TeX, LaTeX and Friends

3:28 AM

@yo' That works great, thanks!

6:07 AM

@DavidCarlisle Something loosing more chars: see raw log for tlb0380 for example. Starts at ^^Z and goes on to what seems to be an arbitrary point. I guess this is a 'Lua side' issue too ...

Joseph Wright

6:37 AM

@DavidCarlisle I think we are going to have to look at alternative ways of grabbing the file contents or get Hans to write ^^ notation, at least for 0/127

David Carlisle

6:49 AM

@JosephWright I could look more at the sketch I posted here the other day, you can get the entire file with no dropped chars as a string, you then just have to pick your way to normalize that as a series of "lines" in some system independent way without using any lua function that has undocumented bad behaviour on control characters....

Joseph Wright

@DavidCarlisle Yes, I think that's the way we'll have to go

David Carlisle

@JosephWright is it worth pushing back to Hans? (probably using native Lua I/O is a defendable position for them, even if it seems rather flaky) :(

Joseph Wright

@DavidCarlisle I can live with the native Lua handling of ^^@ and ^^[ if they would write those awkward values to the log in classical TeX style ...

David Carlisle

@JosephWright well the regression-test and test2e files could inject a suitable callback. at least I thought there was a callback there but the one for \write explictly doesn't apply to the log file...

Joseph Wright

Something like:

logfile = "tlb0380.log"
file = io.open(logfile, "rb")
s = file:read("*all")
s = s .. "\n"
for line in s:gmatch("([^\r\n]*)[\r\n]") do
  print((string.gsub(line, string.char(0), "^^@")))
--print(line)
end

@DavidCarlisle ^^^?

David Carlisle

6:58 AM

@JosephWright yes

Joseph Wright

@DavidCarlisle OK, I'll adjust the code (again) later on: off to work in a second

@DavidCarlisle Means I'll have to revert all of last night's changes ...

David Carlisle

@JosephWright yes same here, need to be in the car "before 8...."

@JosephWright We should send a time sheet to Luatex list how much time spent tracking Luatex changes over last year

Sebastiano

8:08 AM

@egreg I have removed, now. Thanks a lot lot. You are always authorized, if I make mistakes, to edit my questions and answers

Joseph Wright

8:18 AM

@DavidCarlisle Think I have this working: checking everything over now (lots of test runs so not much to actually do, just let the laptop grind away)

David Carlisle

8:28 AM

@JosephWright OK yes it looked plausible, but when I tried it I tried to merge \r\n and ended up never reporting a blank line, which meant the subsequent l3build line wrapping took different choices and showed lots of diffs. I think you are keeping more information in the initial split

@egreg ooh you just pipped me with the Greek answer but with the development latex format I needed to add \DeclareTextCommand{\accdialytikaoxia}{TU}{}% temp defn otherwise get an error that it is not defined. I wonder if I should make that a warning not an error.... (@JosephWright) (composites with the accent are declared but that never gave an error just defined something unusable due to \csname defines \relax) so I added a check for that but.....

Also, without fontspec do you get an error on \< or is that something else just in dev sources?:

\documentclass{article}
%\usepackage{fontspec}
%\setmainfont{CMU Serif}
%\DeclareTextCommand{\accdialytikaoxia}{TU}{}% temp defn
\usepackage[main=british,polutonikogreek]{babel}
\begin{document}
{\selectlanguage{polutonikogreek}
οὔ μοι τὰ Γύγεω τοῦ πολυχρύσου μέλει,}
he said
\end{document}

gives

! LaTeX Error: \< not declared in encoding LGR.

Joseph Wright

@DavidCarlisle Still not quite sorted on Windows (I think a line ending issue)

David Carlisle

@JosephWright do you want [" .. os_newline .."]" or " .. os_newline ? in unix [\n] is same as \n but in windows [\r\n] will match each of \r and \n separately, but I think you want \r\n which just matches the pair, once.

Joseph Wright

@DavidCarlisle Could well be :)

@DavidCarlisle As they've changed the behaviour so we get Unix-like line ends on Windows, at some stage I'll be able to simplify this (depends how many TL versions back we want to support)

egreg

8:47 AM

@DavidCarlisle No error

@DavidCarlisle But the hyphenation patterns are for Modern Greek (it says polytonic, though).

