@UlrikeFischer I have a class which I use to create course materials. I externalise TikZ images, but sometimes, externalisation fails for a picture in some weird way which interacts with Biblatex. If I delete enough refsections, it compiles fine. I can get no useful information about the cause of the failure. For some courses, I have no problem. For others, I do. The difference seems to be the number of refsections. But all attempts to minimise the problem fail. David says it is too much ...

... TikZ, which is fair enough. But I have no idea how to investigate what's wrong. List & make fails the same way. Running the relevant externalised compilation manually also fails the same way. The number of refsections in the externalisation log is much higher than in the document log, but I don't see why this should matter or why I don't get a meaningful error. It doesn't help that this typically only errors after 200+ pages, though my 'minimal' case got that under 100.

@DavidCarlisle I know. It is a bug in hobby that I'd forgotten about. I was excited without cause. Once I add @Jos

... @JosephWright's workaround, I'm back to the same non-error with list and make as I was with the default.

I rewrote a bunch of stuff to reduce the number of refsections and eliminate refsegments, but I still can't generate the last part of the course packet without disabling the bibliography for that part, which I need. Even if I get the externalised image while the bibliography is disabled, when I reenable the bibliography, I'm back to an externalisation error. Which makes no sense. I'm wondering if I can trick it somehow with list & make.

@UlrikeFischer I must be doing something fundamentally stupid, but I have no idea what.

@PauloCereda Cwac <3.

@egreg Yes, I knew that: I remember those days, I was there too! (;-) But today, I think that it’s better to favor consistency over saving memory.

BTW, I think that the best solution to the question I linked to is to define directly \styleForText with \DeclareRobustCommand: if I say \renewcommand{\styleForText}{\protect\footnotesize} I get \footnotesizeText in the written-to file.

@UlrikeFischer Could you, provided @cfr agrees, also send the code to me? Not that I think I have chances to solve the problem, but there is the saying "Also a blind hen may find a corn, sometimes." ;-)

@marmot What code?

Anyone do German-Welsh translation?

@cfr Sorry, I was under the impression you have some lengthy TeX document that causes headache. Sorry that I misinterpreted your chat message.

@marmot Oh, I do. Do you want it? Why are you asking @UlrikeFischer for it?

@marmot It is driving me nuts. Where can I find your email?

@marmot Thank you for editing. Can you get a file dialogue? I can't drag and drop them. I don't have a file browser. (I just use a terminal.)

@cfr If I click on the field where you are supposed to drag the graphics in, I get a file dialog.

@marmot Got it while you were replying ;). I thought I might be able to do it by uploading here first, got stuck again, clicked randomly and found it. They don't make it easy, do they? How are people supposed to know that?

@marmot Yes, but she doesn't have my code. @DavidCarlisle has it and I have it.

@cfr I always did it like this and found it by trial and error. I do not drag images either. Actually, I try to avoid using the mouse, but here it seems that one has to.

@marmot I use a mouse sometimes, but I don't like using it when I'm not in mouse-mode. (If I'm in Inkscape, fine. If I'm editiong .tex, rather not.) The web is annoyingly mixed. At times, I use a mouse and trackpad together, which can be good, too.

@cfr I just have 20 screens between I switch with the keyboard. This allows me to avoid having to use mouse to a large extent, but as you say for TeX.SE and the internet this does not always work.

@marmot If you want my TeX document, you can email me if I can't find your address somewhere. My address is in prooftrees documentation (texdoc prooftrees).

I am blown away!

@marmot Actual screens? I have 12 desktops .... I almost always switch with keyboard.

@cfr Virtual screens of course.

@marmot Well, you might be in air traffic control or something.

@cfr Sure. ;-) (you got mail)

@marmot You got very large mail!

@marmot ooh marmot mail

@barbarabeeton this is a mean hedgehog: vvv

[96/99, 13:32/13:39] auto-install: tikzlings (49113) [422k] ... done

@samcarter ^^^

Anybody who read the TeXbook knows to spell naïve, of course. And you may have come across coördinate. But the other day I came across reëvaluate in the New Yorker. That was new to me! Any other examples of this phenomenon? Maybe reënergise (or -ize), reënter (the atmosphere, like spacecraft do), reënable, coöpt, coöperative, …?

