@UlrikeFischer You are awesome! Tests are passing now.

@manooooh I have no time at the moment, but if you ask a question on the main site I'm sure some being will answer.

@MarcelKrüger I was a bit lucky that I removed something which made travis showing a sensible error message. But I have the impression that failing pdf/pvt-tests are more problematic than failing log/lvt-tests.

Ulrike is awesome

:)

I got a nice answer badge for telling that one can write Œuvres by writing Œuvres ;-)

@UlrikeFischer ooh

@PauloCereda don't believe everything you read on the internet

@DavidCarlisle oh

@PauloCereda you starred that comment? You are mean.

@DavidCarlisle oh no

ooh

@DavidCarlisle <3

@DavidCarlisle oh

@UlrikeFischer: twitter.com/SchwarzerTurm/status/1188431931580604418

@PauloCereda oh god, dancing chess ;-).

@UlrikeFischer :)

The image file name issue is a bummer.

It is biting me again.

@wilx Well, none of my document uses grffile. I consider packages that patch other packages so heavily fragile and so I always kept the rule to keep file names simple and avoid grffile.

@UlrikeFischer Well, I have real file names from real world, like from Twitter. I can't just rename them willy nilly.

@wilx I have real file names too, and I can rename them when saving them.

The "grffile does not work with LuaTeX any more · Issue #73 · ho-tex/oberdiek" has been open now for 21 days. Is there a solution that in sight that does not involve temporary workarounds?

@wilx @DavidCarlisle is working on stuff

@wilx depends a bit why you are using grffile: spaces should work without it, accents too (if you use the -cmdx option when calling lualatex). The only open problem are too many dots in the name.

The file name is BbyaCGZh-bigger.jpg.

@wilx then why using grffile at all?

@UlrikeFischer Inertia, I guess. I do not know LaTeX well enough, so my files are cargo-culted together until they work.

Has LuaLaTeX been much slower recently? It takes like 8 seconds to do one page now on my tweets document.

@wilx that's what's makes supporting latex hard, people post examples that never should have existed.... Harder than fixing grfffile is fixing svg include (I haven't fully traced all the places that breaks, as it has to call out to inkscape via shell escape without breaking filename quoting) and I don't want to have a series of small partial fixes, trying to fix them all in one go, but this is a "free time" activity ...

@DavidCarlisle I get it its free time activity. But this has bitten so many people now.

I get bitten much because grffile is part of Pandoc's default LaTeX template: github.com/jgm/pandoc/blob/master/data/templates/…

Maybe there is a way out of this by introducing gfffile2 or such and fixing everything there properly and deprecating the old one?

I don't know.

@wilx There are no cases where grffile is needed, and with the new filename quoting filenames with spaces work without doing anything special, and non-ascii filenames work (that didn't before) so the only thing that is inconvenient is files with multiple dots, and for those a one line workaround should fix it for now. Are there really so many people who have aaa.bbb.ccc.jpg files who have updated latex in the last month? It's not as if I have a complete fix and am just not releasing it

@wilx the problem is not grffile. The problem is getting the quoting right without it - when this works grffile can basically be dropped.

On a different note, is there a way to get a profile of LuaLaTeX run? 5 minutes for 46 pages seems excessive.

@wilx Yes, seems very slow

@wilx put an empty file grffile.sty alongside the document if you can't configure pandoc not to use it

@DavidCarlisle Is there a condition of like old TeXLive distro and the tools where it is necessary? I can propose a modification of the Pandoc LaTeX default template to remove it.

@wilx well it was never necessary (graphicx by design could handle multiple dots from the start) it just provided a slightly more convenient syntax. But it wasn't broken before the 2019 latex release so if you want documents using it to work as before just load it on latex releases that do not define \set@curr@file

does anyone knows some sensible tool to share/synchronize a list of (email) address with a few other people?

I timed it with time. Two runs of LuaLaTeX to resolve references: make all 1254,92s user 8,45s system 99% cpu 21:04,39 total

This is horrible.

@UlrikeFischer Shared Google Document?

@wilx Each run of my 63-page dissertation takes about 33 seconds, most of which are spent loading graphics. With the -draft option that reduces to about 8 seconds per run. Are you loading several pictures?

@PhelypeOleinik 20 images across 461 .tex files. About 4 to 10 .tex files per page.

It is tweets.

Each file is one tweet.

Each tweet is tcbox.

@wilx Well, tcolorbox has a lot of slow TeX code, but it probably doesn't justify the 10 minutes per run. Can you spot a specific part of the code which is taking long, or is it more or less equally distributed across the document?

