@AlanMunn -- (A U.S. speaker ...) My intuition says that a comma isn't wrong there, but isn't really needed.

@barbarabeeton yes I agree

ooh we Brazilians love commas

The more commas the better

@PhelypeOleinik ^^

@DavidCarlisle twitter.com/AlexKontorovich/status/1236008125645434885?s=19

@CarLaTeX ooh pizza

@CarLaTeX also, today is the traditional Nintendo's Mar10 day. :)

@PauloCereda I'm not a videogame expert :)

@CarLaTeX :)

yay got X applications working from wsl:-)

@DavidCarlisle Yay

@DavidCarlisle yay?

@PauloCereda I couldn't work out why just setting DISPLAY=:0.0 wasn't enough, turns out cgwin X disables tcp connections by default (so doesn't allow non cygwin connections from same machine) but restart X server with -listen tcp and it listens to tcp...

@PauloCereda Eu, Phelype, discordo, dessa afirmação, Sr. Cereda

@PhelypeOleinik ooh :)

@DavidCarlisle oh in Wayland you have to change something similar, but I cannot remember what...

@PauloCereda see it works ^

@JosephWright I think there is something wrong with the dvips backend definition of annotation:

\documentclass{article}

\begin{document}
ccc
\ExplSyntaxOn
\__pdf_backend_annotation:nnnn {1cm}{1cm}{0cm}{/Subtype/Text}
\ExplSyntaxOff

\end{document}

^^^ This prints the content on the page.

@UlrikeFischer OK,. I'll take a look I must have messed up some argument

@UlrikeFischer How are we doing with the PDF stuff?

@JosephWright I'm just trying to setup a dvips-test route with the specialformats table and the first result was that I find this bug ;-)

@JosephWright what do you mean?

@UlrikeFischer Ah, right, yes

@UlrikeFischer As in, are you making progress?

@JosephWright yes, I'm just going through the whole file to check for the missing stuff. Imho apart from xform for dvips most should be okay, but some stuff needs more testing (I'm e.g. not really sure if the calculation order code is doing the right thing).

@UlrikeFischer Right, yes

@UlrikeFischer I'll do this at lunch here

@DavidCarlisle ooh

@JosephWright I'm not getting it. I'm trying to setup a latex-dvips-ps2pdf test and to get rid of the UUid in the pdf. I have simple batch file which does latex, dvips and then ps2pdf -sDocumentUUID=DocumentUUID -sInstanceUUID=InstanceUUID -dCompressStreams=false -dCompressPages=false -dNOSAFER %4. When I use it directly on the test file it works as expected, but as soon as I use l3build the UUID settings are lost.

@UlrikeFischer And you are passing all that using specialformats?

@JosephWright yes, I have an entry dvipsbat = {binary="l3builddvips",format = ""}. It seems to do the right thing - the compilation is done and I get the right output files.

@JosephWright I have to go now, will think about it later again.

@UlrikeFischer Ah, clever

@UlrikeFischer We may need new features; to-date, we've assumed PDFs from direct output only

@DavidCarlisle -- Maybe one source of your problems?

@barbarabeeton @PauloCereda posted same text yesterday:-)

@DavidCarlisle -- Oh. We're just on different schedules, I guess.

@DavidCarlisle mine was sent with UDP, so there was no ack. :)

@PauloCereda -- Yesterday, the temperature here went up to 20C. It would have been almost warm enough for you. Today, though, it's heading back down toward freezing, with rain.

@barbarabeeton ooh :)

I think this must be one of my more useful documents \font\@\@A\end

@DavidCarlisle -- Looks like tex78. ???