David Carlisle

@egreg ah thanks it is the same error/issue it ends up declaring composites of \< with base characters without declaring \< as an encoding-specific command first.

@egreg I suspect I should make that a warning (I already had to update hyperref's PU encoding which was doing that in some cases) (fixed at the last hyperref update)

@egreg I'll take to team list....

David Carlisle

9:07 AM

@egreg you've got mail (I hope:-)

egreg

9:20 AM

@DavidCarlisle Greek has always been a PITA.

Joseph Wright

@DavidCarlisle See Rainer's mail!

David Carlisle

9:35 AM

@JosephWright it was simpler in 1985

so.... @JosephWright, @egreg did you get the message about accents that I just sent to the team list?

Joseph Wright

@DavidCarlisle I did :)

egreg

@DavidCarlisle Yes

David Carlisle

thanks.

egreg

@DavidCarlisle Less simple a few years earlier, though: I remember very long addresses with all possible combinations of @, %, ! and whatnot.

David Carlisle

@egreg oh yes that's true (did you ever use vvencode: the funky tex-variant of uuencode used to pass tex through bitnet/janet gateways without completely destroying them:-)

egreg

9:41 AM

@DavidCarlisle Yes!

@DavidCarlisle I think I installed it in my work space on the Vax/VMS mainframe.

David Carlisle

@egreg the whole of 2e was built based on a distributed source control server Rainer constructed out of RCS and vvencoded mail

@egreg @JosephWright is too young to remember these things:-)

Paulo Cereda

@DavidCarlisle that means you are actually... well... not young? :)

David Carlisle

@PauloCereda age is comparative, compared to @egreg I'm very young

Paulo Cereda

@DavidCarlisle ooh

ilFuria

@DavidCarlisle Age is a state of mind

Joseph Wright

10:43 AM

@DavidCarlisle BTW, I've never even worked out how @latex-project.org mails work: do you know?

David Carlisle

@JosephWright they go to Rainer and he redirects them to wherever you asked for them to be redirected to:-) so david.carlisle @ ... goes to my usual gmail address (not that I use it much)

Paulo Cereda

@JosephWright a dot file in your user area, perhaps? :)

In general, a .forward file. :)

egreg

11:39 AM

@PauloCereda What happened with Ponte Preta? :(

Paulo Cereda

@egreg oh no :(

egreg

@PauloCereda Well, I woke up with a list of notifications ending in the dreaded 1:0

Paulo Cereda

@egreg Indeed. It looks like Ponte Preta's defense was pretty dense, so they scored and retreated to their own field. Thankfully, we were already qualified to the next round. Now we cannot afford losing more matches. :)

Joseph Wright

@DavidCarlisle I think all is now working ...

David Carlisle

@JosephWright want me to run it here? or do you have testsuite files to check in

Joseph Wright

11:54 AM

@DavidCarlisle Give it a try, please

@DavidCarlisle I hope I've sorted the line-end business!

David Carlisle

@JosephWright yes I looked at the diffs in mail, re-installed l3build and running check in 2e base now

barbara beeton

@yo' -- do get well. chicken soup. i hope you find someone to drive with you; yes, it is a long way, but you'll be missed if you can't come. and i hope work on your package goes smoothly.

Will Robertson

12:12 PM

@JosephWright @DavidCarlisle @egreg Sorry I missed you the other night… it's been a busy week/month

David Carlisle

12:31 PM

@WillRobertson we waited for a couple of seconds then decided you probably had better things to do at that time of night than chat to us:-)

Paulo Cereda

@DavidCarlisle play videogames

David Carlisle

@PauloCereda possibly. Unlike some people he has time for such things as he doesn't have a thesis to write.

Paulo Cereda

@DavidCarlisle oh no

barbara beeton

@JosephWright -- someone, probably trying to be nice, has just upvoted 4 answers of mine, raising the vote total on each to 25. but this might well make them (both the votes and the voter) subject to "serial voting" rules, which would have a negative effect in the end. is there any way to warn the voter?

Paulo Cereda

Hershey's chocolate bar... hmmmmm

Joseph Wright

12:45 PM

@barbarabeeton No, afraid not (they won't show up in vote analytics at that level)

@WillRobertson :)

@WillRobertson That way you avoid action points!

Will Robertson

@JosephWright Yeah, good luck with that… I'm already behind reviewing a couple of journal papers; should never have agreed to take them on! (Actually I did manage to push out a fontspec release a brief while ago. At the expense of my reviewing...)

David Carlisle

  All checks passed


real    56m41.154s
user    31m23.534s
sys     24m10.256s

@JosephWright ^^^^

Joseph Wright

@DavidCarlisle Yay!

@DavidCarlisle Slow :(

Paulo Cereda

@DavidCarlisle Oh my! What was it?

Joseph Wright

@PauloCereda LaTeX2e test suite using LuaTeX v1.05(ish)

Paulo Cereda

12:57 PM

@JosephWright Wow

David Carlisle

@JosephWright as I say it does seem a bit slower recently (but that's a lot faster than last time I ran it which took about three hours) running Java at the same time to generate few thousand xml files may have an effect.

Paulo Cereda

@DavidCarlisle SAX?

David Carlisle

@JosephWright how long does it take on your machine?

@PauloCereda xslt (saxon)

Paulo Cereda

@DavidCarlisle oh

Joseph Wright

@DavidCarlisle I'll check: it's a bit tricky as I don't have Unix time on the Windows side

@DavidCarlisle Once v1.05 is available for Mac I'll try 'natively'

David Carlisle

12:59 PM

@JosephWright yes I just added time this time as it was so slow the opther day 9about three hours clock time) so I thought I'd check...

Paulo Cereda

@JosephWright want me to write a .exe timer?

319

Q: How to measure execution time of command in windows command line?

Can this be done with standard means?

@JosephWright: maybe PowerShell can help

David Carlisle

@PauloCereda since it's just a single call to Lua it probably suffices to put print(time()) on first and last line:-)

Joseph Wright

@PauloCereda Probably, but my work PC is slower than my laptop, and on my laptop Windows is a VM, so really I want to know what the Mac timing is

Paulo Cereda

@DavidCarlisle That too. :)

@JosephWright Oh I see

@DavidCarlisle meanwhile, texdoc woes. :)

David Carlisle

@PauloCereda I wonder if I can find a link to the transcript of Jan 12 last year......

Paulo Cereda

1:08 PM

@DavidCarlisle uh-oh

Joseph Wright

1:54 PM

@DavidCarlisle Takes about 45 mins here on the VM: will be faster when I can run it natively

@DavidCarlisle I get a failure for tlb4173 (babel): correct?

@DavidCarlisle Sorry, tools not babel

David Carlisle

@JosephWright I only ran base, will check the others...

Joseph Wright

@DavidCarlisle Ah

Will Robertson

@JosephWright — every now and then I struggle through a few more Travis problems with fontspec btw.

! )) (../unicode-math.sty
****
! ))
! (../unicode-math.sty

Sigh. Any suggestions in the general?

David Carlisle

@WillRobertson well... this morning's new line normaliser might do something different:-) but in general as it (now) grabs the entire log file as a single string there would be the possibility of doing a few extra normalisations at that level before iterating through each "line" of the tlg.... (@JosephWright will be so pleased I suggested this)

  Check failed with difference files
  - ../../build/test/tlb4173.etex.diff
  - ../../build/test/tlb4173.xetex.diff

@JosephWright 4173 fails here as well ^^

Will Robertson

2:10 PM

@DavidCarlisle I'll give it a shot tomorrow hopefully. I'm doubtful that we can really fix this sort of thing up but yeah.

