I found that quote markdown sometimes interferes MathJax as follows.
This is an example that shows how quote markdown can interfere MathJax. $ E
> =mc^2$.
You need to view what I typed as a quote in the edit mode.
@MakeMeSmarterEveryDay it doesn't let me use the edit to see what you used. having to double protect characters that are special to markdown and tex is quite common though, if that is the issue?
@MakeMeSmarterEveryDay I'd say you were optimistic hoping that would work, you can't expect to interleave markdown within the fragment that's going to be passed to mathjax (although it works sometimes)
@MakeMeSmarterEveryDay > This is an example that shows how quote markdown can interfere MathJax. $ E =mc^2$.
@MakeMeSmarterEveryDay what's the advantage in splitting the source over two lines. But I have no information about how stackexchange filters the markdown, they may have some other quoting mechanism in place.
@MakeMeSmarterEveryDay yes perhaps. We don't have mathjax running here, so I haven't much experience with mathjax in stackexchange markdown. So someone may answer but my guess is that it gets classed as a feature
@UlrikeFischer they have added weird offsets for the usual paragraph case so the caption box ends up with kerns of -1in at start and end of the text (followed by parfillskip at the end) so to make the oneline hbox version for small captions it does \unhbox\@ne\unskip\unskip\unpenalty\unkern to try to clean up but in 2021 that final \unkern just removes the new 0pt guard kern leaving a negative inch kern which pushes centering off by half an inch.
@UlrikeFischer but I'm not totally sure where the kern0pt comes from
@DavidCarlisle -- I infer from your last comment on your answer that this is your "ams ping"? When you confirm, I'll send it on. (Have you tested to see what another unkern will result in with TeX Live 2019? I think that's the current production version. If that doesn't work, ...)
@barbarabeeton no I'll send a message to the support address if need be but actually I think we are going to change latex instead (running some tests now) the unkern will be Ok in practice in older releases. It's probably actually more of an issue for @yo' and @PauloCereda as not sure can get a fix out before texlive 2021 ships on overleaf
@Skillmon you just beat me to that lua keys thing but I was held up as I couldn't get it to parse my test option strings
@DavidCarlisle oh, that's understandable. I guess the syntax of what is considered one key and what isn't is a bit strange... :)
@DavidCarlisle I guess the problem is that luakeys tries to guess the data types of the options, and when it hits 1 2 3 4 it treats the 1 as a number, and the rest as separate keys. But I have no idea why it does treat 2, 3, and 4 as separate entries with a space as separator. To get the output you'd expect from a LaTeX point of view you'd have to use quotes: \RequirePackage[aaa,bbb="1 2 3 4",ccc={1 2 3 4},zz=red]{MWE}.
@DavidCarlisle I think so. With this new method you get the worst of two worlds, really. If you want to use the raw options you still have to protect things against expansion, but this way you might not get what you want inside the raw options. But I'm not sure what would've been the best way to not use the \edef. If you get rid of that you definitely break backwards compatibility with old documents.
@Skillmon I was wondering if it would be possible to delay the edef to \processOptions which would probably be ok for standard processing but who knows what patched extensions do
@DavidCarlisle That would certainly work for packages/classes using LaTeX's builtin options, but things like pgfopts or l3keys2e, etc., would perhaps break if they assume sanitized input. This could (in theory) break usages in which a user used a space inside a key name where they shouldn't have been one from the package point of view.
@DavidCarlisle with the new macro patching code in the kernel one could delay the \edef step until the various process-option macros of the packages (hello firstaid...). But that would require quite a few packages being covered by firstaid (not expkv-opt though :)
@Skillmon yes but by the time you've got that far tempting to only pass on the raw list and update the packages to apply zap@space themselves if they need that, but at the time that seemed too big a change
@Skillmon I want to do the edef on the old list in a try-catch .....
already \protected things don't expand, this would do same for \protect and mean most reasonable things would be error free (if teh raw list is being used it doesn't really matter what this results in so long as it doesn't error, but mostly it results in something sensible, I think)
@UlrikeFischer I left a comment, the class is really magical, it's a latex2.09 style with a minimal mostly broken update for 2e so it's never worked but apparently the university has been using for letters all that time
@DavidCarlisle I guess I wrote 2 letters in 12 years, on paper
@DavidCarlisle it's hard if you don't find your LaTeX letter template from 8 years ago that should be on some of the backup thingies
@DavidCarlisle Apart from letters, I had to do a CV in Word because my old .tex one is on an encrypted NAS that I used on my now broken PC, both in the basement ...
@PauloCereda we hard hard weeks with despair and all kind of tries (several roentgen, computer tomography, therapies) at the end it was a salvation for our friend
@PauloCereda but thanks for asking, it's good to remember a friend
sorry to mention in chat, it's life, and you asked, and it's ok to talk when it doesn't hurt anymore, because, it's life.
@PauloCereda my 3 kids are fine, they just complain that dad is too often in the basement
where can we have datacenters, marijuana farms, and meth labs, except in the basement
attic? everybody sees it when the first snow comes and is everywhere except on your roof
now I expect visitors :-D
@PauloCereda I'll send you some domain cc auth codes before they enter