@barbarabeeton works with texlive2020 tex, so long as you copy cmr10.tfm (or anything else) to .tfm it sets A in the font with name ""

@JosephWright I found out why the ps2pdf options were lost: l3build is calling dvitopdf(name, testdir, engine, hide) and that was overwriting my carefully produced pdf. I solved it by setting dviext="xxx" in the config file. I can set CreationDate and ModDate in the document. Now I'm down to this diff:

***** .\BUILD\TEST-CONFIG-DVIPS\catalogAF.tpf
  157:  << /Size 17 /Root 1 0 R /Info 2 0 R
  158:  /ID [<E19F1BD619BE06CEDA5BCDB8AE6BE00F><E19F1BD619BE06CEDA5BCDB8AE6BE00F>]
  159:  >>
***** .\BUILD\TEST-CONFIG-DVIPS\CATALOGAF.DVIPSBAT.PDF
  157:  << /Size 17 /Root 1 0 R /Info 2 0 R
  158:  /ID [<1B8774D59D949FB1EDFD14CC31002619><1B8774D59D949FB1EDFD14CC31002619>]
  159:  >>
*****

I guess that is solvable too. l3build should be able to normalize this.

@UlrikeFischer Do we need a way to customise dvitopdf?

@UlrikeFischer Yes, just have to agree on what we will zap

@UlrikeFischer you could probably use a lua file rather than a batch/shell script for the dvips thing then you could check in a call to it that would work cross platform?

@JosephWright it would probably remove the need to use a special batch file. Basically the code already runs dvips + ps2pdf, one only need to be able to pass the options -sDocumentUUID=DocumentUUID -sInstanceUUID=InstanceUUID -dCompressStreams=false -dCompressPages=false --dNOSAFER.

@UlrikeFischer OK, so something as simple as ps2pdfopts?

@UlrikeFischer OK, I'll sort it over the coming days

@DavidCarlisle yes that was the plan (but is probably unneeded as the compilation is already built in l3build), I started with a batch file as I know there where to put the "echo" and similar stuff to debug it.

@JosephWright regression-test should probably also have something to set the CreationDate and the ModDate. I have the impression the epoch stuff doesn't work with dvips.

@UlrikeFischer Quite possibly not

@UlrikeFischer seems like debian patch gs to meet their reproducible build requirements (which is where the tex support for epoch settings came from) wiki.debian.org/ReproducibleBuilds/PdfGeneratedByGhostscript so not much use for us :(

@DavidCarlisle I found a rather long discussion about it, and it sounded (but I didn't read everything) that the gs developer weren't keen on making reproducible pdf. bugs.ghostscript.com/show_bug.cgi?id=696765

@UlrikeFischer yes saw that as well. But presumably we can normalise that out in the l3build diff so it doesn't cause any actual issues, you just can't test that it is reproducible

@PhelypeOleinik You'll be pleased to know I've been cajoled into another LaTeX3 foray... tex.stackexchange.com/q/531987/2693

@AlanMunn It creeps in, until there's_no_turning_back:w

@AlanMunn -- Not addressed is, why hasn't anyone "improved" TeX to add another pair of delimiters to be checked? TeX is based on 4-bit bytes, which allow only 16 values (0-15), and all 16 have already been assigned to distinct category codes, so any change along this line would be not only a major undertaking, but involve a complete redesign of the structure. I think this addresses the second part of the question, which you avoided.

@AlanMunn -- Totally off topic ... I've been reading The Dictionary Wars, about Noah Webster et al. and American dictionaries. It's reminding me of the story "Meihem in ce Klasrum" (which I first read a zillion years ago when I was in junior high school). I didn't realize then how closely that mirrored history!

@barbarabeeton that's true but the same could be said of \mathchardef, the bit packing is full but extended texs provide Umathchardef with an extended syntax, the same could be done (in principle) for catcodes beyond 16. In practice most documents get on fine with just F as catcode 1 and P as catcode 2.

@DavidCarlisle -- Somehow I think that Peter Flynn would have a problem ... (But point taken. Still, if nothing else, I would expect that error messages would get more confusing.)

@barbarabeeton Peter I am sure would be happy with the latin convention with " having catcode 1 and Y having catcode 2

@JosephWright I'm not sure about the implementation of \__pdf_backend_destination_rectangle:nn. Should this really typeset a box? it doesn't fit the pdftex primititve: