@DavidCarlisle Are you underseas?

Hi Chat!!

> Google Is Shutting Down Its Allo Messaging App, Says Report

@marmot He is overseas, as far as I'm aware.

@marmot no, in a deep burrow.

Blog move done: texdev.net

Just a few DNS issues to iron out

@JosephWright I get a "no certificate" message.

@JosephWright ooh

@UlrikeFischer Yeah, like I said, a few issues :)

@JosephWright and then a "page has redirected to often".

@JosephWright well you show a link so I click ;-)

@UlrikeFischer At present I have to use http://, which then redirects to https:// ...

@UlrikeFischer I think it's a question of working on the DNS settings for a few hours ... this sort of thing is a bit tricky

@UlrikeFischer Warning: Potential Security Risk Ahead sounds more exciting than the usual run-of-the-mill Tex stuff.

@DavidCarlisle I think it's a fontspec error

@PauloCereda blame Will

@DavidCarlisle ooh

@DavidCarlisle yes but the following "this page doesn't work" is then a bit disappointing.

@UlrikeFischer It's so hard to get past that error in firefox nightly (and it sends mail to mozilla each time if you do) that I didn't try.

@UlrikeFischer, @DavidCarlisle, @PauloCereda I think https://texdev.net should now work, but at present the www version doesn't .... not entirely clear exactly what DNS settings I should have (GitHub's instructions are a bit more vague than is ideal).

@JosephWright yes. For souldern.org in the end I made the www one the canonical one so both souldern.org and www.souldern.org work with https but the former redirects to the latter.

@DavidCarlisle Could do with some clue as to the DNS setting you use :)

@JosephWright I still get "www.texdev.net hat Sie zu oft weitergeleitet." (yes the www is added to the message if if I try to open texdev.net).

@DavidCarlisle Of course, now I re-read GitHub's advice, I should have picked a registrar that supports ALIAS ...

@UlrikeFischer Firefox is doing that for me, so I'm using http://texdev.net and getting redirected to https ... but Edge and Chrome seem happy not to add the www

on github I just tick the enforce https and specify www.souldern.org in its CNAME field and at godaddy (where I don't have access) we just set up the CNAME as far as I know to point to www....

@DavidCarlisle Ah, right

@DavidCarlisle I'll try that then

Bring back the old doctor

@PauloCereda Huh?

@JosephWright Doctor Who :)

John SimCARD Simm could be the doctor.

Or bring back the bloke with the cool scarf

Or promote the Pond girl

What's her name

:)

@DavidCarlisle We'll see if that works: currently no certificate so not accessible at all ...

@JosephWright it works now for me if I ignore the certificate message.

@JosephWright github should automatically set up a certificate via letsencrypt as far as I understand it (I didn't do anything special about that)

@JosephWright for me texdev.net and texdev.net both work now (firefox nightly) both redirecting to www form.

@DavidCarlisle Yay!

@DavidCarlisle Good, all working and costing £10/year :)

@JosephWright how much would your old site have cost if you had not switched?

@DavidCarlisle Nearly £200 due to CPU usage

@DavidCarlisle, @UlrikeFischer, @egreg I think a blog post or two on the 'new' XeTeX primitives and \expanded is coming up :)

@StefanKottwitz Hope you like my new blog hosting :)

ctan.org/pkg/changelog looks interesting (@DavidCarlisle!)

@JosephWright Clean and fast! I think I would like blue bars instead of gray bars even more, matching headings and link colour.

@StefanKottwitz I'm probably going to make a post saying 'design ideas or even better contributions welcome': I'm no expert, and just used what I could pick up (basically the LaTeX Project site design)

@StefanKottwitz Blue bars might work nicely, I agree

@StefanKottwitz Yes, speed is very good with a static site :0

Blog title fixed :)

texdev.net/2018/12/06/a-new-primitive-expanded

@DavidCarlisle puh oberdiek update was just in time: tex.stackexchange.com/questions/463504/…

@UlrikeFischer :-)

@DavidCarlisle satisfying ;-) But it naturally shows that the problem with the resources should be handled in the near feature @JosephWright.