Anyway, falling asleep -- good night :)

David Carlisle

@WillRobertson give me write access for a day and I'll make the packages write about 10000th the amount of data to the log file, then there would be less to clean up:-)

daleif

Anyone remember anything that needed to be added to mathtools and friends (either feature requests or bugs that needed to be fixed). On my list of fixes for the coming versions I have

MARCH 2017

* fixed bug in empheq in relation to AMS clases
* renamed macros such that any xxx_yyy: macro should now contain EQ,
MH or HT as a prefix or infix. Thus making everyting truely separate from expl3.
* changed \shortintertext such that as long as 'original-intertext=false' (the default),
\shortintertext now only work in the same context as \intertext
* \shortintertext is now also an error inside the empheq construction

JANUARY 2017

* sporadic space in \noeqref

MAY 2016

* Added \ignorespaces after \noindent, bug inherited from amsmath

David Carlisle

  Check failed with difference files
  - ../../build/test/tlb0735.pdftex.diff
  - ../../build/test/tlb0852.pdftex.diff
  - ../../build/test/tlb3505.pdftex.diff
  - ../../build/test/tlb3649.pdftex.diff
  - ../../build/test/tlb3976.pdftex.diff

@JosephWright do you get that in babel ^^^^

Joseph Wright

@DavidCarlisle ER, good

@DavidCarlisle Haven't checked babel yet: will do

@WillRobertson I'll take a look: I have a VM set up which lets me do 'deep' inspection

@DavidCarlisle :) Don't tell FMi

David Carlisle

2:25 PM

@JosephWright ah babel ones look broken by the breaking out of the individual languages combined with our sandboxing eg first one I looked at is + ! Package babel Error: You haven't defined the language austrian yet. It may make sense to rename testfiles-TU as testfiles-not-sandboxed and put the babel tests there....

David Carlisle

2:45 PM

@JosephWright tlb4173 is fine it's \verb+amsthm+ so shows the new no-hyphens language setting in verb (I'll update this evening)

Joseph Wright

3:09 PM

@DavidCarlisle I guess I still need to sort out the TU business: I'll ad that to my 'to do' for l3build

David Carlisle

@JosephWright yes but I suppose it's up to Javier to decide if the babel tests should restrict just to the core or use a normal format, probably makes sense to do the latter so you can actually test language switching.

Joseph Wright

@DavidCarlisle I mean finding a way of having the TU tests run from the top level

David Carlisle

@JosephWright yes (just following up on my earlier comment about adding the babel tests there) although actually that was nonsense babel tests already use a full search tree, the issue is that I don't have austrian installed so I get ! Package babel Error: You haven't defined the language austrian yet. even using my installed format.. I suppose I just need to install some more babel and re-run...

Joseph Wright

@DavidCarlisle I have 4 failures so makes sense

David Carlisle

5:45 PM

@UlrikeFischer keeping me busy again? :-)

cfr

6:08 PM

@DavidCarlisle @JosephWright @anyone Is it possible to set default options in a class so that packages which pick up global options can pick them up? That is, without passing them explicitly to all packages which might be interested.

David Carlisle

6:26 PM

@cfr yes (although I'm not sure it's a supported interface)

@cfr \edef\@classoptionslist{\@classoptionslist,zzzzz}

cfr

@DavidCarlisle Is there somewhere I can read about it? I'm not terribly bothered if it is unsupported for current purposes.

David Carlisle

@cfr the line above adds zzz

cfr

@DavidCarlisle So after \ProcessOptions before \LoadClass or?

David Carlisle

@cfr I thought you wanted to affect packages loaded in the document preamble, so anywhere in the package would do

the way global options work is that \documentclass saves the option list in that macro so any package can see those options in addition to options on the package. So if you append to that list all packages will see it. Can't say I have ever tried, but what could go wrong? :-)

cfr

@DavidCarlisle Thanks.

Paulo Cereda

6:32 PM

@cfr: quack!

cfr

@DavidCarlisle I was thinking the class needs them too, but, actually, I doubt it will care.

@PauloCereda Cwac!

Paulo Cereda

@cfr <3

cfr

@DavidCarlisle Does it matter if \makeatother gets executed in \AtEndOfClass?

David Carlisle

@cfr well if you do it at the start, before the class itself does \ProcessOptions then it wil see them as well

@cfr probably not but why have it there at all?

@cfr why would you need \makeat... in a class file?

cfr

@DavidCarlisle I want to input a file which I usually input in the preamble of documents. So it includes \makeatletter ... \makeatother. I can switch it back on in the class, but it looks odd. ;)

@DavidCarlisle If I knew what question to ask, I'd ask how to do it properly. Unfortunately, I have no idea.

