« first day (4828 days earlier)      last day (88 days later) » 

7:44 AM
@JosephWright if you don't have to work by expansion \scantokens isn't broken (enough to be unusable). Just do \begingroup\catcode`\^^?=12 \unexpanded{\endgroup\protected\long\def\myscantokens#1{\begingroup\catcode`^^?=14 \scantokens{\endgroup#1^^?}}} and for most stuff it'll just work™. But yes, otherwise it's basically broken by using a pseudo-file, an endline after the last token and an end-of-file as well. Maybe backport LuaTeX's \scantextokens into e-TeX?
 
@Skillmon Well the other issue with \scantokens is things like {%}, but yes I know it has its uses
 
@JosephWright user's fault for wrong input.
 
@Skillmon Sure - I am just saying it's a tricky primitive even if one deals with the end-of-file issue
 
 
1 hour later…
8:53 AM
 
9:08 AM
@DavidCarlisle great minds think alike.
 
@DavidCarlisle Or go with xpdftex.
 
@mickep we would have to resuscitate our dormant children first
@mickep would you write an xuptex ?
 
@DavidCarlisle I leave that for the C programmers on the ConTeXT list.
 
@mickep :)
 
 
2 hours later…
11:11 AM
@JosephWright siunits should probably load color in the begindocument/before hook: tex.stackexchange.com/a/707422/2388
 
@UlrikeFischer vvv
    {
      \typeout{WARNING:~mathml~missing~for~hash\l__math_content_hash_tl}
    }
    \str_set:NV\l_tmpa_str\l__math_content_AF_source_tl
    \str_replace_all:Nnn\l_tmpa_str{&}{&}
    \str_replace_all:Nnn\l_tmpa_str{<}{&lt;}
    \immediate\write\mathhash{^^J
      <div>^^J
      <h2>\string\mml\space\int_use:N\g__math_AF_total_int</h2>^^J
      <p>\l_tmpa_str</p>^^J
      <p>\l__math_content_hash_tl</p>^^J
      <math></math>^^J
      </div>^^J
      }
   \tag_struct_begin:n
     {
       tag=Formula,
@JosephWright could give the str_convert block an xml encoding option to avoid the str_replace :-)
 
Any github action experts around? Beamer has a github action which typesets some files and the doc as test. It worked fine in the past, but for the last couple of commits, it fails on the first attempt (during installation of texlive). If I rerun the action, it works. This now happened for 3 different commits. Does anybody have some suggestions what might be the problem?
Failed 1st attempt: https://github.com/josephwright/beamer/actions/runs/7582977753/attempts/1
Successful 2nd attempt: https://github.com/josephwright/beamer/actions/runs/7582977753/attempts/2
 
@samcarter one just flew in as you posted that:-)
 
@DavidCarlisle Good timing :)
 
@samcarter is that the bad mirror? @MarcelKrüger just changed the latex3 actions to use a curated list of good mirrors to avoid picking up one that kept failing
 
11:29 AM
@DavidCarlisle this might be it. The mirror is different in the second run.
 
@samcarter the second one that worked picked up ctan.mirrors.hoobly.com/systems/texlive/tlnet/…
@samcarter maybe I'm your github action expert, not only the go to guy for language translations
 
@DavidCarlisle you are my 42 :)
 
@samcarter better to be a mouse than a duck
 
@samcarter this cicku mirror is a pain.
 
@UlrikeFischer I guess - this is one of those things that the new hooks mech has made interesting :)
@DavidCarlisle Ask Bruno :)
 
11:32 AM
@JosephWright well generally I think packages should be loaded at the end of preamble and not in the middle of the begindocument mess.
 
@UlrikeFischer Sure, but historically there was only the one hook (in v2 I left it to the user, but that has it's own issues)
 
@UlrikeFischer As the name of the mirror sounds similar to "chicken", maybe David can eat the mirror for lunch and declare the problem solved :)
 
@UlrikeFischer can't we ask texlive and/or ctan to blacklist it from the automatic mirror resolution?
 
@UlrikeFischer One could I think argue that color should always be loaded nowadays
 
@JosephWright or never loaded and use l3 version...
@daleif can I rename the default branch at mathtools to main we did this ages ago for all other repositories under latex3 (and ho-tex) it's just an edit on the site then locally you need to do something equivalent to
git branch -m master main
git fetch origin
git branch -u origin/main main
git remote set-head origin main
 
11:57 AM
@DavidCarlisle If you mave my latest updates, then you should be able to.
@DavidCarlisle Is there an explanation to the version numbers in verbatim.dtx? Can see there is a list of changes at the start of the file, but it does not match the current relealed version. Just do not want to mess something up.
 