@UlrikeFischer Yup

@StefanKottwitz Changed

@StefanKottwitz Wonder if it's a bit too bright ...

@JosephWright at least the bottom may be too bright

@StefanKottwitz Hmm

@StefanKottwitz Happy to take suggestions!

top bar: nice, bottom bar: can be like top, why not

@JosephWright rgba(0, 93, 183,.85); ? (same colour but toned down a bit)

@DavidCarlisle Er, can you give that in HTML format?

@JosephWright that should work in the css, e.g. background-color: rgba(0, 93, 183,.85);

@UlrikeFischer I'll try it then :)

@JosephWright if everything gets pink, blame David ;-)

@UlrikeFischer Looks OK

Thus-far, Jekyll/GitHub pages works well :)

@DavidCarlisle, @UlrikeFischer I've tipped off Karl/@barbarabeeton that I might have some content for an article

@JosephWright that's what I did:-)

@JosephWright hm

@UlrikeFischer translucency, we would naturally blame you

@DavidCarlisle ooh

@JosephWright You haven't got people sending you content as Word files though;-)

@DavidCarlisle :)

see the picture we posted yesterday on souldern.org I could post xii.tex to go with it

@DavidCarlisle ooh season

Poor turkeys

@PauloCereda maybe we'll have duck instead

@DavidCarlisle oh no

@DavidCarlisle I see you have the dwindling chapel congregation problem too ... out Baptists closed recently

@JosephWright since there are less than 200 houses in the village it's tricky to keep these things going, of course in former times every family did the honourable thing and had 10 children and kept the sunday school busy....

@DavidCarlisle oh

@DavidCarlisle :)

Oh great I have write access to another gh repository full of packages I know nothing about:-)

@JosephWright -- looks promising to me.

@DavidCarlisle ?

@UlrikeFischer github.com/wspr/herries-press

@DavidCarlisle oh, and a number of new issues from Frank ...

@UlrikeFischer i applied my usual spam filter and didn't notice those

@DavidCarlisle Marmot burrows are not deep enough to claim that one studied things at great depth.

Anyone know why enumitem started recording fontsizes and why this is not done at begin document? This just broke one of my classes because I load packages first and define fontsizes later.

@daleif it's mentioned in the change list here github.com/jbezos/enumitem

@daleif what does it actually do?

@DavidCarlisle It runs this at load time

IMO that is too early.

I just added an issue on it

@daleif yes seems so, shouldn't assume any fonts are set up before begin document, ping Javier in github I suppose

@daleif good plan:-)

@daleif didn't you want to say that the user "will have to remember to do this before loading enumitem"?

@UlrikeFischer fixed

@daleif what did break in your class? Only enumitem recording font sizes can't do it.

@UlrikeFischer My class is written from scratch, and I load packages first and define font sizes later, thus when loading enumitem the font sizes are not defined (as there are no fall backs in the format). Probably a bad idea on my part, but this the time of recording these values is bad

this the time => here the time...

@daleif packages should probably be able to rely on \normalfont/\normalsize, but the timing is certainly wrong. I'm also not sure if it is a good idea to assume that \tiny or \Huge is defined.

@UlrikeFischer I only ever use the hsl colour model on the web these days. I find that the result of tweaking colours is much more predictable that way. (But not too predictable: One tends to forget that the L component must be set to 50% for maximum saturation. Higher, it fades into white. Lower, it fades into black. (@JosephWright)

@JosephWright Congratulations on the move. Do you plan to still have an RSS (or Atom) feed? The feed is how I used to track the site in the past.

@HaraldHanche-Olsen I don't need to remember this. I simply use the colors knowledgable people like @samcarter or @DavidCarlisle tell me to use ;-)

@HaraldHanche-Olsen Feed should still be there

@HaraldHanche-Olsen texdev.net/feed

@JosephWright interesting page - completly black ;-)

@UlrikeFischer huh? I get an XML file, which shows up fine in Firefox

@JosephWright or with the firefox livemark addon:

@JosephWright with chrome or firefox I see something, with opera I have to add the link as a source and then can view it in the "personal news", in a browser tab it is simply black.

