12:19 AM
@mickep -- Okay; got an answer. There's no definitive documentation; this conjecture is from one of the production editors who was there when the change was made. The amsthm definitions were created to parallel the existing definition in the basic article et al. It's likely that it was felt that the small caps heading, with or without extra space above/below the enunciation, didn't stand out enough, so boldface and left-aligned, with space above, would be more obvious.

8 hours later…
7:50 AM
@marmot It seems that you missed 121,212

@DavidCarlisle can you use texindy in cygwin? (texindy some.idx, if it dies it should die fast with  not a symlink as required for TeX Live)

it does something:
$texindy lb1.idx Opening logfile "/dev/null" (done) Reading indexstyle... Loading module "/tmp/02WH8gW3Iy"... Loading module "lang/general/latin9-lang.xdy"... Loading module "lang/general/latin9.xdy"... Finished loading module "lang/general/latin9.xdy". Finished loading module "lang/general/latin9-lang.xdy". Loading module "tex/inputenc/latin.xdy"... Finished loading module "tex/inputenc/latin.xdy". Loading module "/usr/local/texlive/2019/texmf-dist/xindy/modules/base/texindy.xdy"... @JosephWright I think it's right (but would need to trace) vsize 72 sub is subtracting an inch from vsize (which as far as I can see is stored using bp units, tex.lpro initialises it to /vsize 11 72 mul N ie 11in. Probably your perl reports the correct$OSNAME then, if a perl from a msys bash is used it dies, as this reports "msys" and not "MSWIN".

8:10 AM
@DavidCarlisle your does picture mode work reminds me of Ebenezer Scrooge

@UlrikeFischer well, A Christmas Carol is also art :-)

@yo' like this one?
\let~\catcode~76~A13~F1~j00~P2jdefA71F~7113jdefPALLF
PA''FwPA;;FPAZZFLaLPA//71F71iPAHHFLPAzzFenPASSFthP;AFevP
AGGFRruoPAqq71.72.F717271PAYY7172F727171PA??Fi*LmPA&&71jfi
Fjfi71PAVVFjbigskipRPWGAUU71727374 75,76Fjpar71727375Djifx
RrhC?yLRurtKFeLPFovPgaTLtReRomL;PABB71 72,73:Fjif.73.jelse
B73:jfiXF71PU71 72,73:PWs;AMM71F71diPAJJFRdriPAQQFRsreLPAI
I71Fo71dPA!!FRgiePBt'el@ lTLqdrYmu.Q.,Ke;vz vzLqpip.Q.,tz;

@DavidCarlisle that one wouldn't have the capital A I suppose...

@yo' well it has, but it's catcode 13

8:18 AM
@DavidCarlisle then it doesn't count I'm afraid

8:34 AM
Hey guys, I'm creating a class to set general characteristics for CVs. As a start I'm following overleaf.com/learn/latex/… tutorial
Is there any advice you'd recommend to set the class file or any related issue?
Right now, the class file looks like this:
\NeedsTeXFormat{LaTeX2e}
\ProvidesClass{santi}[2019/04/26 Custom CV class]
\RequirePackage{titlesec}

@DavidCarlisle Ah, cool
@DavidCarlisle Have it working
@DavidCarlisle, @UlrikeFischer Need to deal with the link border stuff in dvips: have to work out if that should be an auxiliary (a no-op for other drivers, apply the border patch in dvips)

8:51 AM
@JosephWright I can probably look only tomorrow, we have to go to another hockey play this evening.

\documentclass{article}
\usepackage{expl3}
\begin{document}

Some text
\ExplSyntaxOn
\box_move_up:nn { \baselineskip }
{ \hbox:n {\driver_pdf_destination:nn { foo } { 200 } } }
\ExplSyntaxOff
with a target which we are going to make longer so I
can actually see where it lands.