@PhelypeOleinik It seems to get progressively (or regressively?) slower with each tweet.

@wilx Hm... Could be. Is anything added to some database or whatnot at each tweet? If so, some code which uses this database might be slowing you down.

@PhelypeOleinik No. I have converted the JSONs retrieved from Twitter's API into .tex files with custom tweet environment. Another .tex got generated with \input{tweet.123} for each separate file. This all gets once more \input{tweets}'d into main document which defines the environment.

@wilx Nothing seems guilty so far... Minimal example?

It is hard to create a minimal example of this. :)

But I can tar it up and put it somewhere.

@wilx Anything I can compile to investigate already helps

@PhelypeOleinik Here: drive.google.com/open?id=1oB296vLk-0DuPE_Etx-zyd-6xeqi-DE4 It contains a directory. The main document is tweet-document.tex. You should be able to compile it. I am using TeX Live 2019 updated just today. There is a GNUmakefile but it requires tweet2latex tool to convert the JSON files into the LaTeX files.

good afternoon ladies and gentlemen

'ello

uia

paulo

qto tempo

Olá. :)

one day i gonna learn the arara software

gg

To this very day I still haven't learned much, so no worries. :)

@wilx (sorry the delay, it took a while to download the fonts) Indeed, each file input makes the process slower. However I don't immediately see why the difference between the first file (which takes 0.3 s to run) and 70th (which takes 0.6 s). The number of lines of \tracingall (which roughly measures the amount of code run) is about the same. However, both do run quite a lot of code.

@wilx Ha! Remove microtype to see the difference.

@PhelypeOleinik Indeed, that's a world of difference.

@wilx Although your document is really unusual, I didn't know microtype could cause such a huge overhead... Really interesting. Try reducing the number of font changes. Most of the time seems to go on font selection...

a random tip: when using the datatool package, never use a command as a header in the csv file

Finally we'll get some nice weather: @PauloCereda

@yo' Yay!

brrrrrr

:)

I got a "good answer badge" for the Œuvres answer - waiting for a Great Answer gold badge now ;-)

I found this advice on the internet. Clearly bad advice not to be believed.

oh wait

@HaraldHanche-Olsen ooh

@PauloCereda already working towards 7777, I see

hellveticafont.com

@JosephWright I see no difference from the single-L version ;-)

@PhelypeOleinik :)

ConTeXt LMTX now has native SVG support. pragma-ade.nl/general/manuals/svg-lmtx.pdf

@PhelypeOleinik Do you think I should fill some sort issue ticket with microtype?

@wilx Hm... I don't believe this is a bug, actually. microtype is a huge package, and it hooks into LaTeX's font selection commands, so it's expected that a document with microtype loaded will take longer to compile.

@wilx Also, if I remove microtype from my dissertation, the difference is barely noticeable, so I suspect your document is doing something unusual with its fonts.

@wilx When I have some time I'll try to track down what makes it take so long to compile, and why the time increases with each file. I'll get back to you if I find something out.

@PhelypeOleinik This is 1 minutes vs 20 minutes. That's not just slower but orders of magnitudes difference.

@PhelypeOleinik Oh. OK. Thank you.

@wilx Exactly. But I've seen this only in your document, which is odd. If course this can be a bug. I didn't dig enough to be sure. Sending an issue to the maintainer won't hurt, of course :-)

@wilx I've seen that you seem to use \newfontfamily\emojifont{...} in your begin tweet code. Could you try if your document builds faster if you move this into your preamble? Loading the font every time might get quite slow.

@MarcelKrüger I tried moving all those definitions to the preamble. Didn't change much. Although it's certainly better to do so (@wilx).

@MarcelKrüger I have tried it right now, it makes no difference. But it is a good idea regardless.

==============================================================
LuaTeX 1.11.1 2019-10-28
==============================================================

First release of luahbtex / luajithbtex,
luatex / luajittex with harfbuzz.
Small bug fixes, code clean up and a couple of new primitives
to match eTeX.
A new callback lua.getcodepage()  for diagnostic purposes.
*** DROPPED the   "page_objnum_provider" callback, replaced with
"page_order_index". ***
See ChangeLog and the manual.

@JosephWright where is this from?

@UlrikeFischer LuaTeX list

@JosephWright hm, I don't see, I wonder if gmane is letting me down again.

@UlrikeFischer Also went to the ConTeXt list ...

@JosephWright yes, I found it. Luigi is working one the non-ascii "we have now a getcodepage callback :" (but I have no idea what it is meant to do).

