@UlrikeFischer What is better than the duck of the day? The marmot of the month! ;-)

Hmm Amazon emails me to suggest some bank holiday reading and suggests I buy a copy of the LaTeX Companion....

@JosephWright @UlrikeFischer seen texlive mail?

@DavidCarlisle Just have: I will take a look

@JosephWright it looks like the start of page \special is causing \clearpage not to clear the mvl and loop but not fully traced

@DavidCarlisle OK, I have some idea where that might come from ...

@DavidCarlisle Frank will be pleased.;-)

@DavidCarlisle oops. Does it happen too if one loads only expl3 instead of mhchem?

@UlrikeFischer yes

@DavidCarlisle Bisecting now

@UlrikeFischer @JosephWright don't need geometry either:

\documentclass[landscape]{article}
\usepackage{expl3}
%\usepackage{geometry}
%\geometry{landscape}
\begin{document}

\tracingmacros1
\begin{center}
  \rule{1.1\linewidth}{1.1\linewidth}
\end{center}
\end{document}

@DavidCarlisle Right you are

@DavidCarlisle Problem is the set up for breaking links:

\documentclass[landscape]{article}
\usepackage{hyperref}
\usepackage{hypdvips}
\begin{document}
\begin{center}
  \rule{1.1\linewidth}{1.1\linewidth}
\end{center}
\end{document}

@JosephWright so you stole bad code:-)

@DavidCarlisle Working on it

@JosephWright thanks, we had an issue previously (right at the start of 2e I think) were we re-organised things and ended up with a start of page whatsit that had this effect.

@DavidCarlisle Problem is in hypdvips:

\global\let\pp@backup@@makecol\@makecol
\gdef\@makecol{%
  \setbox\@cclv\vbox{\unvbox\@cclv\pp@SDictPS{pp.eop}}%
  \pp@backup@@makecol%
  }

yep i saw the Ps Dict special come up in the trace.... but is it clear when it is safer to add it?

@DavidCarlisle You/Frank are the 2e experts ...

@JosephWright are/were/something do you want me to trace or are you in the middle of it?

@DavidCarlisle You know more 2e than me, like I said. It comes down to having to make sure we add the tracking code at the start/end of columns, so hooking into \@startcolumn and \@makecol at present.

@DavidCarlisle Actually, it's the \@startcolumn change that is problematic

@JosephWright hmm so it gets put before tryf(loat)column so you don't really know how many pages are going to be output at that point, i guess i should look at hpdvips.dtx not the .sty...

@DavidCarlisle There is no DTX

@JosephWright so I see, no documented source at all :(

@DavidCarlisle ooh

@DavidCarlisle I think we just need to move the special so we add it to the top of the output for each column: I guess there's a place ...

@JosephWright the \setbox\@outputbox in \@vtryfc and w.. and x.. ?

@JosephWright hm. Do we need more shipout hooks in the kernel? ;-)

@UlrikeFischer from my point of view: Yes. The more hooks the better :)

@Skillmon All the better to shoot your own foot with … but yes, I want hooks too.

@Skillmon be careful what you wish for

Firefox is breaking extensions :)

They are working on a patch.

@DavidCarlisle ooh ein piraten

THE PIRATE HAS AN ARARA IN HIS SHOULDER

@PauloCereda yep, mine is unusable right now :(

@PauloCereda "Oh, ein Pirat"

@Skillmon A ticket is opened in Mozilla's bugtracker, I believe a patch is on the way.

@Skillmon ooh ein Pirat

@PauloCereda Germans set a comma after exclamations, so it has to be "ooh, ein Pirat".

@PauloCereda yes, I saw that, too.

@Skillmon ooh cool!

@PauloCereda perhaps exclamations is the wrong term here, though. @AlanMunn what are those called in English grammar?

@Skillmon the "ooh thingies" :)

@Skillmon interjection?

@HaraldHanche-Olsen thank you very much, that should be the correct one :)

@HaraldHanche-Olsen from that link you posted: "Gesundheit! Are you starting to get a cold?" Do you think the American English speakers know what they are saying there?

@Skillmon I have no idea; but it is explained in their dictionaries, it seems. (Though not in the 1913 edition of Webster's, which I happen to have as an app on my phone.)

@Skillmon Interesting trivia: In the same situation, Norwegians say prosit. Which is, of course, used quite differently in Germany. (Do you think Germans know the meaning?)

@HaraldHanche-Olsen I do and some others most likely, but not everybody. Does every Norwegian know what he is saying?

