12:50 AM
@JosephWright THAT makes sense! That would totally explain why it changed when I updated LaTeX. \ang works fine though. :D

Only 10,010 followers away en.wikipedia.org/wiki/616_(number)
616 (six hundred [and] sixteen) is the natural number following 615 and preceding 617. While 666 is called the "number of the beast" in most manuscripts of Revelation 13:18, a fragment of the earliest papyrus 115 gives the number as 616. == In mathematics == 616 is a member of the Padovan sequence, coming after 265, 351, 465 (it is the sum of the first two of these). 616 is a polygonal number in four different ways: it is a heptagonal number, as well as 13-, 31- and 104-gonal. It is also the sum of the squares of the factorials of 2,3,4. i.e. (2!)² + (3!)² + (4!)² = 4+36+576=616. The 616th harmonic...

5 hours later…
5:53 AM
@FaheemMitha there is: Don't use it on Windows. My personal experience is that on Linux the noise is pretty low, but as soon as I try to benchmark on Windows the measurements are very imprecise.

6:26 AM
Gotta love your well thought out variable names... :)

6:47 AM
@DavidCarlisle @StefanKottwitz wrong @St ping

7:32 AM
http://i.stack.imgur.com/taTzP.png
Am I the only one who read that ^^^ Wikipedia passage and then "5 hours later..." in Spongebob Narrator Voice XD
2

@Skillmon Thanks for the tip. I don't use Windows. And I've never used TeX on Windows, as far as I can recall.

@Plergux yes

@DavidCarlisle good, then I'm still weird. :p

@FaheemMitha I used to use it seldomly on Windows (lately a bit more often, as my work pc is a Windows), but I wanted to see whether the implementations on the different OSes differ in performance of a few things... Needlessly to say, that was futile, no reliable results were obtained on Windows.

@Skillmon I wouldn't expect the timings across different systems to be very informative, but do you mean that even on windows you couldn't get stable comparative times?

7:36 AM
@DavidCarlisle the latter, the noise on windows was too high for me.

@Skillmon I have never used the native windows tex but I wonder how it works on cygwin it's using the same underlying system clock i assume, Never really studied the timing output to see how stable it is...

@DavidCarlisle I guess it could also be because Windows has more processes spawning in the background than my minimalist Arch Linux installation.
@DavidCarlisle in l3benchmark Bruno normalizes the times to ops (I have no idea what exactly is considered 1 op(eration) but the ops are pretty consistent for me, even if my Linux has more load and therefore the timings differ between runs).

@Skillmon I haven't looked but I assumed the op count was counting expansions or whatever so not time based at all. I wonder if l3benchmark has any documentation:-)

@Plergux Hi, long time no see! Welcome back!
Hmmm... how long should I wait for luaotfload | db : Reload initiated (formats: otf,ttf,ttc); reason: Font "URWClassico" not found.? It's on the long side of 5 minutes...

@Rmano not that long normally, have you set it off searching over some slow network-mounded drives?

7:44 AM
@DavidCarlisle I doubt it :)

@Rmano Hey there! Thank you. I took a bit of a break from everything, went to a summer house, had a twentieth anniversary weekend getaway, you know, the usual. :p

@DavidCarlisle No, I just added Classico in my local texmf tree and tried to compile tex.stackexchange.com/a/596987/38080
   PID USER      PR  NI    VIRT    RES    SHR S  %CPU  %MEM     TIME+ COMMAND
19569 romano    20   0  348812 330696   6476 R  99.7   4.1   6:43.00 lualatex

@Rmano blame @UlrikeFischer

@DavidCarlisle It doesn't finish... will wait until @UlrikeFischer can suggest some debugging thing. But I suppose I have to remove classico to have back lualatex...
It seems to be in a very tight loop. No CTRL-C...

@Rmano well it depends on what you really compile. if you just triggered the fira fonts 5 minutes are quite normal ... some font families are insanely large.

7:52 AM
@UlrikeFischer Is there a way to have more info? It's 12 minutes now, and it does not answer to ctrl-c
Yep, it answered to ctrl-c. It stopped, so maybe I just need to be more patient...