@DavidCarlisle Yup

@UlrikeFischer I've not got Opera here ... I guess it doesn't understand XML feeds

@UlrikeFischer opera wow do people still use that I never used it after they switched to using chrome internally

@JosephWright probably, at least not directly in a browser tab, the look is fine in the in-built-rss reader (which I only discovered now, normally I try to ignore all this buttons).

@DavidCarlisle I got used to it, and I have no problems with it that would drive me away currently (and managed to fine tune the site here with it). I have to use firefox or chromes sometimes but that's all.

A case for l3benchmark :) tex.stackexchange.com/questions/463554/…

@JosephWright of course there is also the issue as our tag line says "results matter" speed isn't the only thing that should be tested there:-)

@DavidCarlisle I did think of adding that the L3 FPU is properly tested ...

@DavidCarlisle what matters are no "dimension too large errors". I have again a plot that fails and this is a nuisance ...

@UlrikeFischer Yup: if/when I sort a full TikZ stack ... we might solve it :)

@JosephWright just realized that \g_benchmark_time_fp is gone, it is now \g__benchmark_time_fp. How is one supposed to print to create nice tables?

@UlrikeFischer :)

@UlrikeFischer It's still experimental ....

@UlrikeFischer I imagined this mainly being useful to developers in interactive mode .... but we could add it to be used in typesetting (and thus returning values)

@JosephWright I could also stop to well-behave and use \g__benchmark_time_fp ;-).

@JosephWright how do you print the first one

\documentclass{article}
\usepackage{fp}
\usepackage{xfp}
\usepackage{tikz}
\usepackage{l3benchmark}
\ExplSyntaxOn
\cs_new_eq:NN \benchmark \benchmark:n
\ExplSyntaxOff
\FPmessagesfalse
\newsavebox{\testbox}
\begin{document}

\benchmark{\sbox{\testbox}{\FPupn\result{2 2 root 40 sin *}}}
a \usebox\testbox

\benchmark{\sbox{\testbox}{\fpeval{sqrt(2) * sind(40)}}}
b \usebox\testbox

\benchmark{\sbox{\testbox}{\pgfmathparse{sqrt(2) * sin(40)}\pgfmathresult}}
c \usebox\testbox

\benchmark{\sbox{\testbox}{\directlua{tex.print(math.sqrt(2) * math.sind(40))}}}

@JosephWright Ah, I used to get it from texdev.net/feed/ (note final slash). When I drop the final slash, I do get it.

@DavidCarlisle by adding \result in the box?

@UlrikeFischer I assumed it was somewhere:-)

@JosephWright @UlrikeFischer hmm that;s interesting, gets a different number, looks like it is missing a degrees/radians

@DavidCarlisle I had a head start: I sent a similar test to the spreadtab maintainer (and he now considers to add also xfp as calculation method ...).

@JosephWright as I said speed isn't the only thing, results matter:-) ^^^

@DavidCarlisle 2 2 root 40 180 pi / * sin * perhaps :)

@DavidCarlisle I didn't find a sind operation...

@DavidCarlisle I'll be honest, I didn't check that as I was focussing on performance

@JosephWright \benchmark{\sbox{}} ?

@DavidCarlisle :)

@DavidCarlisle, @UlrikeFischer I think I'll write another blog post: something on this would be good I suspect (the general 'Floating point calculations in LaTeX')

@JosephWright the lua one is quite fast, even though it has to pas the string into lua and back

@DavidCarlisle Of course

@JosephWright do you want to mention apnum or ignore it?

@DavidCarlisle Would be slower if you included pre-parsing for e.g. dimens

@JosephWright or waiting for luatex to start

@UlrikeFischer For a proper post, I should include it, fltpoint, ...