@DavidCarlisle Any ideas on that \special? I feel we need a fix quite quickly, so am leaning toward just stripping the whole breakable link stuff out at present

@JosephWright I tried a couple of things but I was having same thought, that perhaps need to back it out and try again later.

@DavidCarlisle OK

@JosephWright is it a problem of latex itself? I mean, if you would change/add the code to latex.ltx would it still break?

@JosephWright there's a danger of adding it somewhere else so that example works but some combination if float column trials trials drops it again so need to think of some good tests and that's hard to do while under time pressure of a public bug report

@UlrikeFischer Like I've said on the issue page, there must be somewhere one can know that a column has some content. But I think it's likely one for Frank.

@UlrikeFischer if you added it there (in startcolumn) yes

@DavidCarlisle Yes, that's true

@JosephWright i'm trying to understand the specials again, doesn't the start page one need to be paired with this eop one? \pp@SDictPS{pp.eop}

@DavidCarlisle Yes, I think so (to be honest, I've not got all of the detail with that 'tracing' code, but you do need it)

@DavidCarlisle Thinking they could both go in \@makecol?

@JosephWright well yes adding them both at same time, and only if box255 isn't empty, would seem safer but...

@DavidCarlisle Seems to work

You have all enabled 2FA on your GitHub, right? zdnet.com/article/…

@JosephWright but that's the more obvious design so i was trying to find the documentation of why hyperdvi split them up

@HenriMenke no...

@HenriMenke Yes

@HenriMenke I'm so worried ;)

@DavidCarlisle My tests from earlier still work: probably worth a shot

@JosephWright The guy could maintain the source... :)

@DavidCarlisle, @UlrikeFischer Could you test the commit? If it looks OK I'll release: perhaps there is more to do, but at least this fixes the current case and works with the other tests

@DavidCarlisle I wonder if he just saw the name and thought it was a good place to aim at

@JosephWright I can test against the example, but I'm not sure what to test else. Why does it break only with landscape?

@JosephWright ^^ :)

@JosephWright with a bit of luck someone will steal tabu-fixed

@DavidCarlisle ^^

@UlrikeFischer Because of the way the content runs on, not because of landscape per se

@JosephWright ah, because the rule gives an overfull page.

@UlrikeFischer Yes

@Skillmon Most likely not. It's from Latin, meaning something like may it be beneficial, but you knew that of course. ;-) I expect most people have learned and forgotten the meaning.

@JosephWright had you contact with the author of hypdvips? He should correct his code too.

Perhaps stupid Q, but is it not possible to read and write to an ext file at the same time, right? You have to \closeout.. before you \openin.. ?

@vlg I think if you try you will be relying on whatever your filesystem locking does but this works for me (tl2019/cygwin)

\documentclass{article}

\begin{document}

\immediate\openin6=\jobname.foo
\openout6=\jobname.foo

\immediate\write6{hello}

\read6 to \tmp

\typeout{\tmp}

\immediate\write6{hello2}

\read6 to \tmp

\typeout{\tmp}


\end{document}

produces

hello
hello
hello2
\par

@UlrikeFischer True

@DavidCarlisle Fails here: I read from the terminal

@JosephWright how does one load in expl3 a driver like the one for dvipdfmx (instead of dvips?)

@UlrikeFischer \usepackage[driver=dvipdfmx]{expl3}

@UlrikeFischer Usually I test with XeTeX/(u)pTeX for dvipdfmx: it's only dvisvgm that we can't guess at all

@JosephWright I was surprised it worked as it did here actually.

@JosephWright can one suppress the loading of the drivers and do it later? (I'm thinking about interfaces, not tests)

@UlrikeFischer cp l3dvipdfmx.def l3dvips.def :-)

@UlrikeFischer No, not at present

So is everyone having fun with Firefox today? reddit.com/r/firefox/comments/bkhtv8/…

@AlanMunn Yes :(

@AlanMunn I run a more or less stock nightly so extensions not a problem

@JosephWright I think we will need something (even without the question if we want expl3 in the kernel). It looks odd to have to setup something so document specific in the options of expl3 (I sound like someone else ;-)), and it is difficult to write interfaces for it (with e.g. xparse).

@UlrikeFischer I guess we've not had a lot of users wanting the driver code and wanting a 'non-standard' set up

@JosephWright no, but if we want packages like pdfbase/ocgx2 to use our stuff, there must be an interface to select other dvi-drivers.

@JosephWright @UlrikeFischer would it work to delay driver loading to begin document and have some api that lets package code (hyperref, graphic-something, whatever..) or a choose select the driver

@DavidCarlisle Yes of course

@DavidCarlisle I'm still preferring the plan to set the driver and other core stuff before or at \documentclass: things like pdfversion should be set rather early and packages can do curious stuff in the preamble.