@Plergux Yesterday's Earth Observatory Image of the Day may be of interest to you.
2

Ok, it's working now. 13 minutes at 100%
@UlrikeFischer Sorry for the noise

@HaraldHanche-Olsen ooooh! Cool :) I think about half of that was not there when we went there (the south east part that looks like a bone).

@Plergux I was watching at an Australian-Polish-living-in-Iceland girl vlog about the volcano ;-) she was so insanely happy that was contagious...

@Rmano lol. Like Spongebob. :p

8:02 AM
@Plergux Yes! Similar approach to life, I like it

3 hours later…
10:40 AM
The documentation for l3benchmark says:
> In LuaTEX only the CPU time is measured, while in other engines real time is measured (including time waiting for user input).
However, I'm using LuaTeX and want the real time.
Does that mean l3benchmark is not useful for me? If so, what are my options?
This is what l3benchmark is reporting: 3.18 seconds (9.3e6 ops)
Actual timings from time:
real    1m15.567s
user    0m56.407s
sys     0m16.478s
Big difference.
Even sys is much larger than l3benchmark.

11:04 AM
This is strange and unexpected, but it looks like the problem is latexmk!
I should try a newer version.
This thread tipped me off.
9

SOLVED: see bottom of message! I sensed pdfLaTeX was slower in TeX-Live 2020 and decided to run some benchmarking… Computer: MacBook Pro (2018) with SSD Document: The "novathesis" thesis template LaTeX Version(s): MacTex 2019, 2020, 2021 (all versions up to date) The chart shows the execution t...

It seems that latexmk is part of TeX Live, but on Debian it's a separate package.
I did a search, and also checked the latexmk oackage documentation on Debian, but did not find anything. Does anyone know why?

@Rmano Every day is the Best Day Ever! :)

11:35 AM
@Plergux That sounds statistically unlikely.

@FaheemMitha well it's not about statistics, it's about perception. If you have the mindset for it every day has the best of something that was never before this good. And some times it's simply the best day ever because you are still alive and kicking for another day.

@Plergux Being alive is definitely good. That part I can agree with.

@FaheemMitha I try to improve myself every day, so I can be a better me today than I've ever been. And every day I can learn something new is the best day ever. :p And it's also relative. Once it's past you can tell yourself it's of no consequence, and you don't know what's going to happen in the future, so that is of no consequence either. Making each day the best day because it is the only day that matters. :p
#SpongebobPhilosophy :p

@Plergux You should go into the motivational speaking business. :-)
@Plergux Spongebob?

@FaheemMitha Aw, thanks :D
@FaheemMitha Squarepants!

11:46 AM
@Plergux The cartoon character? I don't see the relevance.

@FaheemMitha That's a full run, not a single piece of code - l3benchmark is testing code inside a run (a lot of the startup is loading the format/packages, etc., not doing expansion/execution)

@FaheemMitha Rmano and I were talking about people who are amazingly cheerful which led me to think about Spongebob (that is his character) and the phrase "Best Day Ever" is actually a song that Spongebob sings about how it's the best day ever, just because the grass is still growing and the sun is still shining and he has friends to spend his time with (and he would most likely sing it every day because that's just the way Spongebob is. :p)

@JosephWright Yes, I'm running my file to see how long it takes. What's wrong with that?

@FaheemMitha Nothing wrong per se, but l3benchmark is for testing 'blocks' of TeX code, and that doesn't include the entire format loading, etc.: it's not the same as time on Linux

@JosephWright Well, OK. But I'm still just trying to test my code. I started with the whole file because I wasn't sure where the problem is. Turns out the file isn't the problem at all. It's latexmk. :-)
But the documentation says that it doesn't do actual timing for LuaTeX, which isn't ideal.
@Plergux I see. I listened to a bit of that on Youtube. It's quite noisy. Probably suitable for 3 year olds.

11:57 AM
@FaheemMitha I have the mental maturity of a three year old :p I love noisy. Colour and noise. That's me. :p

@Plergux I think my mental maturity is around 12. Give or take. Being an adult kind of sucks. It's not much fun.

@FaheemMitha Really? I would expect the behavior described in the documentation. If lualatex asks me for input in my benchmarking code while I grab a coffee, I don't want that time counting towards the time measurement of my code. Except you want to benchmark how long it takes you to grab a coffee ;)

@TeXnician Um, not really following that. Why would LuaLaTeX ask me for input?