\vspace*{5em}

Some more text
\ExplSyntaxOn
{ }
{ foo }
text
\ExplSyntaxOff

\showoutput

\end{document}
@DavidCarlisle, @UlrikeFischer ^^^ Should be working :)
@UlrikeFischer Cool. In hyperref this is easy as the individual dict entries are separate But we can't do that, so either I have to 'fish out' /Border[a b c d] or we have something like \driver_pdf_link_border:nnnn{a}{b}{c}{d}

@JosephWright hm, not sure if I understand. You mean with pdflatex one could put /Border in the attribute argument but with dvips this doesn't work?

@UlrikeFischer Yes: look at /BorderArrayPatch in hdvips.def: there is a need to convert units
@UlrikeFischer i'm trying to make sure I pull out all of the 'bits and pieces' from E.g. hyperref

9:07 AM
@JosephWright I don't understand why \Hy@BorderArrayPatch is set to empty in pdfmark and then to BorderArrayPatch in hvips ;-(.

1 hour later…
10:37 AM
Is it possible to insert a vertical line between columns in column environment?
Or at least to increase the separation between columns? I've wandered on the web with no good results

10:56 AM
@santimirandarp Change the length of \tabcolsep, perhaps?

11:08 AM
@santimirandarp Do you mean in beamer?

See how carefully some writers make their notes. In this autograph volume Mozart endeavoured to record the title, date and opening bars of every composition over seven years, including some that were subsequently lost. #MakingYourMark
@egreg ^^ :)

@PauloCereda Always comment your code!
3

@egreg :)

@santimirandarp Something like this?
\documentclass{beamer}
\usetheme{Warsaw}

\begin{document}

\begin{frame}
\frametitle{Test}

\begin{columns}

\begin{column}{1in}
\makebox[1in]{X\hrulefill X}\\
Abc\\
def\\
ghi
\end{column}

\vrule\hfill

\begin{column}{1in}
\makebox[1in]{X\hrulefill X}\\
Abc\\
def\\
ghi
\end{column}

\end{columns}

\end{frame}

\end{document}

12:03 PM
@egreg yes, thanks very much; I've been looking up in the site and found that maybe \multicols is way much flexible than \columns environment, is this right?
And also there is plenty info in the docs

1 hour later…
1:31 PM
@santimirandarp Yes, but not so much for beamer, where a lot of visual formatting is needed.

2:21 PM
@UlrikeFischer Thinking about /B[....], we could just use a third argument at the driver level ...

@JosephWright ooh a bee

2:43 PM
@JouleV Could be worse. ;-)

@marmot Could be raining

@egreg Not at this time of the year.

3:01 PM
@marmot and you will miss 121,314 :|

@JouleV I am only interested in 123456.

@marmot Ohh
Only one week more ;)

@UlrikeFischer Yes, but your numbers are in German.

1 hour later…
4:13 PM
Yay, having puffed rice cake!
@DavidCarlisle ^^ we ducks can eat those, right?

5:01 PM
@PauloCereda you can eat duck with rice (I assume that's what you meant?)

@DavidCarlisle oh no

2 hours later…
7:20 PM
@JosephWright do you want to comment on this github.com/ho-tex/hyperref/issues/91?

8:19 PM
@UlrikeFischer Yup, will do

8:40 PM
@JosephWright I looked at this border question. If I got it right hyperref writes Border [1 0 1]BorderArrayPatch and the keyword triggers then some recalculation later. Did you check if this patch is still needed?

@UlrikeFischer Yes, I did: it is needed
\documentclass{article}
\usepackage{expl3}
\usepackage{hyperref}
\hypersetup{pdfborder = 0 0 3}
\begin{document}

Some text
\ExplSyntaxOn
\box_move_up:nn { \baselineskip }
{ \hbox:n {\driver_pdf_destination:nn { foo } { fit } } }
\ExplSyntaxOff
with a target which we are going to make longer so I
can actually see where it lands.