@DavidCarlisle, @UlrikeFischer I'll draft something up: probably a blog post with the aim of a TUGboat (@barbarabeeton): floating points are useful

@UlrikeFischer There's also xint

@JosephWright i really should do something about \Gscale@div possibly something as simple as if expl3 loaded define it using l3fp ...

@DavidCarlisle Sounds like a plan

@DavidCarlisle Let's be honest, we are pretty close to biting the bullet and just loading expl3 as part of the 2e kernel anyway ...

@JosephWright ;-). I just wanted to suggest it too.

@JosephWright is there a ready packaged division of two dimens in expl3?

@DavidCarlisle \dim_ratio:nn?

@UlrikeFischer sounds like the right sort of thing, thanks...

Hmm, xint doesn't do transcendental stuff (no sine) ... and it does more-or-les point to l3fp

@UlrikeFischer We did discuss it ... I think if we can sort out the rest of the 'move to stable' stuff we might be ready ... or we could just \IfFileExists{expl3.sty}{\RequirePackage{expl3}}{} at the end of latex.ltx

@JosephWright tabularx and tabulary would benefit from that as well, especially tabulary that tries to divide up column widths so they add up to textwidth, rounding erros have caused real world issues before now

@DavidCarlisle With the L3 box stuff, I switched from 'keep it fast, use dimens' ages ago as in the end the FPU is fast and the reliable accuracy is good

@JosephWright in l3box assuming l3 loaded seems reasonable

@DavidCarlisle To be honest, I think there's no contest: really, tieing our hands behind our back is daft. There is no reason not to use expl3 for this stuff.

which reminds me, we are overdue for a release...

@DavidCarlisle Speaking of which, I'll update l3kernel once I get home (I'm in the city)

I do like the description of the LaTeX2e date format :)

@DavidCarlisle, @UlrikeFischer, @egreg (others) I'm think of a blog post more-or-less saying 'Use xfp for calculations'. There are several options out there, but really end users don't need another list of various options, they need an answer ... thoughts?

@JosephWright sounds reasonable (I suspect one day it would be more reasonable to have floating point arithmetic in the engines, but not for a while)

@DavidCarlisle LuaTeX has it, there is no likelihood of any other engine changing to add it

@JosephWright the tex model of doing everything with integers was great for an era where machines had wildly different arithmetic models, but I doubt anyone is using tex on a machine without hardware IEEE arithmentic these days

@DavidCarlisle Well yes ...

@JosephWright well, that depends....

@DavidCarlisle Nicola's son?

@JosephWright no that's on an arm chip, which has ieee doubles

@DavidCarlisle Erm, other than LuaTeX there's only one person who is not Japaneses adding stuff to engines ...

@DavidCarlisle Phew

@JosephWright funny you should mention that.

@DavidCarlisle I've been pondering catcode tables in XeTeX ...

@DavidCarlisle Yeah ...

@JosephWright it's sort of fast enough but doing several thousand tex macro expansions to do something that in a pipleline floating point register operation probably takes one or two clock cycles just seems somehow wrong. especially when people start plotting 3d surface fits rather than just scaling a couple of images

@DavidCarlisle Sure, but you'd still need some of l3fp for the register conversion and the like

@JosephWright yes of course

@DavidCarlisle That' why pgfplots has some Lua-specific code .... if we went the ConTeXt way, l3fp could look a lot shorter

@DavidCarlisle We could have a separate Lua codepath

In other news ... do you think we can sell l3build to O2 and Ericsson ... they don't see to know about unit testing

@JosephWright all our phones are o2...

@DavidCarlisle I'm on Giffgaff ...

@DavidCarlisle Did consider picking up a backup SIM this evening ...

@JosephWright which is same thing i think? Since I only use my phone for about 30 seconds a day as I leave the office, and I've been working from home this week as I've had no car, I could have gone several days without noticing:-)

Hey! Quick question

How can we place |- relation in TikZ?