@HaraldHanche-Olsen It's mainly used in the US, other than things lik naïve: in the UK, we tend to favour a hyphen over a diaresis, so 'co-operative', Etc.

@CarLaTeX yeah!

Is there anybody around who could help reopening tex.stackexchange.com/questions/458802/… ? I'd like to post an answer

@samcarter 1

@DavidCarlisle Thanks! Answer posted :)

@JosephWright ^^^ Is there any chance you include @samcarter's tikzlings to have proper moles in siunitx? ;-)

@marmot hi mr. marmot!

@PauloCereda Hello Dr. Duck!

@marmot if it's not too trouble, I'd like to contact you. :)

@PauloCereda I'm here.

@marmot I mean by some sort of electronic communication. :)

@PauloCereda But this is electronic, isn't it? ;-) (What's your email?)

@marmot ooh marmots are smart

@marmot You can reach me at cereda dot paulo at gmail dot com

@marmot: you've been poked. :)

@PauloCereda better than "I got cooked", I guess. (Got your email but will finish another long email before I answer.)

@marmot ooh marmot news :)

Anyone remember a XeTeX package (or more like a TeX file) for doing Unicode math with Plain TeX?

@KhaledHosny there is this but I never tried it ...

@DavidCarlisle Looks like it, thanks.

Now I remember it used mapping files, not of much use for LuaTeX

@KhaledHosny there are some lua callbacks on the site that try to read mapping files, but it might be more natural to take the unicode-math-table.tex from Will's unicode-math package for latex and wrap it up in some basic plain definitions.

@DavidCarlisle would you expect \font\test=cmr10.0 at 10.0pt \test blub to work?

@UlrikeFischer no, not in general

@UlrikeFischer hmm I see, interesting:-)

@DavidCarlisle well it does with pdflatex but not in lualatex. Which means that the following works in pdflatex but not in lualatex:

\documentclass{book}
\begin{document}
\makeatletter
\def\f@size{10.0}
$x$

\end{document}

@UlrikeFischer this works in pdftex \font\test=cmr10.wibble at 10.0pt \test blub

@DavidCarlisle nice ;-)

@UlrikeFischer I wonder if that's a bug, font file names are in the system dependent part of tex, so the TeXBook is no help, but it seems pretty odd that the extension is simply discarded and not checked against the actual file name

@UlrikeFischer but tex and xetex also accept that.

@DavidCarlisle But is it a latex bug that in the range code it relies on this behaviour and doesn't change 10.0 to 10?

@UlrikeFischer oh in the nfss mappings to filenames, perhaps, actually I'm surprised it does that I thought it would have used the integers from the fd file in <5><6><7><8><9><10><12>gen*cmr% I'd need to trace....

@DavidCarlisle I see external@font ->cmr10.0 at10.0pt in the log.

@UlrikeFischer mail the list and complain to Frank:-)

@DavidCarlisle ok, will do ;-)

@DavidCarlisle ooh

@DavidCarlisle found the one I had in mind github.com/morbusg/unexplain, probably based on that file.

@DavidCarlisle If I want to comment on a conditional \if... doing so on the same line is bad, you told and showed me. Can't remember the examples tho, is the same true even if you input spaces or \relax or?

@vlg ?

Some time ago, you nudged me to not use % everywhere, explained how a, say, \ifnum conditional ended with a %, it wouldn't start reading the T expression as part of the condition to be checked, something like that

@vlg oh so nothing really to do with \if that is just a feature of literal integer parsing:

\count0=12%
3

sets count0 to 123 so if you intended it to be set to 12, don't put a % there.

I know it'd terminate if it encountered a command instead of an integer

@vlg only if the command is non-expandable

So, where best to put comments- above.?

that is, the row above, to guarantee less mistakes

@vlg anywhere that doesn't change the input. A newline is the same as a space so if you have 12<newline> then you can use 12 % any comment as both are 12<space> but if you do 12% any comment you have removed the space so the input is different

@DavidCarlisle thanks

@DavidCarlisle @marmot @UlrikeFischer I think the reason getting rid of the refsection works is just because the picture doesn't get externalised in that case, but I don't know why. I am guessing that the execution of the contents of the page hook (\AddPageHook) changes when the coffin is typeset. (Maybe just because it makes the content shorter.) But this doesn't really explain anything.