@FaheemMitha There are various way in TeX to prompt for user input which you could of course use in a function under test. And the docs you cited (“In LuaTEX only the CPU time is measured, while in other engines real time is measured (including time waiting for user input).”) have this as a prime example. I think CPU time makes sense for this kind of component benchmarking where interaction is usually not what you want to benchmark.

@TeXnician Well, if the code is pausing for some reason in the middle, it would be nice to know that.
Though that would be very unusual behavior for TeX.
As it turns out latexmk is doing "something". I'm not sure what.
But if I run LuaLaTeX directly, there's no problem.

12:02 PM
@FaheemMitha Growing up is entirely optional. :p I've tried it and it's not for me. :p So I'm gonna be a three year old pretending to be a grown-up until forever. :p

But latexmk pauses big time. For no apparent reason.
@Plergux Sounds like a plan.

@CarLaTeX I'm not sure what you meant by Overleaf using tikzducks for logo. We don't use it for the logo per se, we just love the ducks so place them wherever we can :-)

@FaheemMitha Yes, but take for instance OS process scheduling. If you are low on resources processes of high priority might cause normal/low priority processes (like TeX) to pause more often than usual. These things are not controlled by the TeX code that is tested so I see no reason to include the time that the OS is doing stuff including pausing processes. As I said, probably depends on your needs for benchmarking.

@TeXnician True, if other stuff was going on, timings could be all over the place.
But why does LuaTeX behave differently?

@FaheemMitha The point really is that l3benchmark is trying to do a particular job, part of which is being at least qualitatively comparable between engines
@FaheemMitha We have access to different stuff

12:10 PM
@JosephWright OK. Looks like I won't be really using it, after all. Other than what I already did.

1 hour later…
1:13 PM
@yo' Yes, "logo" was not the correct word. I meant ducks in the messages etc.

@CarLaTeX "mascot" maybe :p

@FaheemMitha I'm surprised you can even do the whole file? benchmarking executes the code multiple times to get stable measurable results and you can't normallt do that with the whole file

@Plergux That's the correct word!

@Plergux yep, mascott would work :) And the precise date is actually October 2019, this shall be the first tweet: twitter.com/overleaf/status/1186750838422822912 @CarLaTeX

@yo' So it's more or less 2 years!

1:24 PM
@CarLaTeX yep!

@yo' yay!

@DavidCarlisle Well, the brackets were still inside begin and end document.

@DavidCarlisle very likely in fact. Although why people use linguex ...

I mean, the macro was enclosing whatever was inside the document.

1 hour later…
2:41 PM
@DavidCarlisle Fixed.

3:01 PM
@AlanMunn thanks, I could see the guesses were wrong but couldn't see (in time available) why I couldn't just set it to a wide enough length

@DavidCarlisle Although you were joking about blaming me, I think I discovered the problem originally with Garamond and helped the author with a solution. Unacknowledged it seems.

3:39 PM
@AlanMunn I did know that you knew the package so it wasn't a totally random blame:-)

@DavidCarlisle I really dislike it though, since it has no proper semantic markup really. And the main macro is \def\ex.#1\par{...} But lots of people use it.

@AlanMunn they want to look like they are real linguists like us, perhaps?

@DavidCarlisle That must be it. (The poster of the question is definitely a real linguist though.)

@AlanMunn I did notice the weird \ex.[ syntax in the test example

@MarcelKrüger your pdf root attachement seems to be broken.

3:49 PM
@UlrikeFischer It's weird, when I open the mail I sent again the attachment is fine, but the copy received form the mailing list is broken.

Did the website change its font these two days?

@Symbol1 Yes. There's a main meta site post about it. meta.stackexchange.com/questions/364048/…

@MarcelKrüger the mailing list is weird. In one of the other groups it ate all + symbols.

@UlrikeFischer If I base64 encode the broken attachment I get the correct PDF file without all special symbols...

@MarcelKrüger you could upload the pdf to the github.

4:19 PM
@MarcelKrüger hm, it looks like a good example why marking the stream as mathml is probably not the right idea ... but I think the root must be real content, as artifact you would break copy & paste. Perhaps this is one example where "NonStruct" is of use.

@AlanMunn -- Maybe a real linguist but not a typographer.