\url{tug.org}
\ExplSyntaxOn
\leavevmode
{ /Border[0~0~3]/H/I/C[0~1~1] }
{ /A <</S /URI /URI ( tug.org ) >> }
@UlrikeFischer ^^^

8:58 PM
@DavidCarlisle, @UlrikeFischer What's the binary situation for HarfTeX?
@DavidCarlisle, @UlrikeFischer Ah: W32TeX :)

@JosephWright oh, Akira is really good, I was already wondering ;-).

@UlrikeFischer Now, what do I need to do to build formats ....
@UlrikeFischer I'm going to work on the basis this will get into TL 'real soon now' ...

@JosephWright I don't know, i could build luahbtex on cygwin but haven't tried the new version, been a bit busy lately
@JosephWright "real soon now" has to be tl2020 doesn't it, or do you think they will allow a binary update mid year?

@DavidCarlisle It's not an update ...

@JosephWright Isn't a new binary a "big update" ?

9:09 PM
@DavidCarlisle Yes, but the policy is more that they don't want too much to change: I think we could ask ...

@JosephWright bother. Do you know what the value is? I seem to get similar border with 24 but have no idea where it comes from.

@JosephWright yes, but we should check with Khaled when he's ready to push it for wider use...

@UlrikeFischer It's conversion of inches to pt
@DavidCarlisle Well yes, that's true: I meant more that if it looks good, we could lean on Karl

@JosephWright I'll clone it and run the build script see what happens....

@DavidCarlisle Cool

9:13 PM
@JosephWright is there also a problem is one use a border style directory (BS) instead of Border?

@UlrikeFischer Don't know! Try it with my demo? But we need some way of supporting /Border properly: can't really say 'it might not work, depending on the driver'

@JosephWright build.sh running.....

@JosephWright looks good:
\driver_pdf_link_begin_user:nnw
{ /BS <</Type /Border /W~3>> /H/I/C[0~1~1] }
{ /A <</S /URI /URI ( tug.org ) >> }

@UlrikeFischer Er, not working here
@UlrikeFischer But that still doesn't work if we get passed a /Border array

@JosephWright really? How are you viewing?

9:26 PM
@UlrikeFischer TeXworks (so Poppler)

@JosephWright then probably texworks can't handle the BS dict - not our fault ;-).
@JosephWright probably we should require that a border value is passed with a macro e,g, \driver_pdf_border_array, then one could add the keyword.

@UlrikeFischer Yes, that's my feeling, or we add a separate argument (but that could be an every-expanding list). I'm thinking an expandable function \driver_pdf_border_array:nnn which 'produces a border array suitable for the current driver'
@UlrikeFischer You are the expert on this: where am I supposed to add entries to fmtutil.cnf?

@JosephWright I have a local fmtutil.cnf in texmf-local/web2c and just added
harftex harftex language.dat,language.dat.lua luatex.ini
harflatex harftex language.dat,language.dat.lua lualatex.ini
worked fine: This is HarfTeX, Version 0.1.1 (TeX Live 2019/W32TeX) (format=harflatex 2019.4.26)

@UlrikeFischer Doesn't seem to get picked up by fmtutil-sys ...

@JosephWright perhaps you need to run mktexlsr first. For me fmtutil-sys --byfmt=harflatex worked flawless.

9:41 PM
@JosephWright: To my understanding the /DEST pdfmark expects coordinates in bp, relative to the page origin (0,0) in the lower left page corner. I tried to implement \pdfdest for the usual drivers in pdfbase (gitlab.com/agrahn/media9/raw/pdfdest/pdfbase.sty). Here is a test file with a FitR example (AR preferred):
\documentclass{article}
%\documentclass[dvipdfmx]{article}
\usepackage[a4paper]{geometry}
\usepackage{graphicx}

\usepackage{pdfbase}
\ExplSyntaxOn
\let\pbsPdfDest\pbs_pdfdest:nnnn
\ExplSyntaxOff

\begin{document}

FitR:\pbsPdfDest{myDest1}{fitr}{1}{\fbox{(1) Hello World!}} \pbsPdfDest{myDest2}{fitr}{1}{\rotatebox[origin=bc]{90}{\fbox{(2) Hello World!}}}

@AlexG Yeah, think I have it working :)
@UlrikeFischer texhash it is

@JosephWright I made use of \pdfsavepos and \pdflastxpos/\pdflastypos. Did you use them as well?