@UlrikeFischer Er, good

@UlrikeFischer it's on Twitter too!

@UlrikeFischer Got it

\eTeXglue[stretch|shrink]order @JosephWright

@DavidCarlisle is that in the new version? \hfi?

@DavidCarlisle Sure

@UlrikeFischer changelog for 2019 changes is

2019-10-28 Luigi Scarso <luigi.scarso@gmail.com>
    * New callback lua.getcodepage()  for diagnostic purposes.


2019-10-25 Luigi Scarso <luigi.scarso@gmail.com>
    * Also support {oeps.foo} filenames in openin/out


2019-10-22 Luigi Scarso <luigi.scarso@gmail.com>
    *  *** DROPPED  "page_objnum_provider" callback , replaced with "page_order_index" ***


2019-10-18 Luigi Scarso <luigi.scarso@gmail.com>
	* following tug.org/pipermail/tlbuild/2019q3/004553.html
          we have now define(__sparc)

@DavidCarlisle in expl3 we could let those to the standard names, move the others for out purposes to say \tex_luatexstretchorder:D

@DavidCarlisle Ooh, a bug fix for us :)

@JosephWright yes was going to suggest something along those lines, but you'd have to guard (again) by luatex version

@DavidCarlisle No: you use \ifdefined\etexstretchoder

@JosephWright well yes you need to test something though also I'm not sure what value etexgluestrechorder has if the glue has fi stretch

@DavidCarlisle True, but i'm sure we can find a way

I'm trying to remember that parameter in TikZ for a node to have an extent/size. I thought it was textwidth, but apparently not. Anyone?

@DavidCarlisle If I got it right, it seems to be -1: github.com/TeX-Live/luatex/blob/trunk/source/texk/web2c/…

@MarcelKrüger the newest context beta solves the problem with the tounicode of cmu. I think I should try to import this. Should I use the harfmode-dev branch?

@UlrikeFischer Please use dev for now.

@MarcelKrüger has the dev branch a build setup with a texmf already?

@UlrikeFischer No, but we could do the harnode-dev merge: The test setup is working (I didn't want to merge without working tests), most functionality is fine, test coverage is pretty good. (We are missing Graphite for now, but that will need more work and isn't really important enough to hold off the merge) The other thing is documentation, but we probably should track that separatly.

@MarcelKrüger fine with me. If you merge I can try tomorrow to import context.

@PhelypeOleinik ooh, what I suggested they do:-)

@DavidCarlisle ;-)

@DavidCarlisle @PhelypeOleinik have you guys some archive of quotation worthy chat messages around or something?

@Skillmon there is a very limited repertoire of topics here, so whatever is being discussed, the search box on this page will usually turn up a relevant previous comment

@DavidCarlisle find me something with the combination of noodles with mustard (but not this comment).

@Skillmon marmite will have to substitute for mustard

@DavidCarlisle see, your system is flawed!

@Skillmon No, I just have a more useful definition of equivalence

@Skillmon On a side note, yesterday I found out that you're a Vizzerdrix :-)

@PhelypeOleinik Yay, someone noticed! How did you learn about it?

@Skillmon Some months ago a friend of mine introduced me to Magic. Yesterday he came here and we played a few matches and he had one of these among his (hundreds of) cards

@Skillmon I instantly recognized the picture :-)

@PhelypeOleinik nice. It was my first favoured card. We always called it "Das rosa Kampfkarnickel" (The pink fight rabbit). It was in one of the decks my brother learned Magic with when I was 8 years old or something like that. And it's no card that is quite common nowadays, as it isn't particularly strong for modern cards and quite old, so young players often don't recognize it.

@Skillmon It has a nice background story :-) Indeed, the last edition seems to be relatively old. It's rather strong, but the mana cost is too high. But I heard that the mechanics changed a lot from the earlier days

@Skillmon I have a handful of cards too. I usually play “commander” with a dragon-themed deck. It's a really nice (and neverending) game :-)

@PhelypeOleinik nowadays this card might still have decent P/T, but it's just too slow and doesn't offer any ability. I wouldn't play it anymore. Back in the days it was one of the strongest creatures I knew though.

@PhelypeOleinik I don't play often anymore, unfortunately. Maybe this will change again in the future, but currently I don't have the money to get the decks I'd like and not many players around where I live.

@Skillmon Me neither. It's a once every two months thing. I'm really bad at it too. And indeed, keeping and updated deck is expensive :/

@Skillmon If we meet one day, we'll play :-)

@PhelypeOleinik sounds good, I'd love to.

Going to bed now. Good night everybody!

@Skillmon Good night!