@daleif usually \changes entries go at the point of the change (and verbatim has some later) but global updates and some very old entries are collected at the start, it doesn't matter much as they are all sorted by makeindex anyway
@daleif done
 
@DavidCarlisle Yes, we really need this - I have to sort global colours - perhaps one for Mainz ...
 
@DavidCarlisle I've done those step on my work pc and it seems to now be on main. Will see if I can remember it for my PC at home.
 
@daleif git will remind you if you don't :-)
 
12:21 PM
@samcarter why not use the weekly IslandOfTeX container?
Ooh, @WillRobertson is here again! Might become a regular :) Hi, @WillRobertson!
 
@Skillmon I'll just use what's in the main.yaml file. Not enough knowledge about github actions to start experimenting :)
 
@samcarter that one uses a script to install texlive provided by @MarcelKrüger, so blame him :)
 
@samcarter you could ask me
 
@DavidCarlisle You're not too busy eating cheese (or whatever else mice do :)?
 
@samcarter 42
 
12:34 PM
@DavidCarlisle oh
 
@samcarter The main.yml uses v1 of the install-texlive script, if you use v3 you automatically get the better mirror selection (and more automated caching).
@Skillmon Last time I tried that in GH Actions, GH always had to re-download the image and therefore the Actions started very slowly. Not sure if that's still an issue.
 
@MarcelKrüger Thanks for having a look! It now uses v3 and @JosephWright added a friendly bot to keep things updated :)
 
12:58 PM
Looking at a guide for doing PRs (I need to learn this stuff). Why is this step needed after forking the repro and making a new branch: git remote add upstream orig-repro?
 
1:15 PM
@MarcelKrüger depends on how many people use the images and how often they do. Github keeps a cache, afaik, but if you're unlucky and the container isn't cached anymore you have to redownload it indeed.
@daleif if you want to get updates from the project you forked off of.
@daleif you can then fetch/pull from upstream, and push anything you want to merge to your fork. Then you open a PR from that branch to upstream's repo.
 
@Skillmon ahh, thanks.
 
@Skillmon so the typical chicken duck/egg-problem? It is slow because not enough people use it and people don't us it because it is slow?
 
@samcarter partly, yes. That and there is quite some overhead. If you only need core, there isn't much reason to also install every language...
 
@Skillmon Let's make everybody speak English and remove all these languages from texlive :P
 
1:33 PM
@Skillmon I don't know if my experience generalizes, but I recently backed down from using the IoT docker images because it was taking too long. And not only because of caching/re-download, the task itself took longer. See github.com/gusbrs/zref-clever/issues/20
@samcarter Not an expert, but I was also recently having workflow failures at the TeXLive installation step. I don't know what beamer's case is, and it seems you already figured it out. But, in my case, as far as I can tell, the problem occurred because I ran two parallel tasks (for check and doc), and the installation for both in the same cache made the installation fail in one of them sometimes. I introduced a sleep in the quicker task, and things seem to be quiet since then.
 
@gusbrs That wasn't the problem with the beamer action, but good to keep in mind, thanks!
 
1:49 PM
@PauloCereda and @TeXnician worth looking into? (chat.stackexchange.com/transcript/message/65017956#65017956)
@gusbrs I wasn't aware of that, thanks for sharing.
 
@Skillmon I left that issue open hoping for some further feedback. I really like the model of the docker image. But the workflow was taking long enough to be a pain (15-20min for zref-clever and postnotes). So I backed down (for the time being, at least), and also reduced test times by other means.
 
 
2 hours later…
3:25 PM
@UlrikeFischer Will update for this - am now wondering if I should revise other \AtBeginDocument stuff, though much of that is 'after package loading' so I'm relucatnt
 
3:40 PM
@DavidCarlisle, @UlrikeFischer Any thoughts on github.com/josephwright/siunitx/issues/710 beyond 'shame'/'get Joseph to do ConTeXt-like units'?
 
@JosephWright welll.. if you put locally defined robust commands into the aux file then reference them from elsewhere you need to define the commands?
@JosephWright er actually I take that back having unpacked the zip, why is it undefined?
 
@DavidCarlisle because xr-hyper already expand them when reading the aux. If I remember correctly because of the code that adds the external file. Perhaps one could avoid that. But I'm currently trying to write math files ...
 
3:56 PM
@UlrikeFischer I just checked in a new arxiv example that wrote and read the html format, but I have work call now...
@UlrikeFischer that would be wrong...
 
