Hey@tjt263

ok cool!

In the mean time I am wondering what this code snippet does

\pgfsetbuttcap
{
\definecolor{dialinecolor}{rgb}{1.000000, 0.000000, 0.000000}
\pgfsetfillcolor{dialinecolor}
% was here!!!
\pgfsetarrowsend{stealth}
\definecolor{dialinecolor}{rgb}{1.000000, 0.000000, 0.000000}
\pgfsetstrokecolor{dialinecolor}
\draw (38.400000\du,7.500000\du)--(38.350000\du,24.400000\du);
}

i don't know what a buttcap is. i mean i can kind of imagine it's like a terminator. something that occurs at the end of something else. like an arrowhead or a cap at the end of a line or something

hmm I see

@tjt263 I am trying to learn the syntax of tikz a bit, trying to draw straightlines, like I am trying figures, but they go outside of the pdf page, i want to shrink them any command to insert in the code?

that way i think you should be able to go larger than a regular piece of paper

\documentclass[tikz,border=3.14mm]{standalone}
\makeatletter% from tex.stackexchange.com/a/39698/121799
\def\grd@save@target#1{%
  \def\grd@target{#1}}
\def\grd@save@start#1{%
  \def\grd@start{#1}}
\tikzset{
  labeled grid/.style={
    to path={%
      \pgfextra{%
        \edef\grd@@target{(\tikztotarget)}%
        \tikz@scan@one@point\grd@save@target\grd@@target\relax
        \edef\grd@@start{(\tikztostart)}%
        \tikz@scan@one@point\grd@save@start\grd@@start\relax
        \draw[minor help lines] (\tikztostart) grid (\tikztotarget);

try this

it's not perfect, but it might help you get the hang of it

you can also simply use smaller units of measurement, the result will be scaleable anyway so it doesn't mattter

yup cool

\begin{tikzpicture}[thick,scale=0.6, every node/.style={scale=1.2}]

yeah this is just a thing to show the grid

it didn't work properly, but it's still useful kinda

yup now i m into a bit of creating figures though

so getting to learn tikz as i se by having in mind the coordinates set and u can do what u imagine!

@BAYMAX yes, that's what i use it for. that's why i said it..

i wouldn't lead you astray

if you look through my questions, you will see i usually include a basic example of what i'm trying to draw. then the answers are usually a more complicated version. reading through them might help you advance to my level quickly.

not that i am an expert, but i have made significant progress

oh cool!

I will try though1

thanks!!

@DavidCarlisle You mean the hyphenation? the first nonBAstring? I see that too, that must be a side effect from figure 5.7, where the manual sets preex- and postexhyphenchar, this are global setting and they seem to have forgotten to set them to 0 again.

@UlrikeFischer yes:-)

@DavidCarlisle the test uses filecontents*, the problem is not in the file, but the file date itself, attachfile2 gets it and adds it to the file object.

@UlrikeFischer yes that occurred to me after I went to sleep last night:-) not sure how that can be avoided (I don't think git preserves filedates does it?) unless we get l3build to normalize the date....

@DavidCarlisle in one of the embedfile tests there is \def\pdf@filemoddate#1{}, imho I added this for the same reason.

@UlrikeFischer sounds like a plan....

@DavidCarlisle I just tried it, it looks okay, the pdf not longer has a mod date. I will update the test.

@UlrikeFischer thanks, push straight to master I think

@UlrikeFischer I'm rebuilding normal formats....

@DavidCarlisle yes, I only made a branch yesterday as there were so many changes.

@UlrikeFischer I thought it was to give me some practice in rebasing:-)

@DavidCarlisle ;-) I could need some practice here too - I never know in which branch I should be when rebasing.

@DavidCarlisle ooh is it secret?

@PauloCereda no, @UlrikeFischer broke something, so just business as usual.

@DavidCarlisle ooh

@DavidCarlisle, @UlrikeFischer Karl's working on latex-dev: Frank and I had a mail ...

@JosephWright today's update added the utf8-command line option to texmf.cnf.

@UlrikeFischer OOh

@UlrikeFischer ooh

@UlrikeFischer Cool: so LaTeX is now (or will be from October) more-or-less UTF-8 everywhere

@JosephWright yay

@UlrikeFischer Should be part of our announcement?

@UlrikeFischer Not us directly, but we've basically driven this

@JosephWright yes!

@PauloCereda I've been after this for ages: there's a reason all of expl3 just assumes UTF-8 and some stuff (like case changing) will break with other encodings

@PauloCereda Secret Plan is next ...