@DavidCarlisle Getting lots of experience of Git rebasing here: blog cleanup is interesting

@JosephWright changing the history to show you never posted a typo?

@DavidCarlisle Something like that :)

@DavidCarlisle Want it to be logical for conversion of format: after that ...

@JosephWright meanwhile, I don't know if you followed my chat above with @UlrikeFischer re updating tlg files. that was for oberdiek, but now, in base... If I have 116 failures in latex2e/base what's the official way of updating them (or is writing a bash loop to copy and rename the foo.engine.log files in build/test the right thing to do?)

@DavidCarlisle I run check, then take the list of diffs, edit out the extensions and rerun save ... you'll remember that Frank's not keen on save --all as it means you can just 'shoot from the hip'

@JosephWright yes but still... by the way if you try save without any arguments you get a low level lua error about accessing nil rather than a slap on the wrist saying to supply an argument:-)

@DavidCarlisle Hmm, better fix that

@DavidCarlisle @marmot @UlrikeFischer It's the coffins, isn't it? I need something like

        \enext_tikzsetnextfilename:V \l_cfr_coursepacket_rhannau_handout_enext_tl
        \hbox_gset:Nn \g_cfr_rhannau_handout_box {
          \def\thelocaltitle{
              \coffin_typeset:Nnnnn \l_cfr_taflen_teitl_coffin { l } { vc } { 0pt } { 0pt }
            }
          \cfr_coursepacket_cefndir:V \g_cfr_cefndirrhannauhandout_tl
          \tl_gclear:N \g_enext_figurename_tl
        }
        \AddThispageHook {
%           \hbox:n {
%             \enext_tikzsetnextfilename:V \l_cfr_coursepacket_rhannau_handout_enext_tl

instead of the commented out bit.

... don't I?

@cfr Did you get my mail? I could not unzip your archive, meaning tar -xzvf does not work. How does one unzip it?

@marmot Sorry, I've been photocopying. I just used

tar -xzf cfr-test.tar.gz

I sent @marmot the same archive I sent @DavidCarlisle. He must have unzipped it to know it contained too much TikZ. (He reproduced my error.)

@cfr sorry reproducing the error was about as far as I got:(

@cfr Sorry, my bad. I downloaded the attachment from my university mail server (where I forwarded the mail to), and this caused errors. When I download it directly from gmail, there is no problem. (It's quite strange, though.)

@marmot Then I've cluttered your email unnecessarily by just sending you a fresh copy, thinking it must have got corrupted the first time. Yes, it is odd that the mail server would do that. Not that odd, but somewhat odd. Ours deletes all .zip attachments (sometimes it tells you, sometimes not) and refuses anything exceeding 1.5M. And doesn't support IMAP or any other open standard. And randomly throws my mail away.

@cfr "And randomly throws my mail away." could be a good thing if it would throw away the emails sent by administration, but these are precisely the ones it lets through. :-(

@DavidCarlisle I was hoping you might say whether the above code looked obviously insane, but I understand if you don't have time. I'm not sure whether I should be doing all that inside the box, basically. I'm pretty sure I need a box.

@santimirandarp \centering \begin{tabular}\end{tabular}

@marmot Including those saying you need to clear your office and they'll dump all your possessions if you don't? Not nice, I know, but I would rather have known.

@cfr sorry I don't know the code well enough to say anything at all just looking at it by eye

@DavidCarlisle Fair enough.

Probably the clear should be outside or the set inside.

@santimirandarp which means, for example that if your centred tabular is in a table float then you almost always want \centering rather than center as the float adds its own vertical space.

@DavidCarlisle But the coffin should get typeset with the currently values of everything and the result saved in the box? The problem is, I think, that the coffin is local and I need something which survives the end of the group for it to work in the page hook when the content exceeds a page.

@cfr Oh, so far I did not get such nasty ones...