@DavidCarlisle yes it looks a bit odd. @JosephWright redirect the example to hyperref. We should look at that.
 
@UlrikeFischer Thanks
@DavidCarlisle, @UlrikeFischer Thanks :)
 
@UlrikeFischer as the one true xr package works fine I think we can record this as not my fault
 
@UlrikeFischer I've asked the OP to re-report
 
4:12 PM
@JosephWright yes saw. How do I test if a stream has already been allocated?
 
@Skillmon Probably. But we don't have much insight into GitHub actions. And on GitLab, caching works (to the point that was discussed in my TUG talk). A certain kind of caching is simply not supported with the latest images as I outlined at TUG (and we definitely do not suggest pulling the latest image every time if you can avoid it).
 
4:29 PM
@UlrikeFischer You mean at a low level?
 
@JosephWright no, by name. I simply want to avoid that Davids \mathhash stream is opened twice.
 
@UlrikeFischer Check it exists and is non-negative, I think
 
@UlrikeFischer can't you just make it like the aux so you (not the document as now) opens it and you never need to expose the csname with the write stream as a public variable?
 
4:45 PM
@DavidCarlisle I didn't want to open it always but through a key.
 
@UlrikeFischer yes that's what I mean, if you open it if a key is given, and nothing else knows about that stream, it won't get opened twice will it?
@UlrikeFischer the new example worked more or less out of the box (had to disable amsthm again, but no new issues)
 
@DavidCarlisle if people use the key twice .. (I could naturally use a boolean but thought a test is perhaps easier)
 
@UlrikeFischer user error
@UlrikeFischer I'd just check the csname, so l3 version of \ifx\mathhash\undefined \newwrite\mathhash\else \ERROR\fi
@UlrikeFischer ah I see why I did have to duplicate three expressions usual issue of don't write stuff out during trials.... \caption{The figure on the right shows a sample objective function. The figure of the left is the function $ \max \{0,f(60)-f(x) \}$. This function is not identically zero as $x=60$ is not the global minima of $f(x)$.}
 
@DavidCarlisle yes, I was just wondering about the "l3 version" part. \tl_if_exit:NF\g_xx_iow looks a bit odd.
 
@UlrikeFischer you looking for an \iow_if_exists:NF rather than tl ?
 
5:00 PM
@DavidCarlisle that here could work (perhaps with a few more \unexpanded):
\def\XR@@dURL#1#2#3#4#5\\{%
     {#1}{#2}{\unexpanded{#3}}{#4}{\XR@URL}%
  }%
 
@UlrikeFischer or \unexpanded{{#1}{#2}{#3}{#4}}{\XR@URL}% ? (untested)
@UlrikeFischer meanwhile the dospecials list was doing rather too much with white space so I ended up just making a few things safe:
\cs_new_protected:Npn \__math_AF_mml:w#1</h2>#2<p>#3</p>#4<p>#5</p>#6<math{
  \begingroup
  \@makeother\{\@makeother\}\@makeother\#\@makeother\%
  \__math_AF_mml_verb:w{#1}{#3}{#5}<math
}
which would also be easier to write in l3
 
@DavidCarlisle looks ok too.
 
@UlrikeFischer saves worrying about which argument needs protecting:-)
 
6:05 PM
oh no, empty cookie bin
 
6:35 PM
@PauloCereda eat ducks not biscuits
 
oh no
 
@daleif I have a small howto for basic PR workflow here: github.com/circuitikz/circuitikz/blob/master/CONTRIBUTING.md
 
HEy!
Hey david
@DavidCarlisle
 
6:55 PM
@FerriusUndermine yes?
@Rmano have no bugs or missing features then no one will make a PR on your code
 
Hello David.
 
@mickep go away?
 
Hello David
 
@PauloCereda dinner
 
@DavidCarlisle oh no
 
6:56 PM
@DavidCarlisle I'm home, waiting for the crisps... Will stay.
 
@DavidCarlisle I won't tell what we will have for dinner
 
@UlrikeFischer preemptive oh no
 
@PauloCereda dodo and crisps?
 
@DavidCarlisle ooh an allegedly extinct bird (@mickep we ducks are very good at keeping secrets)
 
7:11 PM
@UlrikeFischer @DavidCarlisle Since you have your hands on it, how about xr(-hyper) support for ltproperties? (If I may bump the request... ;-)
 
@gusbrs probably wouldn't take a lot
 
@DavidCarlisle 🎉
 
@DavidCarlisle that's a plan
 
7:59 PM
@DavidCarlisle I see @UlrikeFischer already did it for xr-hyper (github.com/latex3/hyperref/commit/…). :-) (I don't know if it is merged though).
 
