@UlrikeFischer quicker but shorter than me :-)

quack

@PauloCereda had breakfast elsewhere but there is always lunch

@DavidCarlisle oh no

@DavidCarlisle where? oh the raw encoding

Early morning insomnia.

ooh coffeee

ooh extra e

OH NO

@PauloCereda An outstandingly pointless cartoon.

@PauloCereda you can buy versions of that with the extra e removed, it's called coffee

@FaheemMitha feel totally free to ignore it.

@DavidCarlisle ha :)

@FaheemMitha also, so is your comment. :)

@PauloCereda Ooh.

@FaheemMitha ooh :)

@PauloCereda This bloke does not look very enthusiastic about the advertisement

@samcarter_preparing_for_xmas The funny thing is that Tom Nook always looks that way, even in-game!

@PauloCereda The shirt would suite @yo' :)

@samcarter_preparing_for_xmas ooh definitely!

@samcarter_preparing_for_xmas ^^ the game also features his nephews, Timmy and Tommy.

@PauloCereda :) these look much happier

@samcarter_preparing_for_xmas in my TUG 2020 video, you can see them in the TV behind me. :)

@samcarter_preparing_for_xmas I don't wear green (much) :-)

@yo' arg :) I guess we could select a different colour model to shift this to red

> Japan: Animal Crossing: New Horizons first game to top 6 million sales at retail since New Super Mario Bros

@samcarter_preparing_for_xmas ^^

@PauloCereda 5.99 million due to your TUG'20 talk :)

@samcarter_preparing_for_xmas LOL

@samcarter_preparing_for_xmas Isn't that how many users Overleaf claims to have? Too close to be a coincidence ;-)

@PhelypeOleinik :D Maybe there is a discount if one takes both the game and an Overleaf subscription?

@JosephWright In connection with tex.stackexchange.com/q/572592/4427, wouldn't it be nice if the S column type could also execute a command such as \fpeval{round(#1*100,1)} to every entry in the column? Say with an option command=\makepercent where the command is a one-argument macro?

@JosephWright This would have the limitation that the entry has to be in a format suitable for xfp, so with a decimal point and not a comma, but perhaps the number is normalized… You know better.

@egreg That's in the works for v3, it's just a question of me finalising it: Frank's been after me

@JosephWright ooh the Hound of the Garamonds Helveticas Futuras Baskervilles

@JosephWright after you too ;-)

@UlrikeFischer ooh

@JosephWright Good to know!

@egreg It's really a question of finding the time to finish the corner cases

@UlrikeFischer and me

@DavidCarlisle longtable?

@UlrikeFischer so many, I forget:-)

@DavidCarlisle missing the time to make good todo lists too?

@UlrikeFischer ooh it's in the todo

@samcarter_preparing_for_xmas ooh das Vampir

@PauloCereda :D

I hope you weren't startled :D

@samcarter_preparing_for_xmas I was!

@DavidCarlisle independent.co.uk/arts-entertainment/art/news/…

@PauloCereda obviously not art, no pizza base

To whom it may concern: github.com/latex3/latex2e/issues/441

@DavidCarlisle oh well spotted

@HenriMenke uh-oh

@HenriMenke Frank will have seen it I'm sure

@JosephWright There is still time to roll back the hook management. TeX Live 2020 has not been frozen yet. Since more and more issues keep popping up it might be the best to go back to the drawing board.

@JosephWright I vote for unicode-math not delaying everything.

@HenriMenke 'More and more issues'?

@HenriMenke very few issues have come up other than waiting for package updates the need for which have been known since the first drafts earlier in the year.

@JosephWright I've had to fix several issues in PGF and I don't think anyone has any idea how many of those “templates” hook into different parts of LaTeX in unexpected ways and believe my when I say that people are going to be pissed when awesome-cv et al. stop working.

@HenriMenke for any class in texlive we have a pretty good idea, most private classes are derived from one of those so the number of classes doing really unexpected things should be manageable.

@HenriMenke Templates are very tricky: since many of them rely on internals all over the place (packages as well as the kernel), avoiding breaking all of them is more-or-less impossible for any significant change (one reason I'm still stuck with an unfinished siunitx v3 is worries of this sort)

@HenriMenke that is known. See e.g. github.com/plk/biblatex/issues/1066 for a similar discussion. If you want to ensure an order use labels and the hook rules.

@HenriMenke The pgf stuff was I think all to do with shipout, which as @DavidCarlisle says was essentially 'predictable' in the sense it's very hard to change without matching changes in other places

@UlrikeFischer This is not a solution.

