@Canageek unlikely to be directly related to fonts other than the different fonts caused a different line break somewhere which tipped the column breaking on to a bad path. Do you know how it does twocolumn? (standard latex, multicol.sty, balance.sty ,....?)

@DavidCarlisle What was the problem? (I did not see it in the image)

@mickep @Canageek didn't say but I assume it is that the second page is running off the bottom of the page

@DavidCarlisle Oh, you mean that is a problem? :D Thanks!

@mickep actually it's just Chemistry, so we're probably not missing much:-)

@DavidCarlisle Oh, just don't say it so @JosephWright hear...

ooh chemistry is complicated

Is there a L3 equivalent of \null? Or should one use \hbox:n {} (or if the current mode is guaranteed to be horizontal \box_use:N \c_empty_box)?

@PauloCereda how comes?

@Skillmon there is a lot of elements :)

@Skillmon ooh

@JosephWright: The oxygen went on a date with potassium and it went OK. :)

/sad trombone

@Skillmon At present there is no explicit \null: I'd use \hbox:n { } I think

@JosephWright if in a tabular just to make sure the current line at the end is not gobbled, shouldn't \box_use:N \c_empty_box be more performant?

@Skillmon \c_empty_box is not an hbox

@JosephWright I know, but in horizontal mode placing a void box should do the same, or am I wrong?

@Skillmon That is true, yes

ooh chocolate

@Skillmon two days ago I went to a friend's house to get a haircut and she had a bunny rabbit as a pet. :)

@yo' Šťastné narodeniny ;-)

@Skillmon not normally, if the \mybox register is void, \box\mybox does nothing, but if its an empty hbox, it adds a box.

@PauloCereda keeps it fresh without filling up the freezer

@egreg is that Slovak? :) Anyway, thanks a lot!

@egreg This answer tex.stackexchange.com/a/378213/2693 no longer seems to work (but no error).

@DavidCarlisle oh no

@PauloCereda This is bad chemistry. The potassium has to go on a date with the oxygen twins. It was explosive.

@AlanMunn ooh :D

@PauloCereda tinyurl.com/y3dylscr

@DavidCarlisle ooh

@yo' I knew you'd realize it! :-)

@AlanMunn I'll have a look

@AlanMunn I tried and it reproduces the same exact output.

@egreg My mistake, I must have been playing around with the code at the time and didn't notice the version I had was different from your original.

@DavidCarlisle it works in tricking the \crcr in a last tabular line.

@Skillmon I didn't see the context, just replying to the comment about void and empty hbox being the same in h-mode. But OK (you mean like \relax ?)

@DavidCarlisle yes, that might be even better.

Overleaf should give every package author a membership. :)

@PauloCereda can you provide more reasoning? I mean, I'm a package author, but none of my packages are really wide spread nor of that great use. There is (almost) no quality control in place, as long as it exists it'll be published. As long as the license is free, it'll get included into TL. This would flood CTAN with low-quality packages, so some could get a free OL pro account.

@Skillmon I am just kidding. :) I had to share a project with a colleague and found out that I can only share with one person.

You are right, it would be a mess. Besides, I don't like Overleaf. I am still an offline duck. :)

@PauloCereda that explains why you are never on stackexchange

@DavidCarlisle exactly

Wait a minute...

Oopsie.

> Grumpy Cat has died from complications following a urinary tract infection. She was 7 years old.

Oh no

Poor grumpy cat

@PauloCereda personally I'd prefer my personal git server and offline work for collaborative tasks, but I understand that for others Overleaf can be a great tool.

@PauloCereda, @egreg Seen Dick Koch's mail?

@JosephWright Where?

@egreg OS X TeX mailing list

@JosephWright Not subscribing to it.

@egreg Ah

@JosephWright I tried subscribing

@egreg He's pre-testing a new, signed and notarised, version of BasicTeX to be ready for latest Apple security changes

@JosephWright Is it really useful if packages can't be added?

@egreg I don't think packages are an issue; it's binary signing that's the issue

@JosephWright Oh, yes. Let's see if it catches on.

@egreg Not going to be optional, I think

@santimirandarp Probably you mean you are using some scripting which converts TeX-like input into maths output ...

@JosephWright I tried to write your spotcolor with \driver_pdf_object_write:nn but had some trouble to get the # right. In the end I had to use \pdfescapename:

\driver_pdf_object_write:nx{spotcolor_obj}
{
  /Separation~/\pdfescapename{BarTone~555~GN} /DeviceCMYK
      << /C0~[0~0~0~0] /FunctionType~2 /C1 [0.8~0.2~0.5~0.3] /Domain [0~1] /N~1 >>
}

@UlrikeFischer That's just a question of passing a suitable string from higher up: I have some code at the moment using \tl_replace_all:Nnn, but probably I want an expandable version (easy enough). At the driver level, it's a question of saying 'it has to be a valid string'

\cs_new_protected:Npn \color_spot_const:nnn #1#2#3
  {
    \@@_spot_const:nnn {#1} {#2}
      { \color_parse:nN {#3} }
  }
\cs_new_protected:Npn \color_spot_const:nnnn #1#2#3#4
  {
    \@@_spot_const:nnn {#1} {#2}
      { \@@_direct:nnN {#3} {#4} }
  }
\cs_new_protected:Npx \@@_spot_const:nnn #1#2#3
  {
    \tl_if_exist:cTF { c_@@_spot_ #1 _tl }
      { \__kernel_msg_error:nnn { color } { spot-already-defined } {#1} }
      {
        \group_begin:
          \tl_set:Nn \exp_not:N \l_color_fixed_model_tl { cmyk }

@UlrikeFischer ^^^ Current experiments (not checked in as we have no page resource management in the core)

And Joseph I think the resource management works now:

\documentclass[12pt,oneside]{article}
\usepackage[patches]{pdfresources}
\usepackage{tikz,xcolor}
\usepackage{transparent}
\ExplSyntaxOn
\driver_pdf_compresslevel:n {0}
\driver_pdf_compress_objects:n {0}
%test local resource:
%\driver_pdf_pageresources_gput:nnnn{1}{ColorSpace}{abc}{/Deviceblub}

\driver_pdf_object_new:nn {spotcolor_obj}{array}
\cs_generate_variant:Nn \driver_pdf_object_write:nn {nx}
\driver_pdf_object_write:nx{spotcolor_obj}
{
  /Separation~/\pdfescapename{BarTone~555~GN} /DeviceCMYK

@UlrikeFischer oGood

@UlrikeFischer I think we have most of the ideas, it's just a question of how to get them into the kernel in a way everyone is happy with. Oddly, I'm minded just to add to expl3 then more or less \RequirePackage{expl3} just before \dump in latex.ltx

@JosephWright I'm not quite sure if everything really works at the end of the document correctly - currently I'm simply outputting the object in \AtEndDocument. But that are details.

@UlrikeFischer This is where Hans has the advantage in ConTeXt: can be sure things go into the end of the PDF

@JosephWright and that he has only one driver to worry about ;-). That's the main problem with simply adding expl3: it won't work with dvipdfmx ...

@UlrikeFischer I think that's where I come in and sort things, no? ;)

@UlrikeFischer Also dvips I guess?

@JosephWright dvips is the default driver so the latex.fmt will load it.

@UlrikeFischer Driver loading clearly will have to be fiddled about with if we are slipstreaming into the 2e kernel

@JosephWright yes. Regarding the string: one can use \str_set_convert:Nnnn \l_tmpa_str {BarTone~555~GN}{utf8}{utf8/name} and then use \l_tmpa_str but this is not expandable.