David Carlisle

6:53 PM

@cfr if it was a package @ would be handled automatically:-) the catcode will be reset anyway at the end of the class so an extra resetting won't do any harm unless you have a file that is being input at the end of the class that assumes @ is a letter, but that seems unlikely

touhami

7:51 PM

@DavidCarlisle hello :-) please any suggestion here tex.stackexchange.com/questions/360993/… it seems like a bug

David Carlisle

@touhami sorry too much information although there seems to be some confusion over the page size as set by geometry and the position of the title as set by titlesec but the image indicates the OP was measuring the top margin to the top of the letters which seems the wrong thing to do.

touhami

@DavidCarlisle ok but if you have time please see my edit

Ulrike Fischer

8:26 PM

@DavidCarlisle what did I do?

David Carlisle

@UlrikeFischer told someone on a German forum with a question about 10 different table packages doing something strange if combined to contact me:-)

Joseph Wright

8:49 PM

@DavidCarlisle Ghostbusters?

David Carlisle

@JosephWright ?

Joseph Wright

@DavidCarlisle 'Something strange' you've been ask to fix ;)

David Carlisle

@JosephWright ah.

Ulrike Fischer

@DavidCarlisle Did I? I don't remember but I apologize anyway ;-)

David Carlisle

@JosephWright one of the packages involved wasn't mine so I'll probably be able to blame that (I haven't unpacked the attachment yet:-)

Joseph Wright

8:52 PM

@DavidCarlisle Good plan

David Carlisle

@UlrikeFischer well the mail started by blaming you:-)

yo'

@barbarabeeton Moreover, the deadline for cheaper fees is today, so I'd need to decide now :-/

barbara beeton

9:14 PM

@yo' -- oh. (gloom) i guess we shouldn't expect to see you then. and i know next year will be even more expensive, and it's too much to hope for that you'll find a pot of gold before then. maybe after i retire in a year or so we can put on our traveling gear and wander back to your part of the world. maybe there will be a tex meeting in a location closer to you, in a language i can cope with.

yo'

9:27 PM

@barbarabeeton well, originally I was considering applying for TUG19, but I'll still have a research grant by then and whence I shan't occupy with "fun" ...

hey, why is LuaLaTeX printing my PDF as A4 even if I put a6paper in?!

David Carlisle

9:49 PM

@yo' tradition

@yo' geometry package?

geometry (still) needs luatex85 package

cfr

@DavidCarlisle How do you know when a file should be a package?

yo'

@DavidCarlisle yes, it is involved :-)

anyway, the font is scalable, so I kept a4 and reduced the size in print

David Carlisle

10:08 PM

@cfr if you input it into the preamble it should probably be a package, if you input it in the preamble and it uses @ it should almost certainly be a package.

cfr

@DavidCarlisle Even if it is pretty trivial?

Is it possible to use xindy for a glossary and makeindex for an index? I can't figure out how to stop glossaries trying to use the indexing files. It refuses to stick to its own.

David Carlisle

10:31 PM

@cfr why not? if you save it in foo.tex you have to tell people to \input{foo} and have to set/restore @ if you save it in foo.sty then you just tell people to use foo package and @ is automatic. It's unlikely to be more trivial than indentfirst package (which is exactly 4 tokens of code)

touhami

@DavidCarlisle ok found here tex.stackexchange.com/questions/79232/…

David Carlisle

@touhami ah, OK

cfr

11:18 PM

@DavidCarlisle Hmmm ... maybe ... 'People' is only me, though. Maybe 4 tokens of code gets multiplied by recipients.

barbara beeton

11:29 PM

@DavidCarlisle -- who's maintaining tikz these days. i've just had to work around a problem where \draw[transparent]... was leaving a visible line on the output. (since all it was doing was allowing for extra space in a table cell, a zero-width \vrule substituted quite nicely.) i'll trim it down to a minimum example, and if it still doesn't disappear the unwanted line, who do i send the evidence to?

cfr

11:57 PM

@barbarabeeton transparent isn't standard, is it? Or rather, it is only meant to be used for specifying fadings based on pictures.

Transcript for

$TeX - LaTeX$ TeX, LaTeX and Friends