@JosephWright yay

@PauloCereda Well, actually there's one more bit of prep: getting the debugging of expl3 loaded after the main code. But that's just a question of agreeing interfaces

@PauloCereda The whole latex-dev business is part of the set-up here :)

@JosephWright great work :)

@JosephWright sounds sensible to mention that a setting is needed and this is in a current texmf.cnf (and that it perhaps doesn't work in older windows). (The new setting is at the end and explains also how to disable it if needed.).

@UlrikeFischer Yeah

@JosephWright but two days ago I got a document compiled with texlive 2012 in mac encoding ... Correcting this is a bit of challenge.

@UlrikeFischer Sure, it's tricky for older files. But for expl3 functions I think the situation is OK: it's only a subset that care, and those ones are all things that are 'new'

@JosephWright Ostrich algorithm. :)

@JosephWright oh, I didn't mean that I tried to use expl3 there - only that there are still a lot quite old systems and encodings around ...

@JosephWright en.wikipedia.org/wiki/Ostrich_algorithm

@UlrikeFischer Sure, but they are extremely unlikely to get a kernel update, and if they do then really it's time to re-encode the sources

@JosephWright yes, I don't worry about this.

@UlrikeFischer with standard formats all tests pass except the expect attachfile2, so running again now I see you pushed an update there... (does means we'll have to update a lot of test results when we switch to the updated format, not sure changing wording in font messages is that useful:-)

@UlrikeFischer still get diffs, how come I get two /Producer?

! /Producer (pdfTeX)/Author()/Title()/Subject()/Creator(LaTeX with hyperref)/Keywords()
! /Producer (pdfTeX)/Author()/Title()/Subject()/Creator(LaTeX with hyperref)/Producer(pdfTeX-1.40.20)/Keywords()

@DavidCarlisle Which one? (I have to go now, will answer later)

@UlrikeFischer that was attachfile2.pdftex.pdf.diff (all three fail)

attachfile2 fails with xetex with generated font names ! /CMapName /APWWDF+LMRoman10-Regular-UTF16 def isn't ! /CMapName /KTKDZY+LMRoman10-Regular-UTF16 def

@DavidCarlisle are you sure that you did use tl2019 for the test? The cmapname business came up last year and was resolved by a xdvipdfmx change.

@UlrikeFischer that's what I thought..... I wonder if that never got pushed to cygwin builds...

$ type xdvipdfmx
xdvipdfmx is /usr/local/texlive/2019/bin/x86_64-cygwin/xdvipdfmx

@UlrikeFischer I get

$ xdvipdfmx --version
This is xdvipdfmx Version 20190503 by the DVIPDFMx project team,
modified for TeX Live,
an extended version of DVIPDFMx, which in turn was
an extended version of dvipdfm-0.13.2c developed by Mark A. Wicks.

Copyright (C) 2002-2019 the DVIPDFMx project team
Copyright (C) 2006-2019 SIL International.

This is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or

20190503 sounds new enough?

@DavidCarlisle hmmm

@DavidCarlisle yes, that's the version I have too. Perhaps let's try to find the reason for the pdftex failure first. Can you run l3build check -epdftex attachfile2 and then sent me the diff and the real log?

@UlrikeFischer ok starting from here, I'll mail logs in a bit:

$ git branch
  attachfile-issue69
* master
  split


$ git pull
Already up to date.

@UlrikeFischer mailed

@DavidCarlisle you have a local regression-test.tex. Is this current?

@UlrikeFischer I don't know, do I need to update miktex in user and admin modes?

@UlrikeFischer I deleted that so the test redone using /usr/local/texlive/2019/texmf-dist/tex/latex/l3build/regression-test.tex but same diff

@DavidCarlisle I have a newer hyperref in texmfhome, which contains all the commits which I added in april to remove the producer code ...

@DavidCarlisle Windows woes

@UlrikeFischer I knew we'd discussed this before, do you mean newer as not in master branch?

@UlrikeFischer oh /usr/local/texlive/2019/texmf-dist/tex/latex/hyperref/hyperref.sty can't really look much now, why didn't it use the one it unpacked from the dtx...

@DavidCarlisle no, I'm on master and there are no unmerged branches, so imho my state is simply the head.

@DavidCarlisle you are testing in oberdiek, there is no hyperref to unpack.

@UlrikeFischer ah there is that....

@UlrikeFischer we could release hyperref to ctan, wait a couple of days then repeat the test here.....

@DavidCarlisle but I really need to understand better how l3build handles texmfhome - sometimes it seems to use it and sometimes not.