4:32 PM
@UlrikeFischer I also think that it should be real content, but Boris thinks it should be an artifact. But I don't see how NonStruct would help though since children of NonStruct elements behave like children of the parent of the NonStruct, so I would expect that they have the same issues as using direct children.

@barbarabeeton Actually he's very particular about typography, but in this case I suspect it's a publisher's requirement he's trying to meet.

@UlrikeFischer in an ideal world you'd be able to copy and paste it as the mathml though?

@DavidCarlisle well I think that there will be people who don't care about mathml and simply want to copy&paste √𝑝−𝑞.

@UlrikeFischer sure but it's a lot easier to downgrade the ml to plain text than the reverse (and the clipboard can have both) If you paste mathml into word as mathml for example it will end up as a math zone and editable as such, if you paste it in as plain text then.... people may not think they care about mathml but they may expect the math to stay as math:-)

4:49 PM
@DavidCarlisle well yes, but this looks only at the whole math. I meant, what should happen if people select only a symbol like the root and try to copy it? It shouldn't disappear.

@UlrikeFischer what happens if you try to select a fraction bar?

@DavidCarlisle we could disallow fraction bars and root and tell people to use ^{-1} and ^{1/2} ;-)

@UlrikeFischer If the root is tagged it could be copied as <msqrt /> if MathML c&p is supported.

@MarcelKrüger yes, I see the point. But it relies on "if mathml c&p is supported" and so on a rather vage future idea ;-). What does e.g. adobe do now if you mark the root as artifact?

5:05 PM
@UlrikeFischer forgetting about mathml for a bit isn't the eventual idea that cutting and pasting from a tagged pdf should give you marked up html on the clipboard not just plain text (even if that doesn't work now)

@MarcelKrüger how did you create the bdc-marks for the math?
@DavidCarlisle at least as an option probably yes, but I don't know how much it already does.

@UlrikeFischer Basically the code runs \tag_mc_begin:n{...}, saves the attribute values and then runs \tag_mc_end:, afterwards it modifies the attributes of the affected nodes to the saved ones. It's rather hacky but it works at least for simple cases.
@UlrikeFischer See the pdfstructelem branch (but there isn't a compatible example file yet)
\RequirePackage{pdfmanagement-testphase}
\DeclareDocumentMetadata{uncompress,pdfversion=2.0}
\documentclass{article}
\usepackage{unicode-math,tagpdf,amsmath,luamml}
\tagpdfsetup{activate-all,interwordspace=true,}
\ExplSyntaxOn
\let\BLUB\luamml_flag_structelem:
\ExplSyntaxOff
\begin{document}\tracingmathml=2
\tagstructbegin{tag=Document}
\tagstructbegin{tag=Formula}
$x_{1/2}=-\frac p2\pm\sqrt{\frac{p^2}4-q} \BLUB$
\tagstructend
\tagstructend
\end{document}

5:46 PM
@MarcelKrüger hm, I think my slides will end with "and next time in this theater: Math tagging with Marcel"

1 hour later…
7:10 PM
Hi, my Travis build stopped working, with:
tlmgr: fmtutil-sys --byfmt cont-en --no-error-if-no-engine=luajithbtex,luajittex,mfluajit --status-file=/tmp/inyNnIEbeZ/fF_KBo2cqJ failed (status 255), output:

Unknown option: status-file
Has anybody any idea on what can be going awry?

7:34 PM
@Rmano what is --status-file ?

@DavidCarlisle no idea, it is doing it a tlmgr update --self --repository http://mirror.ox.ac.uk/sites/ctan.org/systems/texlive/tlnet/
Well, sorry, have to go now... will try again tomorrow

@Rmano oh I misread, it's a fmtutil-sys argument not tlmgr, OK

Probably a glitch, will try again tomorrow.

2 hours later…
9:38 PM
@UlrikeFischer pageslts :(

@DavidCarlisle it is not ours ;-)

@UlrikeFischer not yet

@DavidCarlisle hopefully never, I don't even understand what it does and the docu has a chapter "a few warnings" with 8 sections.

10:05 PM
@UlrikeFischer probably enough to ignore it all apart from first line of the abstract which says it makes a LastPage and VeryLastPage \label and do the same with th epage hooks, and call it done

@DavidCarlisle something like that, we should get rid of these external last page packages anyway.