@HenriMenke a solution not requiring document changes is for unicode-math to arrange that it inserts its code before any top level preamble inserts, which should be possible.

@DavidCarlisle It's not specific to unicode-math. Any package that uses \AtBeginDocument will trigger this bug.

Quick (I hope) question. It is generally the case that a package that provides a key=value interface for commands can also do it for other packages?

@HenriMenke it is not a bug. It is by design. If you want a specific order you have to setup a rule. Or you will have to use another hook, e.g. \AddToHook{begindocument/end}{\renewcommand\vec{\mathbfit}}.

@HenriMenke It isn't a bug, it's a known feature.

@LaTeXereXeTaL ?

@UlrikeFischer Don't be ridiculous. This document is absolutely trivial and it worked in 2019 but doesn't in 2020. This is clearly a regression.

@HenriMenke yes there is a change. But if I may say we broke more documents and more seriously. All document using catoptions (through for example xwatermark) don't work at all after the changes. I just told today someone that their class no longer work as I disabled kvoptions-patch (and they took the message with much more calm than you). And your document is not trivial as it has a hidden top-level \AtBeginDocument in the file of the external library which you could avoid by adding a label.

@DavidCarlisle I'll try again. Is is true that a package providing a key=value interface for ordinary commands also provides a key=value interface for packages? In other words, can one package do both? Hope that was clearer.

I think the answer is yes but I'm trying to be certain.

@LaTeXereXeTaL do you mean as package options? then no not in general, eg keyval provides a keyval parser but doesn't touch the option system (kvoptions package attempts to patch the package option system to use teh keyval parser)

@DavidCarlisle Yes, as package options.

Isn't kvoptions having some problems nowadays?

@LaTeXereXeTaL any package that claims to add a key value option system to \usepackage has problems. (I may try again to fix that before texlive 2021 but it is tricky see discussion above, any change breaks something, and every latex document uses \documentclass so there is a real chance of breaking everything

@DavidCarlisle Given that, what's your recommendation for package options?

I had some small demos working with kvoptions before I became aware of the recent problems.

@LaTeXereXeTaL with latex as it is (as it has always been) then \usepackage{hyperref}\hypersetup{a=b,c=5,1=2} is much more robust than \usepackage[a=b,c=5,1=2]{hyperref} here hyperref supports the second form as a shorthand for teh first but it only works in some cases.

@DavidCarlisle I like that interface too!

@LaTeXereXeTaL kvoptions is essentially the code from hyperref made more general so it applies to other packages.

@LaTeXereXeTaL which the hypersetup interface or the usepackage option interface?

@DavidCarlisle the hypersetup interface

@DavidCarlisle I would just like to use the same setup for commands and packages if that's possible.

@LaTeXereXeTaL package options shouldn't be a problem if you don't try complicated stuff with spaces, see biblatex or hyperref. That all works. But normally I would avoid it if possible, as David said, using a setup command is a better interface.

@UlrikeFischer I can't foresee using anything but spaceless (i.e. one word) keys so I don't think spaces will be problematic.

@LaTeXereXeTaL that is much easier to do and much safer, so not offering the usepackage version is a realistic possibility. all keyval parsers take great care to step around , and = to split thing sup without destroying the values being passed, but the argument to \documentclass and \usepackage is put through \edef\zzz{\zap@space...} as teh first thing which typically destroys the values. We just were not expecting keyval there.

@DavidCarlisle @UlrikeFischer I think that answers my question. Thank you both!

@DavidCarlisle @UlrikeFischer Sorry to be a pain! LOL

@LaTeXereXeTaL it is not just the keys. If you have in keyval a setting such as font=\bfseries then the internal key associated with font will get \bfseries if however you put that in the optional argument of \usepackage it will blow up with horrible internal errors due to premature expansion before teh keyval processing even starts.

@DavidCarlisle That's good to know.

I blame @UlrikeFischer for adding \zap@space here.

I plan on doing lots of experimenting tonight.

@DavidCarlisle you shouldn't have written keyval, then nobody would complain about the \zap@space.

@UlrikeFischer I shouldn't have told Sebastian how to patch the package option handler to give it a kv interface...

This post is particularly interesting to me. tex.stackexchange.com/q/525261/218142

@PauloCereda small change in the image: can you make the word "Change" bigger?

or would it break something :-D

@LaTeXereXeTaL yes, I should get a grip and implement the colors.

@PauloCereda you got a macro for this layout?

@StefanKottwitz good idea! Will reproduce it in TeX. :)

@LaTeXereXeTaL best response is obviously this: tex.stackexchange.com/questions/525261/…

@DavidCarlisle Oh that literally made me laugh!