@UlrikeFischer actually I think this is near enough a pass that I'm tempted to upload hyperref and oberdiek to ctan....

@DavidCarlisle yes, I don't see a problem here. But one probably will need to check the changes for the announcement and the docu.

@UlrikeFischer I'll look at both this evening if I can

@DavidCarlisle Marcel just made a pull request, I will check and then pull.

@MarcelKrüger good timing ;-) we were just considering an update.

@UlrikeFischer yes just seen I made a comment, looks good basically but version numbers and dates will need bumping, possibly merge the PR then fix the versions?

It's tempting to drop luatex.lua altogether but I guess not this time....

@DavidCarlisle I was also thinking if it even makes sense to change luatex.lua given that it mostly appears to be a historical artifact. But I decided for consistency.

@UlrikeFischer Great. This are changes from the category: I need it for luametatex, but I think it is better for LuaTeX too.

@DavidCarlisle I have forgotten what will explode if we drop it?

@UlrikeFischer some unknown person's document which has \usepackage{luatex} I removed all the dependency on luatex.lua from the files in this collection some time ago,

@DavidCarlisle tempting ...

@UlrikeFischer I really should get my split branch working so we can slit this up into sensible collections and then luatex.sty/lua can be in its own explicitly obsoleted ctan upload.

@DavidCarlisle I tried to find all the places where one has to adapt a version, apart from the direct dtx it seems to be oberdiek.tex and oberdiek-bundle.bib, or is there more?

@UlrikeFischer you need README to have the newest date or Petra will be cross:-) To save keep changing the README and oberdiek.tex I tend to use the same version date for all, that is it wouldn't bother me if you use yesterdays date on all the files in this PR as there has not been a CTAN release, but updating the date is fine too

@UlrikeFischer the main pain is how many times the date appears in each dtx file

@DavidCarlisle OK I will adapt the readme dates too as we are planing an update. I will use today everwhere ...

@DavidCarlisle Use the same approach I have for L3/L2e? marker date in the README?

@JosephWright we only use l3build for check not build, (it builds into a complicated tds structure across latex and generic and I'm not sure l3build can do it) (but yes the bash script could sed in a date at the end)

@JosephWright I think we discussed it before, oberdiek/mkctan does:

# fill tds tree
for i in `cat manifest.txt`
do
    cp `basename $i` $i
done

That is, it populates the tds tree from an existing manifest file rather than from heuristics based on the file name

@DavidCarlisle Yes, we did: it's a feature we could do with for the most complex cases

@DavidCarlisle We could basically add that as an approach: Will's used manifestfile for something else, but we could have something like install_locations or whatever: workable?

@JosephWright yes the above is really all there is to it (although I note I have arranged elsewhere the subdirectories are in place which should probably also be here so then it's simply a matter of looping through the file and copying things, also it would mean @UlrikeFischer could build the ctan releases.., currently building requires bash setup. for install_locations would you use a lua table to specify the targets or as above just specify an existing manifest text file (Either would work)

@DavidCarlisle Hmm, rather than a separate file I think a table makes sense: just needs to be a list of files

install_manifest =
  {
    "bibtex/bib/oberdiek/oberdiek-bundle.bib" ,
    ...
  }

@DavidCarlisle ^^^ Something as simple as that? If the manifest table is non-empty, we use it for entries which are found?

@DavidCarlisle All I have to do is loop over the table, convert to a key-value set up (keys are file names), then use that in the install routine where there is an entry

@DavidCarlisle. @UlrikeFischer Sound OK? If so, I think implementation is pretty easy

@JosephWright yes but (in the files I picked up from Heiko) the manifest.txt was already part of the public distribution, so the code was designed to ensure that what I generated didn't change that. You could use the lua table as a list then use the existing manifestfile to generate a manifest file which would be the same thing but that seems slightly circular

@DavidCarlisle Hmm, OK ...

@DavidCarlisle How about we have install_manifest_file = "manifest.txt"? Logic being we can (a) read a file (b) use a table (c) use the default locations (as now). For HO stuff, (a) works, for any new complex sets ups probably go with (b), (c) is the standard 'just guess' approach

@JosephWright but I'm not too bothered as I don't think the manifest file is actually useful to many people. It doesn't tell you anything that unzip -l oberdiek.tds.zip doesn't tell you.

@DavidCarlisle Well yes (which is why I've never used Will's l3build manifest either ....)

@DavidCarlisle Would be cleaner not to have to support it: we are looking at a one-shot conversion after I add the feature

@JosephWright so go with whatever seems most natural, I'll get this release out as it is, but then adjust to fit for teh next release.