@UlrikeFischer well i was wondering if end of preamble was too late but doing it before documentclass means documenting a more or less non standard syntax which would be nice to avoid.

@DavidCarlisle a keyval-syntax in the \documentclass options could work too, but after \documentclass is imho to late if you consider large classes which loads lots of packages like pdfx or hyperref.

@JosephWright Is it by choice that the stars in interface3.pdf are not links any longer?

@egreg Er no

@egreg Did they used to be?

@JosephWright Yes. Try /usr/local/texlive/2018/texmf-dist/doc/latex/l3kernel/interface3.pdf and look for \cs_to_str:N on page 4

@egreg Working on it

@JosephWright Not a big deal: all links point to the same two places.

@egreg But something must have changed

using 6 or \newwrite shouldn't make any difference but the actual behaviour is likely to be very system dependent

achso

@DavidCarlisle, @UlrikeFischer Thoughts on github.com/latex3/latex3/issues/570 welcome

@egreg Bisecting to find the issue

@JosephWright I was thinking...

@JosephWright I suppose main options are: take it out, leave it in, or perhaps just have expl3 (in lieu of the format) add empty hooks at start and end of \@makecol and then have a package level interface inserting the specials via those hooks....

oh you just said something similar in the issue

@DavidCarlisle Yes. What I could do is make the standard PostScript into no-ops and take out the hooks. Then we just need to patch one macro (or have a kernel change that enables the hooks), plus a load of PostScript header code

@DavidCarlisle I'm working out why @egreg has spotted an issue, so bisecting, then I might make a branch which has a 'l3driver-extras' module that we can use to collect up things that need non-existent kernel hooks

@UlrikeFischer On the driver business, is it so bad to say 'if you want a driver that can't be auto-detected, you need to \RequirePackage[<driver>]{expl3}before \documentclass (@DavidCarlisle)

@HaraldHanche-Olsen yes, I had 6 years of Latin in school (and have not forgotten everything up to now :)

@JosephWright it's workable:-)

@DavidCarlisle if expl3 adds hooks to \@makecol then we have the same situation: expl3 is interfacing with the latex format. Then it could add the postscript code too. I'm not sure what the correct solution is to this, l3driver.sty?

@UlrikeFischer yes but that isn't necessarily a problem (adding the postscript is more risky, as if anything else changes any definitions in that area things go wrong, but just adding a couple of empty hooks is safer)

@UlrikeFischer That's what Chris is saying: no (more) fiddling with LaTeX2e in l3kernel, do everything in packages

@JosephWright yes I got this. And I see the point, patching latex2e commands could be problematic. But I don't see why it is better if it is done in expl3-packages ;-). Perhaps one should start to add hooks for expl3 in latex.ltx.

@UlrikeFischer one possible argument for splitting things that way is that it makes it easier/safer to preload expl3 into the format as that is then just a load optimisation without affecting running code if its not used, but it still leaves an explicit package in the document required to tie things up the the output routine...

@UlrikeFischer That's also what Chris would like to see

@DavidCarlisle Or we only offer that functionality if the format is sufficiently new that the hook is there

@JosephWright don't like that idea, just leads to years of "why doesn't abc work on overleaf" questions.

@DavidCarlisle yes, for my driver-code question it would be good. ;-)

@DavidCarlisle So a package then? Plus hopefully a hook built in that we can later move to (or perhaps move to immediately with the package 'auto-loaded' in that case)

@JosephWright how difficult is it to move the driver code out? I saw that e.g. some box-commands uses \driver-commands, are there other dependencies?

@JosephWright yes as I say I think adding the hooks in expl3, and using them in a package leads to a future where the hooks and expl3 are in the format and the package interface stays the same, perhaps...

@DavidCarlisle won't this happen anyway if they don't have the newest expl3?

@UlrikeFischer Depends where you look. I really was keen to avoid having to load lots of separate driver files

@UlrikeFischer yes but updating a package is easier even on overleaf, they tell you how to make the files part of your project, but updating the format is harder

@UlrikeFischer For example, boxes, color, drawing all need drivers

@JosephWright but presumably the drivers can be stubs until you actually need to set a box or draw something, which shouldn't be before \documentclass?

@DavidCarlisle Sure: the drivers don't have to be loaded at all until some higher-level thing needs them. @UlrikeFischer's case of PDF level or whatnot is the harder one

@JosephWright and how hard would it be to setup things so that you can reload/overload a driver file?

