@JosephWright @DavidCarlisle @egreg Is tex.stackexchange.com/q/512689 a tabu/colortbl issue? If so, can it be fixed or is ther an alternative to ot-tableau?

This is pdfTeX, Version 3.14159265-2.6-1.40.20 (TeX Live 2020/dev)

@JosephWright ^

@DavidCarlisle Oooh

@egreg thing is that this list is at the begin of a chapter and has some introductory and requirements for that chapter. Afterwards follows the first paragraph of the chapter, so that one shouldn't be indented no matter what.

@DavidCarlisle What's the trick then?

@JosephWright that was Build --no-x-windows then it stopped in lacheck, so cd Work/texk; make then it built lots of stuff but stopped in detex, but it had built pdftex luatex and xetex by then. So not great but almost there....

@DavidCarlisle OK

@DavidCarlisle I'm trying reading the .travis.yml to track everything down, but it's got a Docker image of TeX Live (!) which I'd like to avoid

@Skillmon Then the first paragraph should be indented, as it doesn't appear just after a title.

@egreg I don't think so. Would you indent the first paragraph after an inspirational quote? Or after an abstract? Imho the first paragraph shouldn't be indented, this also holds true if there is a block prior to it that is typeset with distinct margins.

@egreg (in the actual non-minimal environment the lists body is 4cm narrower than the main body width)

@DavidCarlisle OK, I've got something building pdfTeX ... now, do I add a TeX system so it's testable? Hmm, don't remember how I did this last time

@JosephWright what are two trying to do? Test development version on travis?

@UlrikeFischer pdfTeX updates: PDF2.0 primitives

@UlrikeFischer I want to have enough time to test properly ...

@UlrikeFischer I did this last year for XeTeX, but don't remember the 'recipe' ...

@JosephWright oh good. Can't one install a binary from something like tlcontrib?

@JosephWright last time I just copied the binaries into my main texlive area and ran them over the existing tree

@UlrikeFischer Probably, but what I need to do is make sure I get a working pdfTeX

@DavidCarlisle Sure, but that's your main system: I'm building on WSL whilst my working system is w32. So I'd need to cross-compile for that to work, and w32 is not set up on Travis-CI so I've no real clues

@JosephWright oh well yes I recall now, that's why I tried so hard to keep it working on cygwin, as if I built a wsl binary I'd need a full wsl texlive tree....

@JosephWright that said you should be able to set TEXMFHOME in your wsl shell so it finds your windows texlive tree

@DavidCarlisle I've grabbed the Ubuntu texlive package, copied my new pdfTeX binary to /usr/bin: seems to work

@DavidCarlisle Right, I got everything to build!

============================================================================
Testsuite summary for TeX Live Scripts 2020/dev
============================================================================
# TOTAL: 1
# PASS:  1
# SKIP:  0
# XFAIL: 0
# FAIL:  0
# XPASS: 0
# ERROR: 0
============================================================================
make[6]: Leaving directory '/home/joseph/texlive-source/Work/texk/texlive'
make[5]: Leaving directory '/home/joseph/texlive-source/Work/texk/texlive'

@JosephWright yes that's probably simpler if you have space for a new tree, otherwise you should be able to see your existing windows one. eg my cygwin one looks like /mnt/c/cygwin64/usr/local/texlive/2019/

@JosephWright any magic switches to get lacheck to build?

@DavidCarlisle @DavidCarlisle The trick is to read the .travis.yml file. There are some pre-install bits plus a list of Ubuntu packages to install. I did that, ignored Docker stuff, ran ./Build -C

@DavidCarlisle There are a few touch lines mentioning lacheck, plus some extra headers I've now got

@JosephWright touch lines such a hack:-)

before_script:
  - find . -name \*.info -exec touch '{}' \;
  - touch ./texk/detex/detex-src/detex.c
  - touch ./texk/detex/detex-src/detex.h
  - touch ./texk/gregorio/gregorio-src/src/gabc/gabc-score-determination.c
  - touch ./texk/gregorio/gregorio-src/src/gabc/gabc-score-determination.h
  - touch ./texk/gregorio/gregorio-src/src/vowel/vowel-rules.h
  - touch ./texk/web2c/omegafonts/pl-lexer.c
  - touch ./texk/web2c/omegafonts/pl-parser.c
  - touch ./texk/web2c/omegafonts/pl-parser.h
  - touch ./texk/web2c/otps/otp-lexer.c

@DavidCarlisle ^^^

@JosephWright yes I see it. When the make was stopping I wondered about using touch, but I thought that would be too much of a hack and I should investigate the proper way....