@DavidCarlisle OK

@DavidCarlisle install_manifest a good name?

@JosephWright it would be trivial to turn the existing file into a lua table so probably go with that. (If I felt motivated to keep manifest.txt I could populate the locations lua table from that file in custom lua in the oberdiek build.lua )

@JosephWright tds_locations ?

@DavidCarlisle Yup

@DavidCarlisle Good name :)

@JosephWright imho being able to tell l3build directly where to install a file would be quite usefull, I had cases too where I fighted against the heuristic.

@DavidCarlisle, @UlrikeFischer If we have tds_locations, how does it interact with other stuff? For example, does it take wildcards? Do we install 'just what it shows' or use the standard locations for other stuff? In some ways a simple switch is easiest ...

@DavidCarlisle, @UlrikeFischer Does it error if a file is missing?

@UlrikeFischer Sure, I've had to make some 'interesting' settings myself

@DavidCarlisle, @UlrikeFischer If we sort this, there's then really only the 'auto-versioning' stuff preventing people like pgf or biblatex using l3build: should be able to cover everything

@JosephWright I was wondering whether to allow wildcards, it might be convenient sometimes, to be able to say all *.lua goes to generic/scripts or whatever, on the other hand my current use was to ensure a pre-existing exact file list (but reverse engineering a build script from a public ctan release is hopefully not the common case)

@DavidCarlisle Yes, exactly: a simply list of wildcards would fit with our existing variables too: you still list all the installfiles or whatever`, and just do a pattern match for the final destiantion

@DavidCarlisle, @UlrikeFischer I'll sketch something out later

@DavidCarlisle @JosephWright interesting problem: magicnum.sty does require("oberdiek.magicnum") which means that it finds the version in the texlive texmf and not the one in build/test and so the test fails as the lua has the wrong version ...

@UlrikeFischer I hoped at one point to get rid of the dotted name weirdness but when I tried I broke something, if you look at mkctan in obediek you'll see I have some comments about having to copy those to install locally before things work. If we coul dnot do that at all it wul dbe better, but I'm not sure of teh compatibility implications

@UlrikeFischer :

## Lua need to be here for doc generation must fix this one day
cp \
    luacolor.lua \
    pdftexcmds.lua oberdiek.pdftexcmds.lua \
    magicnum.lua oberdiek.magicnum.lua \
    oberdiek.luatex.lua \
    scripts/oberdiek

Are these oberdiek.* filenames a legacy of the module use that @MarcelKrüger just removed?

@UlrikeFischer or to put it another way if the .sty did require("magicnum") instead, what would go wrong?

@DavidCarlisle well imho now that Marcel changed this, one could remove the "oberdiek.", but this probably will mean that require could also find a "weird/magicnum.lua" and not only "oberdiek/magicnum.lua"

@UlrikeFischer yes but that's always the case with any kpse lookup

@yo' -- Hi! How is scout camp?

@DavidCarlisle yes, but if we change it, we should at least check for name clashes ;-). The alternative would be to rename the lua, (there is an oberdiek.luatex.lua)

@UlrikeFischer I did before replying:

$ kpsewhich -all magicnum.lua
/usr/local/texlive/2019/texmf-dist/scripts/oberdiek/magicnum.lua

@UlrikeFischer oh yes I think there used to be luatex.lua and oberdiek.luatex.lua but I wasn't brave enough to delete both...

@DavidCarlisle I know deleted the oberdiek in magicnum (it was already gone in luacolor and pdftexcmds). And I could try to remove luatex.lua ...

@UlrikeFischer, @DavidCarlisle Grr, differences between Lua and TeX file searching

@JosephWright ?

@JosephWright can't you tell lua to search like tex via the lua api to kpathsea?

@UlrikeFischer The oberdiek.<name> business: Lua doesn't require unique names, and mixes up tree and file name ...

@DavidCarlisle Yes, that's what we do in l3build :)

@UlrikeFischer, @DavidCarlisle I think making a tds_location approach is quite workable; hopefully later today

@JosephWright oh that, I think we are just going to remove the oberdiek. versions....

@JosephWright I removed "oberdiek." and the test now passed.

@UlrikeFischer Cool

@UlrikeFischer just removed oberdiek. from the call, or no longer generate the file?