@santimirandarp not alignment but vertical space when a float is re-inserted into the document it gets vertical space added (\floatsep and related parameters) so you don't normally want vertical space inside the float box as you would get from center

Gonna be viva-ed on my thesis on Tuesday!

@Krishna good luck with that:-)

Thank you so much

I made that picture a transparent background in the blank page following my cover page by using the eso-pic package

@Krishna Where? (Country, I mean, makes a difference whether it is a formality or more than that.)

@Krishna Hmhh, always thought "eso" in "eso-pic" would not stand for esoteric... ;-) (Just kidding!!)

@JosephWright Thanks! After reading Mary Norris's book “Between you & me: Confessions of a a comma queen”, I got the impression that this is more specifically a New Yorker phenomenon these days; that the rest of the US has more or less left this diaeresis madness behind. (At least I think that's where I got it from.)

@cfr sorry I'm quite cueless, I didn't saw your code and even don't know if you get an error or something else. ;-) If I should look I would need the code too.

@JosephWright couldn't one at least have a file with a list for the failures? E.g. with lines like

@UlrikeFischer Same issue: would make it easy not to review manually ...

@JosephWright must say I think it's a non-argument, making people script the copying or script getting from the list of diffs to a save list doesn't increase the quality of any review.

@JosephWright I just updated the tlg in base and it as easiest to first copy them in, then check with git-diff that only the expected white space differences were in the tlg.

@JosephWright if I wanted to cheat I could always do it in other ways. But currently reviewing and resaving 10 tests is already quite a nightmare - I at least can't remember the names of ten failed tests. A list that would allow to check the handled files would make the reviewing much less error prone.

@DavidCarlisle What can I say, we have different workflows ...

@JosephWright meanwhile I just broke the build on travis:-)

@DavidCarlisle Ooops

@JosephWright oh not really me I just checked in tlg in base but it fails in cyrillic:

@DavidCarlisle you seem to have forgotten a few spaces.

kpathsea: Running mktextfm wnr10
/tmp/texlive/texmf-dist/web2c/mktexnam: Could not map source abbreviation  for wnr10.
/tmp/texlive/texmf-dist/web2c/mktexnam: Need to update ?
mktextfm: Running mf-nowin -progname=mf \mode:=ljfour; mag:=1; nonstopmode; input wnr10

@UlrikeFischer I only did l3build check in latex2e/base and they all pass here

@UlrikeFischer which spaces?

@DavidCarlisle the build shows space differences in tlb2888. ! (ot2enc.def

@UlrikeFischer yes as I say that's cyrillic, but it's failing to make a font with the error ^^^ I haven't touched any of the non-base files yet

@DavidCarlisle do you know if it ever worked? It seemed not to have stopped the test.

@UlrikeFischer oh I see it never worked. hmm what was it testing:-)

@UlrikeFischer I'm just concentrating on base at present (as I want to test the optional parameter list changes don't break that before pushing out more changes)

@DavidCarlisle then the travis failure is to be expected or not?

@UlrikeFischer yes travis's copy of texlive has picked up the latest l3build so all the tlg will need updating

@DavidCarlisle I was lucky - none of the luaotfload tests broke with it ;-)

@UlrikeFischer I was thrown by the font error, I didn't notice the actual test diff was just the expected white space

@UlrikeFischer I suppose I should update the tlg in required/* as well so travis passes again :(

@DavidCarlisle ;-).

@UlrikeFischer have I mentioned it's painful bulk updating tlg files:-)

@UlrikeFischer You realise this is all your fault?

@DavidCarlisle you can blame me - I made the tests that started all this ...

@UlrikeFischer I just did!

@UlrikeFischer in particular you used Windows, if your logs had been based on /usr/local/texlive/2018 then the tlg would have been the same and we wouldn't have noticed the system dependency.....

@santimirandarp -- what \centering doesn't do is turn itself off. for that you need some kind of group. of course, \begingroup ... \endgroup would work perfectly well for that.