@DavidCarlisle Then there's apt-get install -y --no-install-recommends bash gcc g++ make perl libfontconfig-dev libx11-dev libxmu-dev libxaw7-dev build-essential

@DavidCarlisle Anyway, now I know what to do, and can just use make in the texk dir once I've done some edits

@DavidCarlisle I should make some notes somewhere: perhaps a blog entry ...

@DavidCarlisle, @u

@JosephWright yes but I have those anyway that's just installing the needed utilities, but touching files to make them look new when they are not is "the make file is broken, let's work round it"

@DavidCarlisle, @UlrikeFischer With a build system working, I'll sort out the \pdfmajorversion I think while we are thinking about next steps for Secret Plan

@DavidCarlisle Ah: I didn't have all that stuff ...

@JosephWright I'll have another go on cygwin, otherwise I could always use wsl as well.

@DavidCarlisle It should be much the same, no?

@JosephWright compiling stuff on cygwin is what I'm supposed to do with this machine

@JosephWright in theory but sometimes harfbuzz or icu or some other random package doesn't build on cygwin even upstream as it's not first priority platform for testing. It always gets sorted out "in time" but if there issues there it's easier just to go with the linux build on wsl

@DavidCarlisle ooh

@PauloCereda @JosephWright the BBC is concerned about your well being bbc.co.uk/news/uk-england-derbyshire-50081386

@DavidCarlisle ooh auntie

@DavidCarlisle They used the WhatsApp emoji in their pamphlets. :)

@JosephWright with those touch lines, make in Work/texk builds without error, so I have built everything under texk, may go back and try a full top level build later but that's enough for now.

@DavidCarlisle Cool

@DavidCarlisle I'll start on \pdfmajorversion later: we really do need this (OK, it's in LuaHBTeX ...)

@JosephWright just symlink pdftex and xetex to luahbtex, add a couple of compatibility macro packages and call it done?

@DavidCarlisle At some stage, I think that probably is the future ... but there is still a bit to do!

@DavidCarlisle Did you read Arthur's article in TUGBoat about XeTeX?

@JosephWright yes, I think I had seen a draft previously but maybe just emails. We should really bring him on board to working on adding any xetex compat via lua code over luahbtex, really another engine isn't what we want.

@DavidCarlisle One for the team list, but probably we (you/me/@UlrikeFischer) should draft a mail to him from the team as a whole

@DavidCarlisle I think he'd be best to look at some Lua code to emulate the interchar stuff reliably, and call it a day for XeTeX, really: LuaHBTeX does do everything else

@DavidCarlisle For example, I have thought about trying to add catcode tables, but the effort doesn't look worth it

@JosephWright yes interchar stuff is the tricky one. I think you can meet any real use cases using the existing lua callbacks but it is quite hard to make a complete drop in replacement as the timing of when interchartoks is a bit odd, the input buffer callbacks are too early and the node related callbacks are too late (but probably better meet the real need) if we had ended up with harftex, an easier option would have been just to add the primitives, but that may be harder with luahbtex

@DavidCarlisle Yes, my understanding from Hans is that it's not a required feature, in the sense LuaTeX can already do all the stuff

@JosephWright that depends how you define "can already do"

@JosephWright I'm tempted to ask on xetex list for any non trivial examples that could be used as test cases for luahbtex

@DavidCarlisle Good plan

Meanwhile, starting on \pdfmajorversion, trying to work out where to handle \pdfinclusionerrorlevel: I think that really applies at the C end ... would helpf i I knew C (or indeed WEB)

I think I'll sort the other part first ...

@JosephWright has anyone ever set that to anything?

@DavidCarlisle Good question

.

@JosephWright @DavidCarlisle as I'm seeing Luigi next week I could try to speek with him about this, but I would perhaps need a clearer idea what is needed for a suitable interchar implementation @MarcelKrüger

@UlrikeFischer If he were to do anything I think the most natural luatex-like thing to do would not be implement \interchartoks in the luatex sources but add a lua callback at the point that xetex adds those tokens, then you could implement the xetex behaviour easily in Lua. But I haven't compared the xetex and luatex sources to see if that is a reasonable thing to suggest.

@UlrikeFischer it may already be possible to hook to more or less the right place with the ligaturing or kerning callbacks (which also deal with inter-character behaviour) the luatex manual does not inspire confidence in putting too much code there: Messing too much can push LuaTEX into panic mode.

@UlrikeFischer ooh

@PauloCereda wrong Luigi

@DavidCarlisle oh

THIS WOMAN IS MEAN

@DavidCarlisle I don't think that's the right point. In LuaTeX, ligaturing and kerning are called directly before hyphenation, so only after the full node list for a paragraph is already constructed. E.g. changing the current font as it is often done in interchartokens wouldn't have the intended effect there.