@DavidCarlisle I removed it from magicnum.sty: require("magicnum")%

@DavidCarlisle I have three failures, embedfile-test3 because of different line endings in the embedded file, pdfrender-test3 because it still has the additional /Pdfproducer, and pdfcolparallel-test1 because luatex decided to set an xheight from 430 to 431 (blame Will?). So nothing serious, imho one can ignore them. I push this. Should I try to remove luatex.dtx?

tdslocations =
  {
    "bibtex/bst/foo/*.bst"
  }

@DavidCarlisle, @UlrikeFischer ^^^ Imagining something like the above for non-standard locations?

@DavidCarlisle, @UlrikeFischer I'd wondered about including all of the standard stuff, but particularly for tex/latex, the file extensions could be 'whatever' and I'm reluctant to impact on the common case of 'just go with the defaults'

twitter.com/marcocongiu/status/1154289457874526208?s=09

@JosephWright looks good.

@DavidCarlisle ^^^

@JosephWright I prefer explicit settings - much easier to correct and debug.

@UlrikeFischer Huh

@UlrikeFischer I'd thought of

tdslocations =
  {
    ["*.bib"] = "bibtex/bib/" .. module ,
    ["*.bst"] = "bibtex/bst/" .. module ,
    ["*.idx"] = "makeindex/" .. module ,
    ["*.lua"] = "scripts/" .. module ,
    ["*.tex"] = ""
  }

but it gets tricky and I think a bit verbose

@UlrikeFischer The idea is to allow wildcards so for example *.cbx but at the same time to also support single file names

@JosephWright I think your first suggestion is fine - I did understand it directly ;-)

tdslocations =
  {
    "bibtex/bst/foo/*.bst" ,
    "bibtex/bst/bar/odd.bst"
  }

local tdsmap = { }
for _,pathglob in pairs{tdslocations} do
  path,glob = splitpath(pathglob)
  tdsmap[glob] = path
end

for glob,path in pairs(tdsmap) do
  local pattern = glob_to_pattern(glob)
  if string.match(NAME,pattern) then
    PATH= path
  end
end

@UlrikeFischer ^^^ Some partially-complete code. We have a table of defined targets, which is used to make a new table on-the-fly for lookup of names. If we get a hit, use the path given, if not, I'll code it to use the default paths we've had from day one

@JosephWright looks good imho.

@UlrikeFischer Main thing is covering uninstall/dry run, rather than the actual core idea!

@DavidCarlisle, @egreg, @UlrikeFischer, @PhelypeOleinik Interesting mail from Karl .. will forward

@JosephWright actually more imagined something like {"*.bst", "bibtex/bst/foo"} that is a list of mappings of local name to target directory

@DavidCarlisle That's my second suggestion above:

tdslocations =
  {
    ["*.bib"] = "bibtex/bib/" .. module ,
    ["*.bst"] = "bibtex/bst/" .. module ,
    ["*.idx"] = "makeindex/" .. module ,
    ["*.lua"] = "scripts/" .. module ,
    ["*.tex"] = ""
  }

@JosephWright sorry just catching up...

@DavidCarlisle Saves me a re-write step on-the-fly, is slightly harder for the user to write

@JosephWright so for the first version the last step may be a glob and is matched against the current dirctory where things have been built? sounds OK to me

@JosephWright which one is easier for you?

@UlrikeFischer It's much of a muchness: all I have to do for the first version is use about 4 Lua lines to create the second version

@DavidCarlisle Yes, I do path,glob = splitpath(pathglob) then the glob could be a single name or a real glob, either way converts to a pattern for matching

How did I end up as the pdfTeX maintainer?

@JosephWright There's a sucker born every minute. PT Barnum

@JosephWright Saw it. Non more hassles with unexpandable \input. ;-)

@egreg Not quite ...

@egreg Frank and I have talked about this, the issue is that we can't do a controlled file lookup with \input, other than in LuaTeX of course

@egreg I've got to write the code first!

@AlanMunn Sure, but somehow the team seem now to be in charge, more or less, of all of the engines. We've accumulated all the power jobs

@UlrikeFischer, @DavidCarlisle I have proof-of-principal for the TDS stuff: I'll push a branch for checking?

@JosephWright Seems like a nice change, except for who will write the code :-)

@PhelypeOleinik Er .. yes

@PhelypeOleinik Do you know WEB? ;)

@DavidCarlisle, @UlrikeFischer Take a look at TDS branch: I have to tidy up uninstall and --dry-run, but the idea should be workable

@JosephWright Does anyone (except for DEK) know?

@JosephWright No, just saw some of TeX's source but didn't understand much. I'm searching for the code to see if I can understand something...