@AlexG No ... I'll have to look at your code
@AlexG Sorting HarfTeX at the moment

@JosephWright Oh, what's this?

9:48 PM
@JosephWright error: explicit instantiation of ‘struct std::atomic<icu_63::number::impl::NumberRangeFormatterImpl*>’ in namespace ‘icu_63::number’ (which does not enclose namespace ‘std’)

@DavidCarlisle Huh???

@JosephWright harftex build script dying trying to build ICU :(

@JosephWright Does it produce pdf or dvi?

@AlexG I had to debug a problem with pdfbase a few days ago. A document included a pdf through \AtBeginDocument (not a good idea...) and pdfbase then constantly said Rerun to get internal references right!.

@AlexG It's more-or-less LuaTex: I think both are possible

9:50 PM
hmm google seems to think this is a known feature unicode-org.atlassian.net/browse/…

@AlexG Give me two minutes to check

@JosephWright No rush, going to bed now.

@AlexG it is luatex with harfbuz linked to the executable so it does everything luatex does including building dvi files that are virtually unusable by any normal dvi driver:-)

@DavidCarlisle Phew!

@AlexG Does both

9:54 PM
@DavidCarlisle I got an error about ICU too - but only because I forgot to copy the icudt64.dll ;-).

@KhaledHosny I think I am hitting unicode-org.atlassian.net/browse/…
build.sh ends with
In file included from /home/harftex/source/libs/icu/icu-src/source/i18n/numrange_impl.h:11:0,
from /home/harftex/build/libs/icu/../../../source/libs/icu/icu-src/source/i18n/numrange_fluent.cpp:12:
/home/harftex/source/libs/icu/icu-src/source/i18n/unicode/numberrangeformatter.h:189:33: error: explicit instantiation of ‘struct std::atomic<icu_63::number::impl::NumberRangeFormatterImpl*>’ in namespace ‘icu_63::number’ (which does not enclose namespace ‘std’) [-fpermissive]
template struct U_I18N_API std::atomic<impl::NumberRangeFormatterImpl*>;

@DavidCarlisle I don't understand anything at all from this, but why is your build using icu_63 when my binary needs icu 64?

@UlrikeFischer I just did git clone ...; cd harftex; build.sh don't ask me:-) perhaps Akira updated ICU while building the w32tex version?

@UlrikeFischer, @DavidCarlisle Key point is: at least on Windows, it's pretty trivial to install and test HarfTeX. I've added it to my TeXworks menus, and have been able to check it against the new driver code :)
@UlrikeFischer I think I might add something for shipout support in l3trial. My feeling is we really don't want to code this twice: it should just go directly into the expl3 kernel, we then say 'resource management = require expl3' and try to convince people (mainly @HenriMenke as we control e.g. hyperref) to get on with it.

@JosephWright yep there are also linux and macos binaries on the harftex gh releases page, just hard for people who insist on minority platforms like cygwin...

10:10 PM
@JosephWright yes.

1 hour later…
11:22 PM
@JosephWright I'm happy to adapt to that but there are two prerequisites: (1) The shipout stuff has to be part of the LaTeX format, because loading times of expl3 are absolutely abysmal. (2) There has to be the freedom for some level of abstraction, because PGF/TikZ has to support different formats.

@HenriMenke ? it takes ages to load pgf due to the multiple small files, expl3 on pdftex is nothing compared to that (a bit longer for unicode-texs due to loading the unicode data files)

@DavidCarlisle Exactly, loading PGF takes ages, so I don't want to bloat that even further with exp3.

@HenriMenke (but for latex we should get it in to the format anyway:-)
@HenriMenke concatenate the pgf files and you'd save far more time than you would ever save by not loading expl3, loading lots of small files kills the load time.