@JosephWright Maybe it's too late (you've already working on Cygwin) but I can provide some information on building TL win32 binaries; currently I have no trouble using my own win32 binaries github.com/aminophen/w32tex-build, including pdftex and xetex. (I couldn't build luatex for TL2019, sorry)

@MarcelKrüger yes that was the conclusion I came to when I last looked at this (that the right point wasn't available) I just had a quick look down the list while saying that and wondered if they would be near enough. Of course it doesn't help that the xetex manual is totally silent about when these are added with respect to expansion.

@MarcelKrüger .. The xetex timing is really weird, it is after the first token is all expanded to character data and gone to the hlist but the next token is expanded just far enough to know its first non expandable token is a character of the right class, so if you insert tokens that grab arguments you see rather arbitrary internal state of the macro expansion.

@HironobuYAMASHITA I am in total denial that this is a windows machine:-)

@MarcelKrüger I suspect most real use cases could be addressed with the process_input_buffer, but of course that's pre macro expansion so isn't really the same thing...

@HironobuYAMASHITA Would be great: I'm not using Cygwin, just WSL or a WM

@DavidCarlisle I think that the beginning of hyphenation would be the best place for most actual usecases, but that does require a full rewrite of interchartoken code.

@MarcelKrüger yes.. Do you have any feeling how hard it would be to move the xetex primitive to luatex (or add a callback to the same place that would allow a lua implementation)

@MarcelKrüger but if the real use cases are adding spacing for classifications of cjk characters the existing luatex node level callbacks are most likely fine and interchartoks isn't needed at all. Hard to know what people use it for..

@DavidCarlisle I fear that a callback at that point would be too slow, even if it would be relativly easy to implement. I'll have to look into XeTeX to see how it works, but I think porting the primitives should be relativly easy.

@DavidCarlisle Looking at the PDF version stuff, I'm going to have quire a knock-on in the C part: currently the major version is assumed, so we need an extra parameter all over the place :(

@JosephWright yes I had a quick look last year, the alternative is internally (or even externally) allow minorversion to be >100 with(1+) the hundreds part being output as the major version that would require fewer changes but may be too weird, however not that dissimilar from luatexversion=110 meaning 1.10 not separate minor and major version primitives

@JosephWright Joseph leading the way! duckoftheday.co.uk/2019/10/16-oct-2019.html

@UlrikeFischer ooh a parade

@DavidCarlisle Yes except the C code does check minor and major version as the 'external' PDFs are set up that way

@JosephWright sounds vaguely familiar (adding the separate integers is probably more maintainable really anyway..)

@PauloCereda Please, is there a way to link specific search results in texdoc?

@yo' you mean, the online search?

@PauloCereda yep

@yo' There's the hyperlink for each result, but apart from that, there's no way... :( I could implement something...

@PauloCereda hmm REST .... :-)

@DavidCarlisle ooh I like it

@PauloCereda http://texdoc.net/?query=duck

@DavidCarlisle good idea. :)

@PauloCereda especially if it returned a load of menu suggestions

@DavidCarlisle texdoc.net/pkg/duck is close, but it's like "I am feeling lucky" from Google.

@DavidCarlisle or simply http://texdoc.net/q/duck (but this takes some directory-handling fu)

@yo' yes!

@yo' as long as it returns Chinese crispy duck pancakes, I don't mind the URL format.

@DavidCarlisle just please, don't replace the first two letters with fu

@DavidCarlisle oh no

@yo' it's can be done through rewriting URL rules. :)

@PauloCereda nicely illustrated: bbc.com/news/uk-politics-50084090

@UlrikeFischer ooh

@UlrikeFischer ^^ :)

@PauloCereda now Joseph is offended

@DavidCarlisle oh no

wait

@DavidCarlisle ^^ :)

@PauloCereda as I work for a UK-based company, should I get one of these?

@yo' ooh hold on

@yo' ^^

@PauloCereda :) mine are at the customs :( just now I'm filling in the paperwork

@yo' oh

@JosephWright I get an error ! LaTeX3 Error: Backend configuration already set. with the following:

\begin{filecontents}{driver-test.sty}
\RequirePackage{xkeyval}
\DeclareOptionX{dvisvgm}{%
  \PassOptionsToPackage{driver=dvisvgm}{expl3}
}
\ProcessOptionsX*
\RequirePackage{expl3}
\endinput
\end{filecontents}
\documentclass[dvisvgm]{article}
\usepackage[dvisvgm]{driver-test}
\begin{document}
blb
\end{document}

It somehow has to do with xkeyval, but I don't understand what is happening.

@UlrikeFischer -- one problem with that theory: the duck in front is a female mallard. Regardless of what Joseph's avatar looks like.

Hand position for typing in $\LaTeX$ - this question was put on hold on Mathematics. I suppose it would be off-topic on TeX - LaTeX, too, right?