@PhelypeOleinik Hans, Taco, pTeX people

@PhelypeOleinik tug.org/svn/texlive/trunk/Build/source/texk/web2c/pdftexdir/… line 12330 onward

> in many respects it would be much better if the old syntax wouldn't work at all, but for this we would get killed I guess by angry users

Hi there, typesetting aficionados

I have a question for y'all

@EmilioPisanty Hi

what's with papers with en-dashes instead of hyphens in their titles?

case in point:

> Three-dimensional skyrmions in spin-2 Bose–Einstein condensates

doi.org/10.1088/1367-2630/aac2a8

@EmilioPisanty En-dash is used where 'and' or 'or' would fit, '-' is strictly a hyphen for word breaking

@EmilioPisanty Here, you have 'Bose and Einstein', so two people, en-dash to join them, not a single surname

but "spin-2"?

I guess that's an adjectival phrase?

@EmilioPisanty That's a hyphen, or it should be

@EmilioPisanty Yes

@JosephWright I guess I learned something today =|

I think that's a bizarre convention

but I can see that it has an internal logic

@EmilioPisanty Like I say, a hyphen does one very specific job

I guess the correct way to put this in a .bib file is just using a double dash?

@EmilioPisanty Yes, easiest at least

@JosephWright would biber sort Bose--Einstein differently from Bose–Einstein?

@StrongBad That I don't know: if using Biber I'd stick to UTF-8

@JosephWright thanks =)

@JosephWright but if I get e.g. Stable particlelike solitons with multiply quantized vortex lines in Bose-Einstein condensates with just a hyphen, then I just keep the hyphen?

@EmilioPisanty Well ... this is tricky as a lot of authors just mess these things up. I guess formally you should cite as seen, but this is so borderline I'd correct

@JosephWright thanks

@JosephWright well, what if it's house style at Phys Rev journals to use hyphens in Bose-Einstein?

from the smallish sample I've seen it looks like they take that tack (but I need a bigger sample to confirm)

@EmilioPisanty Well, that's their call: I'd write it 'right' and let them worry if Iwas submitting a paper

hmmmm

but isn't that changing the title from as-seen?

or rather -- is it really that strong a convention that Phys Rev is so wrong that it needs to be corrected regardless?

@PhelypeOleinik the person who added \expanded to xetex must know WEB, we could get him to do it

@DavidCarlisle Apparently he's the maintainer of pdfTeX now, so it seems good ;-)

@PhelypeOleinik best to concentrate blame in one place

@UlrikeFischer what's the current plan? still generate \file{oberdiek.pdftexcmds.lua}{\from{pdftexcmds.dtx}{lua}}% but don't use it anywhere?

@DavidCarlisle I haven't really looked at pdftexcmds, as it already used require("pdftexcmds"). Is this really still somewhere? On my system unpack creates pdftexcmds.lua.

@UlrikeFischer it makes it with and without oberdiek (copy of same file) see

 ls /usr/local/texlive/2019/texmf-dist/scripts/oberdiek/
luacolor.lua  oberdiek.luatex.lua    oberdiek.pdftexcmds.lua  pdftexcmds.lua
magicnum.lua  oberdiek.magicnum.lua  pdfatfi.pl

@UlrikeFischer I'm happy to leave in the generation of the files for now, just avoid using them.

@UlrikeFischer so oberdiek and hyperref can go out?

@DavidCarlisle yes, but I could also push pdftexcmd and magicnum without the oberdiek-file.

@UlrikeFischer probably safer to leave them in for now.

@DavidCarlisle fine.

bcd2725 (HEAD -> master, origin/master, origin/HEAD) remove dot prefix in lua call
ceb9ee3 Pulled PR#70 and adapted versions
5e7ae76 Don't use module
d613a64 remade attachfile tests without mod date
e9915f8 (origin/attachfile-issue69, attachfile-issue69) date
90e9c2b Merge branch 'attachfile-issue69' of github.com:ho-tex/oberdiek into attachfile-issue69
ac1b5bc test for attachfile2
f05832d added luatex driver for issue #69
3d8e76d test for attachfile2
7aa4bf3 added luatex driver for issue #69
1526bac date

If I have three complementary options to a package, what's the preferred way of implementing them with kvoptions?

@JosephWright if I can get oberdiek and hyperref out the door i'll try to set up build scripts for them using l3build/TDS,

@AlanMunn have one option, with three values?

@DavidCarlisle Cool

@JosephWright building oberdiek takes forever, lualatex every dtx + oberdiek.tex three or four times each....

@DavidCarlisle And then make 3 conditionals for the bits of code that each value implements?

@AlanMunn something like that. but it's easier for the user to see that foo=red,foo=blue is bad than to see red,blue is bad.