@DavidCarlisle Empty document with only \usepackage{tikz}:
real	0m0.586s
user	0m0.409s
sys	0m0.165s
Empty document with only \usepackage{expl3}:
real	0m0.690s
user	0m0.553s
sys	0m0.128s
With LuaLaTeX of course.

with pdftex:
\documentclass{book}
\pdfresettimer
\typeout{===\the\pdfelapsedtime}

\usepackage{tikz}

\typeout{===\the\pdfelapsedtime}

\pdfresettimer
\typeout{===\the\pdfelapsedtime}

\usepackage{expl3}

\typeout{===\the\pdfelapsedtime}

\begin{document}

\end{document}
load time for tikz around 5 times that of expl3...

11:37 PM
@DavidCarlisle When using LuaTeX, loading expl3 in addition to tikz roughly doubles the startup time of the document. I don't want to do that to my users.

@HenriMenke as I say we should get it in to the format (and can possibly do a minimal version for plain, the time is almost all loading the unicode data) for lualatex users they pretty much always load expl3 anyway (for fontspec) so it's not really extra time (but having it the format would solve that anyway)

@DavidCarlisle I will think about concatenating the PGF files, but that will be non-trivial. There might be circular dependencies.

@HenriMenke I just thought it funny that the tikz maintainer of all people should call the expl3 load times "absolutely abysmal" as tikz is by far the slowest package to load that I know of:-)

Really?
\documentclass{book}
\pdfresettimer
\typeout{===\the\pdfelapsedtime}

\usepackage{tikz}

\typeout{===\the\pdfelapsedtime}

\pdfresettimer
\typeout{===\the\pdfelapsedtime}

\usepackage{expl3}

\typeout{===\the\pdfelapsedtime}

\pdfresettimer
\typeout{===\the\pdfelapsedtime}

\usepackage{pst-all}

\typeout{===\the\pdfelapsedtime}

\begin{document}

\end{document}
@DavidCarlisle ^^^^

@marmot pstricks must have got bigger since I last used it:-) (but still just goes to show expl3 load time negligible...

11:47 PM
@DavidCarlisle On my MacBook I get tikz: ===12959, expl3: ===16638 and pstricks: ===41744.
I use latex (obviously pdflatex won't work with pstricks)

@marmot I get 13868 for tikz, 11042 for expl3 and 20567 for pst-all

@marmot expl3 looks very wrong I get from your file:
$latex bb611 | grep === ===0 ===135708 ===0 ===29145 ===0 ===175656 @DavidCarlisle You should tell @JosephWright and @egreg that expl3 looks wrong, not me. ;-) @DavidCarlisle There is a very simple explanation: my computer is better trained at loading tikz and yours better at loading expl3. ;-) \documentclass{book} \directlua{socket = require("socket")} \directlua{start = socket.gettime()} \usepackage{tikz} \directlua{io.stderr:write(socket.gettime() - start, "\string\n")} \directlua{start = socket.gettime()} \usepackage{expl3} \directlua{io.stderr:write(socket.gettime() - start, "\string\n")} \directlua{start = socket.gettime()} \usepackage{pst-all} \directlua{io.stderr:write(socket.gettime() - start, "\string\n")} \begin{document} \end{document} $ lualatex test.tex > /dev/null
0.14264607429504
0.32327008247375
0.37016105651855

@HenriMenke hmm I get:
$lualatex ltest > /dev/null 1.152636051178 0.94291400909424 1.8418300151825 11:56 PM @DavidCarlisle I still have TeX Live 2018 on this machine. If you're loading plain tikz shouldn't you load plain pstricks as well? Loading all libraries seems unfair :-) @PhelypeOleinik Pssst! ;-) yes but it's 10 times faster than mine to load tikz anyway:-) (mine's not the fastest machine but it does have ssd shouldn't be )that_ slow :( Just tried on a machine with TeX Live 2019 and a much faster CPU and I get $ lualatex test.tex > /dev/null
0.09126091003418
0.39997220039368
0.22430205345154
However this machine has a HDD, my laptop has a SSD.

@marmot Oops :-)