9:46 AM
texlua ~/texmf/scripts/luaotfload/luaotfload-tool.lua --aliases | wc -l 6681 @MarcelKrüger ^^ seems to work:-) @DavidCarlisle ooh 10:04 AM @DavidCarlisle Now we just need someone to adjust fontspec. 23 hours ago, by David Carlisle @UlrikeFischer just needs Will to make it work:-) @MarcelKrüger incidetally Karl mentioned off list whether texfonts.map may be better than the aliases file, format is essetially the same. looking now I see texfonts.map was used in earlier times to access postscript internal font names: % Allow people to use PostScript font names. % Except these r* fonts have not been the best practice for years, and % are now only part of the cspsfonts package (strange). Let's not do this. % --karl, 16jul07. % rpagk AvantGarde-Book % rpagko AvantGarde-BookOblique % rpagd AvantGarde-Demi @JosephWright I'm playing around with "create-style" keys, and while it works for flat key systems, there are lots of problems as soon a selective key setting is wanted. I tried with subdividing and groups but nothing allows to execute only a part of a meta-style-key. @DavidCarlisle Wouldn't you need TFM files when loading fonts from there? @DavidCarlisle ooh @MarcelKrüger ooh @UlrikeFischer ooh @MarcelKrüger that's what I thought too, actually (given that he was saying it would be too much work) I guess he was suggesting extending the map to apply to opentype fonts as well, presumably this would potentially have a benefit of not impacting on searches of anythng other than fonts. (I'm not sure if havig an aliases file that big has any measurable impact on searching?) 10:17 AM @PauloCereda Good morning! How's the ooh counter going? @Rmano ooh we need to take a look :) @Rmano and good morning, mr. squirrel! <3 @Rmano he needs to get them in quick as it will soon be lunch time @DavidCarlisle oh no @DavidCarlisle Yes, you're right --- I have to start to prepare @Rmano OH NO 10:19 AM @PauloCereda almost vegetarian today, don't worry. Just a couple of anchovies but the rest is vegetable ;-) @Rmano ooh anchovies I need a pizza. :) @PauloCereda ^^^^^ @DavidCarlisle I haven't tested it, but I wouldn't expect it to make a big difference. (I would assume that it's basically a hash table which should have more or less O(1) lookups.) Using the map file seems quite messy, especially since it can already contain otf and ttf files with legacy encodings. @Rmano ooh @MarcelKrüger yes seems that way, the texfonts map seems only set up for tfm files I guess it just came to mind for Karl as he would have done this for the similar sounding issue of using PostScript names back in the beginning 10:47 AM @UlrikeFischer I guess I'll need a demo @JosephWright quack @PauloCereda QUACK @JosephWright ooh a duck 5 German is so easy @PauloCereda :) @JosephWright for example this. I want to avoid that the B-key is executed. But if it is burried in the meta, then it seems to inherit the group from the C key: \documentclass{article} \begin{document} \ExplSyntaxOn \keys_define:nn {path}{A .code:n = A#1,A .groups:n = GA} \keys_define:nn {path}{B .code:n = B#1,B .groups:n = GB} \keys_define:nn {path}{C .meta:n = {A=#1,B=#1}, C .groups:n={GA}} \keys_set_groups:nnn {path}{GA}{A=a,B=b,C=c} \ExplSyntaxOff \end{document} I want only "AaAc". 2 hours later… 12:36 PM @PhelypeOleinik I've tried to skim over the alignment code in texdoc tex, but honestly, I have way too little of an idea of the TeX code base to understand where I should look for the issue with the master counter state being ignored when an unexpandable token is found during align_peek. @PhelypeOleinik but the curious behaviour we have as an issue is documented in the TeXbook in a footnote: \footnote*{The token list\alphashould not be empty, however, because \TeX\ expands the first token of an alignment entry before looking at the template, in order to see if the entry begins with |\noalign| or |\omit|. The master counter value that is considered to be present at the beginning of an entry is the value in the counter just after the u\$~part'' of the
template has been entirely read.}

@UlrikeFischer Ah, yes, that might happen ....

@Skillmon can you remind me of the example where you want different behaviour?

12:53 PM
@DavidCarlisle if \keyval_parse:nnn is used at the start of a column but the user code isn't fully expandable, a single \group_align_safe_begin: at the start of \keyval_parse:nnn doesn't suffice to make \keyval_parse:nnn alignment safe. See github.com/latex3/latex3/issues/896#issuecomment-830864624 and the code in fix #1.

@Skillmon but is that really associated with the \omit look-ahead? I get the same error from xxx\foo{a=a&b} \\ % doesn't work so it's related to being nested in halign but not specifically being at start of cell?

@DavidCarlisle well it is. If you apply fix #1 and put zz\foo{a=a&b} in the table it'll work.
@DavidCarlisle currently \keyval_parse:nnn doesn't try anything to be alignment safe (as it never was alignment safe in the first place...)