@DavidCarlisle Right. That makes sense.

@AlanMunn do it quick as I'm just rebuilding the oberdiek bundle now, so if I upload to ctan, probably nothing will work at all.

@DavidCarlisle lol

In a tabular p{} col, is there any way of suggesting it split at a particular place?

Hmm, surprisingly, \newline works. I learned that from

@DavidCarlisle when rebuilding those dtx files, do they need stuff from each other? Otherwise run then in parallel (I would assuyou already did). I recently converted some of my build scripts lots of single doc abstracts into running parallel, saves so much time.

@FaheemMitha it's just a \parbox so anything that works in parbox works there, except \\ has been stolen to end the table row

@DavidCarlisle Stolen?

@daleif hmm I could but then if I convert the build to l3build I'd need feature request @JosephWright...

@FaheemMitha redefined

@daleif how many nodes do you have?? on this laptop it's dual core with allegedly two hyperthreads per core so only 4 so there are not vast amounts of parallism although if it was 1/3 of the total time that would be a useful saving...

@daleif since a full build takes well over an hour if I recall correctly

@DavidCarlisle Er, yes, the design brief was 'belt and braces'

@DavidCarlisle I've wondered about getting an i7

@JosephWright this is an i7 with an ssd and it still takes over an hour

Processor Intel(R) Core(TM) i7-7600U CPU @ 2.80GHz, 2901 Mhz, 2 Core(s), 4 Logical Processor(s)

@DavidCarlisle Ah.

@DavidCarlisle I hope you disabled the failing test.

@UlrikeFischer mkctan doesn't run the checks it just builds the zip of what is there, as I say, it takes long enough to do that:-). (I did run l3build check and check I just got the attachfile diffss)

@AlanMunn oh it's got as far as building kvoptions doc....

@FaheemMitha if you use \raggedright as I see in the quote above then \\ would have worked as a linebreak except it suggested \arraybackslash which just redefines \\ back to mean end of table row.

@DavidCarlisle my office pc is a quad core, so up to eight threads. I gave the script 7 threads to use.

@DavidCarlisle I'm not using \raggedright.

@FaheemMitha \newline (or a blank line) are the things to use then

@DavidCarlisle ah right, I forgot ;-)

@DavidCarlisle Yes, I used \newline.

A 2 core i7? I would what thought it would be at least a quad core

Is p{} generally recommended if one wants to control column width?

c, l, and r don't let one do so.

Are you still running things under cygwin?

@daleif parallelism from Lua seems hard/non standard surprisingly enough, so getting l3build to do thet seems hard

@daleif yes, I know....

@daleif I guess TeX doesn't get a lot of corporate funding.

@FaheemMitha probably

What is needed is a supercomputer.

@FaheemMitha they do with array.sty (see w modifier

@DavidCarlisle Oh. I don't really know array.

@FaheemMitha it gets none but this isn't a tex machine, it's a work one

I'll check it out. There are so many packages, it's hard to keep track.

@DavidCarlisle NAG?

@DavidCarlisle that's why my script is in perl. But the fork controller is probably not in tlperl

@FaheemMitha the little bit of tabular syntax you showed above is array package syntax.

I'm getting yet more TeX practice, doing my taxes.

@DavidCarlisle You mean p{}?

@FaheemMitha no >{\raggedright...}

@DavidCarlisle I didn't mention it. You did.

@daleif I don't think we buy the very best at any point but go for some resonable compromise, this machine is a couple of years old or so, but was not a bad spec when new I think, core i7 and 500Mb ssd

@DavidCarlisle That's an Apple machine?

@FaheemMitha only because it is shown in the fragment you quoted here:

@FaheemMitha no dell running windows 10

@FaheemMitha \newcolumntype and the > syntax are from array package.

@DavidCarlisle Oh, that. Right. I just took the \newline bit from there.

@DavidCarlisle Ok. My mistake.

Hmm, I got a blank when I tried to typeset Василий Буков.

That's Cyrillic, I suppose.

Missing fonts, perhaps.

@DavidCarlisle My office PC is from 2012, running Ubuntu. i7 quad core, 32Gb of RAM. I do have its successor in my office (also an i7 quad core, just a newer one).

@FaheemMitha you need a font (the log will be complaining about missing characters)

@DavidCarlisle Would it tell me what to get?

@daleif we do have access to machines with thousands of cores but I probably shouldn't use them for building eh oberdiek bundle for ctan:-)

@FaheemMitha no

@DavidCarlisle we have on some of our faster machines. Why not if no-one else is using them. I can try and make you a script. I know your tool chain should be in lua, but if this can help you guys safe some time once in a while.