Hmm, now I have noticed that this was already asked here: Hand position for typing in Latex. So ignore my message above.

@DavidCarlisle Ok. (Sorry for the slow reply.)

@DavidCarlisle See fun with spaces in file names!

@JosephWright I was just looking at your comments, but not sure I followed it (just got in) what's the bug?

@DavidCarlisle Well \pdffilesize{name.tex } should fail as it's not the same file name as \pdffilesize{name.tex}, but it doesn't

@JosephWright ah

@DavidCarlisle Our older code (using \ifeof) must have the same bug, not allowing for a space-at-end-of-name

@DavidCarlisle LuaTeX emulation gets this right

@JosephWright I can test but does \pdffilesize{"name.tex "} do the right thing?

@DavidCarlisle No

@JosephWright well fortunately we know someone with the pdftex sources open in an editor....

@JosephWright I get

\documentclass{article}

\begin{document}

\typeout{no space \pdffilesize{name.tex}}
\typeout{space \pdffilesize{name.tex }}
\typeout{space \pdffilesize{name.tex" "}}
\end{document}

no space 3
space 16
space 16

so the first one checked name.tex and the second two checked name.tex with pdflatex?

How do you managed to get get two values? Do you have a file with a space at the end?

xelatex gives

no space 3
space 16
space

@UlrikeFischer doesn't everyone:-)?

@DavidCarlisle I tried to create one but windows didn't let me do it ;-(

uplatex gives same as pdfatex

@UlrikeFischer echo "abc" > "name.tex " in the wsl bash

@JosephWright \typeout{space \pdffilesize{"name.tex "}} finds nothing in xetex and the file with a space in pdftex which is a bit of a pain as it means you can't consistently " surround filenames, need to unquote here.

Have there been changes in the kernel or anything else that would cause pdfcrop to behave differently from earlier? A package I made that uses it doesn't crop correctly any more.

@AlanMunn possibly.

@DavidCarlisle Hmm. I'm trying now with an older version of TL. I'm just using pdfcrop foo.pdf > bar.pdf and it's cropping far too much.

@AlanMunn lots of commands (including \begin and \end) that were fragile are now robust, and filehadling has changed 9to allow spaces and non ascii characters in more places) so it depends what pdfcrop was doing...

@AlanMunn oh more likely it's ghostscript, let me dig out a thread...

@DavidCarlisle Yes, that's what I'm thinking now.

@AlanMunn ghostscript removed a whole load of functions in a release a few months back..

@AlanMunn tug.org/pipermail/tex-live/2018-September/042358.html

@AlanMunn not sure that directly affects pdfcrop but..

@AlanMunn pdfcrop does use -dSAFER :

#   2009/07/16 v1.16: Security fixes:
#                     * -dSAFER added for Ghostscript,

@DavidCarlisle Hmm. So this is annoying. And switching to a previous TL doesn't help, since Ghostscript is installed separately.

@AlanMunn you could confirm if that is the issue by editing the perl script and removing -dSAFER

@DavidCarlisle Hmm. No that seems to make no difference.

@DavidCarlisle So it must be something with my code. It's a tiny package that has worked for years. Let me see if I can make a MWE.

@AlanMunn blame @UlrikeFischer

@DavidCarlisle Ok here's a MWE:

\documentclass{article}
\usepackage{fontspec}
\pagestyle{empty}
\begin{document}
\shipout\vbox
{\parbox{1.75in}{%
\centering
\raggedleft
\vspace{.25in}\par\fontsize{11}{13}\selectfont
Some Text\par\vspace{12pt}
\fontsize{10}{12}\selectfont
Some more text\par%
\vspace{12pt}\fontsize{8}{10}\selectfont
Even more text\par
\vspace{10pt}
Something else\par}
}
\end{document}

Output from pdfcrop:

luatex or xetex?

@DavidCarlisle Same with either. xetex preferably in this case.

@AlanMunn let me see..

@AlanMunn same here,

@DavidCarlisle Reproducibility is a start. :) So who should I blame, apart from @UlrikeFischer?

@AlanMunn not sure, it seems to be the initial bbox calculation, if I use --verbose so it outputs the bounding box and add 5 to the 3rd and 4th component so pdfcrop -bbox "139 673 215 750" cc271.pdf new.pdf it works but no idea where that calculation went wrong

@DavidCarlisle Should I post to the TL list? Or is this sufficient for a report?

@AlanMunn TL list I guess (I am not sure I can trace this much more) or you could try Heiko, but he doesn't always answer mail these days, his tex time is limited

Hello friends. Which math font do you recommend to be used with \usepackage[sfdefault,lf]{carlito} in size 40pt?

Maybe \usepackage{sansmathfonts}?