@UlrikeFischer That's a question of \cs_new_protected:Npn versus \cs_set_protected:Npn

@UlrikeFischer Mainly, at least: we'd have to watch register allocation too

@JosephWright if I do some setup at/before documentclass I can also load a driver, and if expl3 is loaded after \documentclass it could load it as usual.

@UlrikeFischer Well yes I guess: one for the team list perhaps?

@UlrikeFischer the main problem with reloading drivers is not so much loading the definitions twice, but if the code has already added (say) dvips specials setting page size or loading some postscript header then you switch and specify dvipdfmx, you can't undo those specials very easily (well unless you force a call to the output routine and discard the entire mvl constructed so far...

@DavidCarlisle yes, doesn't sound like a sensible idea. it is better to have an option to delay the loading, at least if dvi mode is detected.

@UlrikeFischer What drivers other than pdfmode, dvipdfmx, dvips, dvisvgm are you thinking we need to worry about?

@UlrikeFischer We could go with delaying until \AtBeginDocument unless some special marker macro is used ...

Eating @marmot can be dangerous. themoscowtimes.com/2019/05/03/…

@JosephWright imho that is the lot, and only the three dvi-mode (with latex) really needs delaying (or do we need to consider the case, that people change the output mode?).

@UlrikeFischer I've never considered supporting changing output mode: it's really a developer thing (and even we don't bother other than for low-level testing)

@JosephWright I don't think that delaying so much is really needed. If expl3 is loader after \documentclass it could simply do what it ever did.

@UlrikeFischer OK, so check for \documentclass?

@JosephWright I would say so, it is a clear "border", and we won't run into troubles if packages try to use driver dependant commands in the preamble ...

@AlanMunn yes

@AlanMunn but I am faithful FF user. :)

@AlanMunn Well, "Despite the fact that eating marmots is banned, Citizen T hunted marmot.". There is not much one can add.

@PauloCereda FF is slow, as compared to Chromium and descendants.

@AlexG I have better user experience with FF than with Chrom*.

Besides, I do not trust Google at all. :)

@PauloCereda : Try this: agrahn.gitlab.io

@AlexG slowmo. :)

@PauloCereda And now with Chrome. You'll be stunned!

@PauloCereda Me too. And I also don't want more Google in my life.

@PauloCereda sarahkabine.files.wordpress.com/2010/05/13-vampirente.jpg

@AlexG I did, quite impressive indeed. :)

@UlrikeFischer ooh vampires and mirrors

@AlanMunn , @PauloCereda Btw: Chromium is Open-Source.

@marmot Yes, although it's part of the cultural tradition, so I think the ban is not for the marmots' sake. :)

@AlanMunn I had two add-ons, so I can wait. :)

@PauloCereda I followed the instructions in the reddit post I linked to and all is fine for the moment.

@PauloCereda Working again for me

@AlanMunn The thread is gold: I am pissed! You have any idea how many singles in my area are now trying to contact me?? I never wanted the fame!

@JosephWright WAIT A MINUTE IT'S WORKING FOR ME NOW

@PauloCereda lol

@AlanMunn It is really cruel to eat marmots!

@marmot definitely, stick to eating duck

@DavidCarlisle oh no

FF is back! Yay!

@JosephWright what happens if both expl3 and hypdvips is loaded? Isn't then the postscript code doubled and is this a problem?

This is probably a dumb question, but can TeX and friends actually downsample images? I.e. if one includes a JPG which is (say) 2550x4204 pixels in a LaTeX document. This is like 300 ppi at US Legal size (8.5 in x 14 in). Can one tell TeX to effectively include a JPG with less pixels? I don't think it is possible, but I thought I would check.

I know \includegraphics can do scaling for example, but I'm unclear what that means, exactly.

@UlrikeFischer The code is doubled but I've tried to avoid the same names, etc.: in testing, everything worked OK

@JosephWright you are thinking about everything ;-)

@Diaa Huh? What are you trying to achieve?

@JosephWright get a rounded angle to 2 decimal places (i.e. the same output of \ang{2.33})

@Diaa \ang[round-precision=2,round-mode=places]{2.3307}

@FaheemMitha No, usually TeX just includes the image and tells the viewer in which dimensions it should be displayed.

@TeXnician That's what I thought, but thank you for the confirmation.

@JosephWright This is what exactly happens when trying to learn by checking Q&As without reading the documentation. It is even stated in the "siunitx for the impatient" section in the manual XD Thanks for the consideration

:50165572 hahahaha It is stated in a section entitled "siunitx for the impatient", which makes the irony intense.