@DavidCarlisle Yes, owned by O2 but much cheaper

I have (A.west) |- node[above]{text} (B.east) but I want to move |- to the left/right

@UlrikeFischer hmmm

@DavidCarlisle ? Doesn't it work?

@UlrikeFischer define work, it returns 8060928/8192000

@UlrikeFischer I suppose wrap in \fp_to_decimal:n

@DavidCarlisle yes that's what the documention says.

@UlrikeFischer if you say so:-)

@DavidCarlisle ^^^^

@UlrikeFischer @JosephWright @UlrikeFischer

\documentclass{article}

\usepackage{graphicx,expl3}
\begin{document}


\makeatletter


\def\divtest#1#2#3{%
\Gscale@div\@tempa{#2}{#3}%
\typeout{div #1: \@tempa}}

\def\sintest#1#2{%
\CalculateSin{#2}%
\typeout{sin #1: \UseSin{#2}}}

\def\tantest#1#2{%
\CalculateTan{#2}%
\typeout{tan #1: \UseTan{#2}}}

\scrollmode
\def\test{
\divtest{A}{123pt}{125pt}
\divtest{B}{123pt}{.1pt}
\divtest{C}{12345pt}{.01pt}

\sintest{D}{90}
\sintest{E}{91}
\sintest{F}{0.1}

\tantest{G}{90}
\tantest{H}{45}
\tantest{I}{0}

@manooooh Not sure I understand, are you asking how to move the corner of the path?

@DavidCarlisle looks good. You only need to handle the tan 1 error.

Please help to reopen tex.stackexchange.com/questions/462533/…

@samcarter Surely this is a duplicate of some other beamer question?

@samcarter Done (welcome back, BTW :)

@AlanMunn Yes, that's why I'd like to mark it as such. (an add some details for this specific situation)

@PhelypeOleinik Thanks!

@samcarter Ah, ok. I forgot that you have to open it to close it again as a duplicate. I've added my reopen vote.

@samcarter -- opened. (and welcome back from me too.)

@barbarabeeton Thanks! I'm glad to be back -- I missed the friendly members of the chat room!

@PauloCereda I'm not surprised a duck wants the pond girl as doctor :)

@samcarter Welcome back!

@CarLaTeX Thank you!

DEK would touch wood if he read no. 10 here: tex.meta.stackexchange.com/a/7948/101651

@samcarter where did you go??

@TorbjørnT. yes? I wanted it to partially solve this question: tex.stackexchange.com/questions/462914/…

@manooooh Actually, I don't understand what you want to do, unless you mean something like tex.stackexchange.com/questions/45347/…

@manooooh TeX.se was so nice to send me on a week long holiday :) I made a lot of nice pictures

@TorbjørnT. look:

@manooooh Right, see the link I posted.

@samcarter I do not know if I should consider that you work formally for TeX.SE as to receive holidays or not but welcome back! It seems that the duckling has had a lot of fun :D :D that's great

@TorbjørnT. I do not remember that we have to add a (new) library for that, it was simpler. Are you sure that we do not have an easier way to achieve it?

In the last image of the link I posted you can see that the part which contains |- is sticked to Entity C (it should be in the middle of the path). What do you think I am doing wrong?

@manooooh That was just a joke, I don't work for tex.se (if I would, the design of the site would not be so broken ...)

@manooooh There were a couple of other answers there, did you look at the accepted answer? Or for example tex.stackexchange.com/a/117281

@samcarter haha it is cute

@TorbjørnT. nice, \draw[dashed,->] (A.south) |- ($(A)!.7!(B)$) -| (B.north); is what I more or less remember, thank you

It was -| ++($(-1.25,0)$) |- (changing different values) lol

Forget that omg, -| ($(C.west)!.5!(C2.east)$) |- works better!

@samcarter Welcome back!!! (And enjoy your holidays. ;-)

@marmot Thanks! I'm currently working on the last-overlay question. Instead of one slide missing like in your answer, I get one slide too much. Do you think it is OK if the mean value of both answers gives the correct number of slides? :)