@gusbrs I had forgotten that. I merged it now.
 
@UlrikeFischer Great! Thanks!
 
8:18 PM
@PauloCereda Very good indeed.
 
8:37 PM
@UlrikeFischer I may be missing something, but reading the commit I don't see how the imported label would know the "url" of the target property.
 
8:48 PM
@gusbrs hm. I think I need an example.
 
@UlrikeFischer I could provide one. But the point seems straightforward. If you want to build a link to a target in a file different from the current one, you need the "url" to the file to compose the link. \XR@addURL handles that for \newlabel. But I see no similar provision for \new@label@record.
 
9:15 PM
@gusbrs but \ref (for \label) is building the link so needs the url up front but for general properties aren't you just getting the key values and it's then up to you to build a link and you know the url?
 
@DavidCarlisle But you can store target property in a file in a file and then (now) import that with xr-hyper. This is independent of \label (assuming you just set a ltproperties one). How would you build the hyperlink to a target in the external file then?
 
@gusbrs I'm not sure I inderstand, in your main document don't you just do \href{\urltodoc#\extractedPropertyValue} or some such?
 
MWE: A.tex
 
@DavidCarlisle the problem is if he has say two external files and imports their properties. How can he build a target if he doesn't really know if label A (with target blub) if from file 1 or 2.
 
\documentclass{article}

\usepackage{hyperref}

\begin{document}

\section{Section 1}
\label{foolabel}
\RecordProperties{ltp:foolabel}{label,page,target}

\end{document}
B.tex:
\documentclass{article}

\usepackage{xr-hyper}
\externaldocument[A-]{A}
\usepackage{hyperref}

\begin{document}

\makeatletter
\hyper@linkfile{\ref*{A-foolabel}}{\getrefbykeydefault{A-foolabel}{url}{}}{%
\getrefbykeydefault{A-foolabel}{anchor}{}}

% untested anyway, since I don't have the new commit available to start with
\hyper@linkfile{\RefProperty*{A-ltp:foolabel}{label}}{<what would you put here?>}{%
\RefProperty*{A-ltp:foolabel}{target}}
\makeatother

\end{document}
@UlrikeFischer @DavidCarlisle ^^^
 
9:23 PM
@UlrikeFischer ah you want to insert an original-document key into the saved data...
 
@DavidCarlisle we could use something like this:
     \edef\next{\noexpand\new@label@record{\XR@prefix#2}{\unexpanded{#3}{file}{\XR@URL}}}%
      \next
 
@UlrikeFischer in the example that follows @gusbrs knows labels starting A- come from the document A (if you don't use the optional argument you don't know)
 
@DavidCarlisle Why not just store it, as is done for \newlabel, instead of trying to guess from the prefix?
 
@UlrikeFischer can we make xr and xr-hyper be the same file at the same time:-)
 
@DavidCarlisle yes, but that is similar with the normal xr-hyper. Thats why it add \XR@URL in the fifth argument. With the command above you get in the property the values {counter}{}{abspage}{1}{file}{suppinfo1.pdf} and so you know the file. The main question is which property name one should use? file? url?
 
9:29 PM
@gusbrs because it maintains the truth of my earlier statements? (which is clearly more important than actually fixing your problem:-)
 
@DavidCarlisle I should have known. :-D
 
@UlrikeFischer xr-url so it's specific to this and not clashing with any existing generic url property?
 
@DavidCarlisle or xr-file as it is not really an url?
 
@UlrikeFischer everything is a url if you squint at it the right way, but OK, file
 
@UlrikeFischer I never understood why "url", "file" seems more to the point. But then I've never seen a use to this property other than xr(-hyper). Perhaps I miss some other intended use case.
 
9:32 PM
@DavidCarlisle I'm fine with both.
@gusbrs you can actually add as optional argument a full url. E.g. with \externaldocument{myfile}[https://abc/blub/xxx]
 
@UlrikeFischer yes which was the main point of xr-hyper over the original xr
 
@UlrikeFischer Ah, I see.
 
I think url is better really as that's definitely a url but may not be a file
 
@DavidCarlisle ok so xr-url. I will add that.
 
@UlrikeFischer Thanks again!
 
9:38 PM
@UlrikeFischer I tried an arxiv example from theoretical physics rather than math, hadn't noticed before (or I had forgotten) that amsart has a faked almost but not quite amsthm which makes it harder to work round than the previous examples
 
 
1 hour later…
11:03 PM
user image
2
ooh cowculus
 

« first day (4828 days earlier)      last day (88 days later) »