@Diaa Indeed. :)

@AlanMunn Amen XD XD

@Diaa Another user on the site once described himself as "turning documentation into reputation points".

@AlanMunn Impressive recycling of online PDFs XD

@egreg Found it: github.com/latex3/latex3/commit/…

@Diaa This must be an oxymoron since the impatient will never read this because they do not read documentations. ;-)

Unsure if this warrants an official question, but why does load order matter so much with tikz things? Yesterday I was hunting what exactly I'd changed for it to not be able to compile at all- Rather than loading pgf,pgfplots,pgfplotsable and a usepgflibrary{dateplot} using \usepackage inside a \AtEndPreamble{.. I'd moved them to somewhere else, regardless, just putting them back made it at least work.

I have no idea what it's doing, and the errors it spits out aren't helpful and I dunno what exactly to do other than move something, build, and repeat until it works.

e.g., currently loading them at lastly in the .sty with RequirePackage makes it so graph externalization yields empty output, i.e., nothing; whereas loading them using usepackage within a AtEndPreamble yields graphics (where the \pgfplosset{..} also is to be found- but regardless, if they're loaded beforehand or JUST beforehand I don't understand where the difference is)

@ me, if you know sth, or think it's a reasonable questions for the main site

@vlg As with your other question, it looks like a reasonable one for the site if you can construct a MWE.

@JosephWright imho the plan sounds good, but I can't test until tomorrow afternoon. Btw: I found a problem with the AR-fix/ocgx2: gitlab.com/agrahn/ocgx2/issues/10

@UlrikeFischer Oh great. I want to adjust how we do that anyway, but it needs to be done in line with the shipout hook. Do you think we should drop the fix?

@UlrikeFischer I've checked in an idea: want @DavidCarlisle and Chris and Frank to look, really

@JosephWright let's wait what Alex says. Perhaps it means that every annotation should use the same flags, and then we need something to sync this anyway. Btw: hyperref uses \pdfstartlink to create the text fields, so it looks as if I need a \driver_pdf_link_begin_user:nnw version which doesn't add /Subtype/Link, unless I want to redo everything to use pdfannot.

@UlrikeFischer Right .. hmmm

@UlrikeFischer Well isn't that conceptually \driver_pdf_field_begin:nnw or something? It might use the same structure as a link, but it's not actually a link

@UlrikeFischer I'll take a look at hyperref tomorrow

@JosephWright yes saw it, that was what I meant with the plan. But I can't look now, I have a chess match tomorrow and should go to bed ;-).

@JosephWright yes something like this would be ok. Perhaps one should also have a field_end - even if it does the same as link_end.

@UlrikeFischer Taking a quick look, these are all field things, yes?

@UlrikeFischer Sure, conceptually they should have the right names

@JosephWright yes, \@Textfield, \@Submit etc., they are all /Subtype/Widget imho (but I don't know how easy it would be to prevent hyperref from adding the /Subtype entry itself).

@UlrikeFischer We could just add nothing at all at the driver level, and expect things 'higher up' to add /Subtype /Link. I think the only wrinkle is that goto links need to know whether or not to include that as some drivers do it automatically. But for link_begin_user we could just drop it

@JosephWright would be ok with me too, and it would be probably be better - less driver commands.

@UlrikeFischer Right, OK: I'll adjust

@DavidCarlisle Is \@reargdef used somewhere? In latex.ltx it is defined but not used...

@PhelypeOleinik vvvv

%  \changes{LaTeX2e}{1993/12/20}
%                {Kept old version of \cs{@reargdef}, for array.sty}
% \changes{v1.0b}{1994/03/12}{New defn, in terms of \cs{@yargdef}}

@DavidCarlisle Ooh, sources. Thanks :)

@PhelypeOleinik I stole bits of newcommand for \newcolumntype but then we decided to do 2e and I recoded \newcommand and we never had time to bring them back in sync:-)

@DavidCarlisle Interesting :)

@DavidCarlisle Where can I find LaTeX 2.09? Just out of curiosity, to see the differences :-)

@PhelypeOleinik there is the historic archive with every release but a first place would be ctan.org/pkg/latex209?lang=en

@PhelypeOleinik or pre-history latex 0.9 ftp.tug.org/historic/macros/latex-saildart/latex-0.90/latex.tex

@DavidCarlisle Looks good. Thanks :)

@DavidCarlisle for use with TeX version 0.95 -- Wow, I didn't realise LaTeX was that old... I thought it was created much later.