

@yo' I'd be curious if anyone could figure out a way to do it totally automatically.

@DavidCarlisle thx. It still doesn't fix the issue with \myreplace{\textbf e} I had there...

@Skillmon no

@Skillmon the official line is that the braces around the argument of \textbf are not optional, sometimes of course that rule need not be enforced but here I think enforcing that is better than always bracing the letter

@DavidCarlisle the texlive pdflatex man page says there are better ways than setting the environment variables. It then says RTFM, I would rather ask you. Any idea what the better way is?

@StrongBad hmm l3build sets the enviornment (which is where I jut grabbed the settings from) Just had a quick look at the pdftex manual, (texdoc pdftex) it doesn't suggest an alternative

@DavidCarlisle I think it is here: tex.stackexchange.com/questions/229605/…

they added new primatives

@StrongBad ah yes sounds vaguely familiar:-)

@DavidCarlisle it looks like those new primitives may have cost you a green tick.

@StrongBad the world is unfair, but I'm used to that by now.

@StrongBad it's changed again since texlive 2017 the note in the accepted answer that even if you set the epoch environment that the filepath gets used in the internal id so not reproducible if built in a different place is no longer true in texlive development svn (a bug report from debian I think which tests all their software for reproducibility and asked for a change here)

@DavidCarlisle yeah I saw that. Depending on how the system works, I bet I could figure out the identities of a fair number of academic reviewers.

@DavidCarlisle 'Joseph decides on something and cleans everything up ...'

@StrongBad No, has to be done from the environment (this stuff all comes from the reproducible build community, and they want the environment to be 'known')

@JosephWright but you can do it from either place. It is nice from a build perspective that it is reproducible without changing the source. The new primitives seem nice since they give a little more control.

@StrongBad dvips is still tricky

@StrongBad At some stage I plan to revise the l3build set up to work more smoothly: first, though, I have some feature requests from the rest of the team to address