@Skillmon Oh, who would have imagined the documentation contains the answer ;-)

@Skillmon yes the align_safe groups are needed for the nested align case, but I thought you were saying it was only failing because of the lookahead

1:10 PM
@DavidCarlisle well, yesn't. If TeX respected the state of the master counter during align_peek for the alignment group a single align_safe group would suffice to make \keyval_parse:nnn alignment safe. Since that's not the case, we have to put a group per parsed key/keyval pair, or use one of the other strategies.
@PhelypeOleinik :) That's what we get for reading the documentation...
Off for a while, until later!

@Skillmon yes true, I was just commenting on the initial example really

2 hours later…
3:33 PM
@Skillmon is it clear why all these work, why isnt the last one the "empty alpha" case?
\documentclass[]{minimal}

\ExplSyntaxOn
\NewDocumentCommand \foo { m } {\iffalse{\fi\keys_set:nn { foo } {#1} \iffalse}\fi }
\keys_define:nn { foo } { a .tl_set:N = \l_tmpa_tl }
\ExplSyntaxOff

\begin{document}
\begin{tabular}{c}
\foo{a={a&b}} \\ % works
x\foo{a=a&b} \\ % doesn't work
\foo{a=a&b} \\ % doesn't work
\end{tabular}
\end{document}

@DavidCarlisle because \foo is a robust command, so TeX isn't searching for \omit etc. anymore.
@DavidCarlisle (I guess, not sure how the rules in this area are)

@Skillmon yes was just aout to answe that myself... hmm but in a latex/expl3 context do you care about non protected commands
@Skillmon yes that was the main change we got at etex 2 over etex1 that \protected tokens stopped the lookahead

@DavidCarlisle yes, because the real bug is that \keyval_parse:nnn isn't alignment safe, it is just a consequence of this that currently, also \keys_set:nn isn't alignment safe. Making the latter alignment safe is trivial, making the former alignment safe takes a bit more effort.

@Skillmon which probably means that you can guard the definition with \protectedempty defined as \protected\def\protectedempty{} can't you?
@Skillmon oh we were a few days ago: you can't do that if you want to expand finally to \omit grrr

@DavidCarlisle but is this the correct thing to do in a macro that is documented as expandable?
@DavidCarlisle correct.

3:41 PM
@Skillmon making the documentation tell the truth can be done later it isn't necessarily an unbreakable contract th yatou have to implement exactly what was originally documented.

@DavidCarlisle I guess the real discussion now is which route to choose for the fix (and maybe there is a bit of benchmarking ahead of me to make an educated decission)
@DavidCarlisle we could also document \keyval_parse:nnn as not being alignment safe, and instead only make \keys_set:nn alignment safe.

@Skillmon or .. oh you just said the same I was writing...

@DavidCarlisle I have yet to benchmark how big of an impact it would be to make \keyval_parse:nnn alignment safe. And I guess the only two really safe variants of this are either the \expanded or the delayed output route. I don't trust the user code with the variant which closes and reopens the align-group around it everytime.

@Skillmon I think this is not unreasonable, parsising is almost always embedded in some other code and if that code is being used in alignments it has probably already guarded itself, think of a tabular variant with a keyval option that tabular code can guard what needs guarding before handling its arguments
@Skillmon as in the code I posted above, that is perfectly safe and in this context (as in most contexts) the trailing \iffalse}\fi isn't doing any harm, so any contortions that need to be done to remove it in advance are not needed.

@DavidCarlisle I'd like Frank's, Bruno's and Joseph's input on this as well... I don't feel comfortable making decisions in the name of the team. I volunteer to implement whatever route you guys choose though :)

3:49 PM
@Skillmon I'll comment in the issue

4:34 PM
I need some tutorials on how to write scripts to change fonts.

5:24 PM
@Rmano which fonts are you hoping to avoid?, you could use egreg's user stylesheet to get the site in comic sans.

@DavidCarlisle I am not so fond of Ubuntu, especially the mono font

@Rmano switch OS to windows:-)

@DavidCarlisle a bit radical solution ;-)

@Rmano in firefox I already use a user stylesheet that ensures I don't get the left sidebar, makes the right sidebar smaller and omits some of its more annoying boxes, so if I don't like the font I could easily change it. But I just tried switching manually to segoe UI and I think it's OK

@DavidCarlisle interesting. The left bar is effectively quite annoying...

5:34 PM
@Rmano there is an option in the site settings to not have the left bar but I go further:

@DavidCarlisle yes I've seen the option but then I have no "newest" in the top button bar bar (which you removed, I see ;-)

@Rmano the "newest" link is on the list of questions at tex.stackexchange.com/questions it isn't on the individual question pages is it? oh was it on the left sidebar, it's so long since I had that enabled I don't remember:-) You can get back to the newest questions list from the three bar menu at top left

@DavidCarlisle Aaahhhhh thanks!

6:09 PM
christianity.stackexchange.com uses serif fonts and it is hard to read. :-)

6:19 PM
Hey folks! :) What do you reckon is a fitting math font for the source serif pro font?

6:30 PM
@KimJongUn I like serif when the size is not small --- otherwise, the font seems irregular. I agree that good serif fonts are more difficult to find than sans.