@DavidCarlisle Ok.

@daleif No:-)

@DavidCarlisle My current machine is i5 :(

@DavidCarlisle Performance was not one of the criteria at design time, and if we want it I'd go for using Travis-CI

hour and 5mins in it's started making oberdiek.pdf so getting there....

yay:

real    66m50.527s
user    25m48.518s
sys     26m39.619s

@JosephWright what exactly is the benefit of tracis-CI in relation to LaTeX? I don't get it.

@DavidCarlisle exactly what are you compiling here? The entire core/all that the group supports?

@daleif just making oberdiek.zip for ctan so that is: run tex on the .ins file to make all the packages, run lualatex on each dtx file to get pdfs, zip everything up, stop.

@daleif so not latex project code at all, officially.

@DavidCarlisle hmm, I'll try that at some point. Ought to be able to get the files from CTAN right?

@daleif well easiest to check out (or grab zip of) github.com/ho-tex/oberdiek then run the mkctan bash script...

Are there any rules on the PDF generation? like special purpose index setup etc? or will latexmk -lualatex do?

@daleif currently it's hand tweaked lualatex and bibtex and makeindex calls to do the right thing, no idea if latexmk would do the right thing.

@daleif the pdf making part of the script is trivial:

# make pdf
rm *.bbl || echo no bbl
for i in *.dtx
do
lualatex ${i}
if [ -f ${i/.dtx/.bcf} ]; then
    #    biber ${i/.dtx}
    bibtex ${i/.dtx}
fi
lualatex ${i}
if [ -f ${i/.dtx/.idx} ]; then
    makeindex -s gind.ist ${i/.dtx/.idx}
    lualatex ${i}
fi
done

lualatex oberdiek
lualatex oberdiek

@DavidCarlisle I seem to be able to run it - at least it creates documentation ...

@UlrikeFischer you have a windows bash?

@DavidCarlisle sure, I have git, there is bash included.

so it is just a bash script

@daleif yes

@UlrikeFischer ah I forget (never used the windows git) and on other platforms you don't use an embedded bash

@UlrikeFischer do you want to build hyperref then?:-) (I'm just running l3build check there...) (oberdiek gone to ctan)

@daleif Travis-CI runs stuff on cloud systems; it's fast and not using local resources. So we could have it build our zip files

@DavidCarlisle Looking forward to hearing if my adjustments help with building

@UlrikeFischer @JosephWright oops hyperref test failed

! ! Undefined control sequence.
! <recently read> \unquote@name
! l. ...Normal \includegraphics{hog}

@DavidCarlisle which test is this?

- ./build/test/test0.luatex.diff but the diff may be my local setup, checking....

@UlrikeFischer oh graphics.....

@UlrikeFischer hmm TEXMFHOME latex-dev searched before standard latex tree :(

Here is an interesting one:

\newcommand{\idxmark}[1]{{\let\gobbleone\relax\markboth{\sffamily\bfseries#1}{\sffamily\bfseries#1}}#1}

that local \let\gobbleone\relax never seems to be running before #1 hits the headers and I do not understand why.

I've set the normal \gobbleone definition to \DeclareRobustCommand\gobbleone[1]{\typeout{gobbled: #1}}

Hmm, seems more likely to belong where the headers are executed

@daleif your redefinition will be out of scope by the time gobbleone is run won't it (as you made it robust)

@DavidCarlisle that was what I just got to. The normal version is not robut

@UlrikeFischer all checks passed

@DavidCarlisle why did it find latex-dev first?

@DavidCarlisle you can probably parallelize that bash part using GNU parallel. Though I got interesting lualatex errors when I tried a simple version. Might need some logging

@DavidCarlisle I tried, but it fails at the zip-step, and it polutes the folder with lots of file git then wants to track. Must some build folder exist?

usual issue that you reported that the fallback search of all of tex gets in the way:

$ kpsexpand '$TEXINPUTS'
.:{{}/home/davidc/.texlive2019/texmf-config,/home/davidc/.texlive2019/texmf-var,/home/davidc/texmf,!!/usr/local/texlive/texmf-local,!!/usr/local/texlive/2019/texmf-config,!!/usr/local/texlive/2019/texmf-var,!!/usr/local/texlive/2019/texmf-dist}/tex/{kpsewhich,generic,}//

@DavidCarlisle Documentation is overrated.

@AlanMunn definitely

@DavidCarlisle but why for a lualatex run? It should have latex first -- unless you still have the faulty lualibs (or something isn't still right there).