@UlrikeFischer but joking aside I think I had better back out my experimental kernel changes and get all the tlg updated so travis passes

@barbarabeeton with a blank line before \endgroup

@DavidCarlisle but at least I didn't use miktex to create the tests - then we would perhaps have more diffs ;-)

@UlrikeFischer actually you should do that at some point, check they pass with miktex

@DavidCarlisle -- good point. (or \par, or, being very conservative, \endgraf. yes, i know you're going to tell me that those aren't latex. but sometimes it pays to not complicate things.)

@DavidCarlisle well at least for the luatex tests I wouldn't take bets. But currently it doesn't make sense as miktex hasn't the newest l3build yet.

@barbarabeeton they aren't latex (and neither are \begingroup, \endgroup:-) in any case normally it's in some environment such as figure that already scopes the centering and adds a trailing \par so none of the commands are needed.

@PauloCereda We blame you for all things abntex don't we?

@DavidCarlisle is the reason for the failure the non-ascii name of the option or something in the code?

@UlrikeFischer yes funny really I'm just testing an update to the option handling and then this... the utf8 characters get stuck in an edef by the option code and blow up...

@JosephWright ^^^ (my plan of having separate raw lists but leaving the original option handling don't help here)

@DavidCarlisle I don't find the option in the class. Who is using it?

@UlrikeFischer I'm not sure if there is a default option handler, I may ask the OP

@santimirandarp yes, but you need a blank line before the }

@santimirandarp there is an answer here from egreg

@DavidCarlisle I can't find it in latex styles, it means "hyphenation" but one wonder what it should do ...

@UlrikeFischer just re-checking babel, then I think I have updates for tools,amsmath,cyrillic and babel for latest l3build

@UlrikeFischer Happy to send it if you're curious, for sure. Would greatly appreciate your taking a look. Where can I find your email?

@cfr it is easy to find. Look e.g. at my profile.

@UlrikeFischer If I use \hbox_gset:Nn ... and typeset a coffin in it, will it survive the local group? That is, will it be typeset when I make the box and not when I use it?

@cfr yes, if I understand the question correctly.

@cfr I think it will be typeset as a box (but not usable as a coffin)

@PauloCereda Cwac.

@DavidCarlisle That's OK. It has to go in a box anyway. That's all I want to do with it. I just need it to survive.

@UlrikeFischer Thanks. You have mail. Sorry it is so large. I think I could probably construct a more minimal example now, if I'm right about what's wrong. But if I'm right about it, I don't need an example any more ....

@DavidCarlisle I think I'm getting the error when the content of the input handout is not longer than a page without the bibliography, but is longer than a page with the bibliography. So when I process the document in the main run, the coffin still has its content when it is called in the after-page hook, because that happens before the end of the group, which is on the next page. But when the image is externalised, the coffin is empty, because that happens after the group ends on the same page.

@UlrikeFischer travis happy again:-)

@cfr sounds likely, why do you need to use page hooks?

@cfr got It, but it is late now, so it will have to wait.

@DavidCarlisle ;-)

@DavidCarlisle I'm not actually sure. I'm putting a picture at a certain point on the page. I'm not certain why I'm doing this in a page hook, though.

\cs_new_protected_nopar:Nn \cfr_coursepacket_cefndir:n {% sylweddolir ar awgrymiad David Carlisle
  \begin{picture}(0,0)
    \unitlength=1mm
    \put(-25.4,25.4){% align to top left corner of paper
      \file_input:n { #1 }
    }
  \end{picture}
}

Apparently, the picture bit is based on something you suggested. Basically, the file input is a tikzpicture which takes bits of information about the title and typesets them inside nodes, with some other stuff jazzing it up a bit.

@cfr you might want to locally define inside that picture anything that might go out of scope

@DavidCarlisle Since when are you recommending answers of mine? :-)

@egreg yes I felt bad about it

@DavidCarlisle Don't worry: ManU lost to City and Juventus beat Milan. :-D

@egreg didn't know Juventus played cricket

@DavidCarlisle @egreg Now that @samcarter's tikzling package is out, you have no longer an excuse for keeping your profile pics. If you wait too long, the best options will be gone. (OK, looking at my profile pic, I realize that the best option is already taken, but there are still very nice ones. ;-)

@marmot but @egreg and I use natural photographs of our real selves, we don't need tikz