6:51 PM
@Rmano: :-)

2 hours later…
8:26 PM
hey all, been a while
is there a ubuntu PPA for TeXlive 2020? i'm on 20.04 and don't wanna hassle with the expired key thingy
20.04 comes with TL 2019

@DavidCarlisle Hi, I have an idea/question (I don't know if it is feasible) for texfaq, in the past it was distributed as a pdf file and I was looking at (github.com/Wandmalfarbe/pandoc-latex-template), do you think something like that can be applied to texfaq?

looking at it now

8:47 PM
@PabloGonzálezL well it's all prettty smple markdown, maybe pandoc can make something of it I don't know, never used it really. the old version was written in tex so pdf was its natural format with any web version sbeing conversions from that.
@AGoldMan why not just install upstream texlive 2021 from tug?

yeah, that's what I'm doing
was hoping to be more lazy and not have to set it up
but i guess I could even do a user install, no sudo needed

@DavidCarlisle Haaa...I thought it was the other way around, from html to PDF, but since everything is now in .md it made more sense to go from .md to .tex :)

@PabloGonzálezL I think the uktexfaq is older than html

@DavidCarlisle hoooo :D
@DavidCarlisle Another question, htlatex or make4ht ...which is the correct executable? (I know I read somewhere in the forum that htlatex is deprecated).

9:06 PM
@PabloGonzálezL make4ht

@DavidCarlisle Thanks, I was just reading (tug.org/tex4ht/doc/mn-commands.html), but there is no mention of make4ht :(
@DavidCarlisle perhaps the line should be updated (github.com/latex3/hyperref/blob/…)

9:28 PM
@PabloGonzálezL make4ht is a more or less separately developed wrapper around tex4ht, michal has a manual for it somewhere, let me find....
@PabloGonzálezL why, it works? hyperref build is complicated enough without making chages with no claimed benefit. make4ht uses the same tex4ht conversion as htlatex so what benefit would it have apart from possibly missing some argument syntax change and breaking it?

@DavidCarlisle Great thanks :D ...I hadn't found that link (I searched for a while and nothing).
@DavidCarlisle I had misread the meaning of "depreciation" as referring to htlatex:(
@DavidCarlisle There is really no benefit (in this case) to modifying the line (it was a headache to get everything to work in hyperref) :p
@DavidCarlisle (kodymirus.cz) ?? is from @michal.h21

@PabloGonzálezL no idea about the rest of the site, it's a bit "sparse" :-)

9:44 PM
@DavidCarlisle Jejeje :D

@PabloGonzálezL I know. github pages is refusing to auto-redirect, just go straight to the https version.

@DavidCarlisle :( ... from my mobile it tells me that it is not safe and does not let me enter (android and its things).

@PabloGonzálezL how did you get the url originally? click on this texfaq.org
@PabloGonzálezL you could ask @JosephWright to look at this again:-)
@PabloGonzálezL Beware: a man in the middle may steal Robin's 30 year old latex hack

@DavidCarlisle Great, no problems from there :D
@DavidCarlisle hooooo

@PabloGonzálezL yoy get warned about this every time you make a commit to the site, so it's not likely to get forgotten:-)

9:52 PM
@DavidCarlisle Promoting names with spaces texfaq.org/FAQ-grffilenames

@PabloGonzálezL almost all of the faq text predates 2010, so almost every page could do with an update.... the advice the grffile package will help isn't exactly harmful, admitedly it could say the grffile package is completly empty and does nothing at all but still....

@DavidCarlisle In short, it's not your fault :D
@DavidCarlisle I have been looking at the changes and in this last week they have updated several things :D
@DavidCarlisle And I just found a broken link to (texdev.net/index.php?s=programming+latex3) @JosephWright ^^

@PabloGonzálezL yes we had a pull request updating the conversion page.

@DavidCarlisle Now I found it, just today I went to my parents' house, I knew it was somewhere and I finally got it, my printed copy of texfaq from my college years :D

10:02 PM
@DavidCarlisle Yes !!

@PabloGonzálezL fixed

@DavidCarlisle Heheh

@PabloGonzálezL (@JosephWright) I'm not sure what to do with that page: the whole page is basically wrong now, we could delete it or collapse it to a line that says "spaces in file names used not to work, but now they do."
and this is more or less the same re multiple dots. texfaq.org/FAQ-unkgrfextn

@DavidCarlisle I think "removing it" is the best option (assuming the vast majority will have a modern version of TeX installed).

@PabloGonzálezL but as you just observed removing links breaks references to the faq (and theer are millions of those)

10:09 PM
@DavidCarlisle Or you can put a "404" message and blame the usual person :D
@DavidCarlisle And some of your magic with greep + regex?

@PabloGonzálezL we don't control the server so can't have a custom 404 page, and redirects mean basically having a page there anyway as we can not do server side redirects

@DavidCarlisle :( ... Mmm... what do you think of this one (texfaq.org/FAQ-acrobat)

@PabloGonzálezL everyone should use DVIPSONE
27 mins ago, by David Carlisle
@PabloGonzálezL almost all of the faq text predates 2010, so almost every page could do with an update.... the advice the grffile package will help isn't exactly harmful, admitedly it could say the grffile package is completly empty and does nothing at all but still....

Jejjeje