12:50 AM
Happy Easter!
4

6 hours later…
6:48 AM
@UlrikeFischer Sure
@UlrikeFischer Yes, exactly

2 hours later…
9:11 AM
CHOCOLATE

9:51 AM
@JosephWright \driver_pdf_version_minor: is expandable but \driver_pdf_version_major: isn't. \PackageInfo{test}{pdf-Version: \driver_pdf_version_major:.\driver_pdf_version_minor:} gives Package test Info: pdf-Version: \driver_pdf_version_major: .5 That's imho a bit odd.

@UlrikeFischer Ooops, a typo

@JosephWright better now ;-). (and I should add git pull to my pop-up).

1 hour later…
11:12 AM
Apparently I’m incapable of writing fast Lua code, I wonder if I should just suck it and write everything in C.

@KhaledHosny oh is your code that awful? I am sure it's quite fast.

BTW, luahbtex is back again and now called harftex github.com/khaledhosny/harftex
3
@PauloCereda Processing the same file is 2.5 times slower than both luaotfload and xetex, so neither HarfBuzz nor Lua is slow, which leaves only my code.

@KhaledHosny oh

@KhaledHosny ooh good to know:-)

@DavidCarlisle The typesetting. There is the profiler in lualibs (called debugger, as camouflage I think), and the majority of time is spent in my Lua code.
Funnily, calling HarfBuzz takes only a small fraction of the time.

2 hours later…
1:06 PM
@marmot In mathematics, with very few exceptions (I can't think of any at the moment, though I think I've seen some), authors are listed alphabetically without regard to seniority or who did the most work.

1:24 PM
@HaraldHanche-Olsen Does that mean in Italy they have a strange alphabet in which C comes after F? ;-)

2:20 PM
@marmot In Italy, everything is possible.
Sri Lanka, we love you!
5

3:00 PM
@marmot Ubi maior…

@egreg Wer ko der ko?

3:25 PM
@DavidCarlisle congrats to 500k, @UlrikeFischer congrats to 200k @JouleV congrats to 14541 :D
happy easter to everyone

@thymaro :-)

3:41 PM
@JosephWright I think the definition of \driver_pdf_link_begin_goto:nnw is wrong. \documentclass[a4paper]{article}
\usepackage{expl3}
\ExplSyntaxOn
\driver_pdf_compresslevel:n {0}
\driver_pdf_objects_disable:
\ExplSyntaxOff
\begin{document}
\ExplSyntaxOn
xx
\ExplSyntaxOff
\end{document}
this gives an action of /A << /S /GoTo /D (/Subtype/LinkDest) >> instead of /A << /S /GoTo /D (Dest) >>

@UlrikeFischer shame about the ducks:-) (thanks for the accept though!)

@DavidCarlisle sorry ;-). but they are round plump ducks ;-)

@UlrikeFischer just right for easter sunday

@thymaro I commented on your comment to my meta answer, thinking you were the OP. I've deleted the comment now (since you're not the OP), but in case you saw it and were confused, that's what happened.

@JosephWright and why do the driver commands end with w? \driver_pdf_link_begin_goto:nnw

4:02 PM
@UlrikeFischer Could well be: I've not tested these ones (also, what subtype is right ...)
@UlrikeFischer They have to be matched because for some drivers (dvips currently) they have to collect a box. So we have the same as \hbox_set:Nw - the command does or may collect an argument in a w sense, not simply require a matching ...end:
@UlrikeFischer It's largely a reminder that for dvips we are actually collecting something, not just relying on the back-end to draw the box correctly

@JosephWright ah, ok, it looked a bit weird, but I see now the analogy to the hbox commands.
@JosephWright imho the /Subtype/Link should go out of the internal command.
\cs_new_protected:Npn \driver_pdf_link_begin_user:nnw #1#2
{ \__driver_pdf_link_begin:nnnw {#1} { user } {/Subtype /Link ~ #2} }

@UlrikeFischer But they are all links ... so don't they all need /Subtype /Link?

@JosephWright yes, but the goto name type adds it automatically. Try \pdfstartlink goto name {Dest} xxx \pdfendlink:
/Type /Annot
/Rect [149.266 671.357 167.033 677.638]
/A << /S /GoTo /D (Dest) >>

@UlrikeFischer Ah, right

　＜⌒／ヽ-､_＿_
／＜_/＿＿＿＿／
￣￣￣￣￣￣￣

∧_∧     I SHOULD TRY
＿|　⊃／(＿＿_
／　└-(＿＿＿_／

zzz
＜⌒／ヽ-､_＿_
／＜_/＿＿＿＿／
￣￣￣￣￣￣￣
3

4:23 PM
@PauloCereda \pdfstartlink goto name {Bed} xxx \pdfendlink: is then a more sensible destination name :-)

@UlrikeFischer ooh :)

@PauloCereda ah, that's where the \zzz command by @DavidCarlisle's comes from!

@CarLaTeX ooh

@PauloCereda haven't you got important psmith PR's to handle? :-)

@DavidCarlisle ooh that's true :)

4:37 PM
@JosephWright regarding the naming. Imho \g__driver_pdf_resource_extgs_prop sounds good (one could use ...pdf_xobj_resources... for objects resources). Should we call tl that holds the objectnumber/name \g__driver_pdf_resource_extgs_tl? Or should there be somewhere a reference that it is an object?

@UlrikeFischer Do we need to provide access to this number/ref? I was imagining it's internal (so is the prop, but we have a special case due to the need to access from the 2e kernel). (If it's internal, we can use an int)

@JosephWright well the name is internal (two underscores), but can it be an int? I thought dvips/dvipdfmx would use names like "@extobj" in the specials.

@UlrikeFischer The name is like that ... but for our tracking the unique part is a number. In any case, if it's not public then the form can be engine-dependent

@JosephWright thinking about it, it is not a \g__ but a \c__ variable: the object number doesn't change, when we have reserved it. But if a tl or an int is more suited is your choice, you have to do the cross-driver stuff.

5:02 PM
@UlrikeFischer I'm going to send a CTAN update today: probably we will need the driver code available as-is for further testing, plus sorting the resource issue is non-trivial

@JosephWright did you correct the goto code? I see no commit.

@UlrikeFischer Yes, local at the present
@UlrikeFischer Try now

@JosephWright I think there is ~ missing after /Link

@UlrikeFischer The next entry should be a key, so there should not need to be a ~ ...

@JosephWright ah, yes you are right, I had problem with the Dest, but this is not relevant here.
@JosephWright are the other drivers ok or should I check them?

5:09 PM
@UlrikeFischer Totally unchecked

@JosephWright then I will take look. Should go fast.

5:26 PM
@JosephWright with dvips the user link work but the goto gives a postscript error. with xelatex for both I can't find no annotation in the pdf, only the "dummy" is there <</Ff 65537/FT/Btn/Subtype/Widget/Rect[130.062 651.336 133.051 654.325]>>. I have to make dinner now, but can look at it later.

1 hour later…
6:30 PM
@UlrikeFischer On it

@JosephWright the link problem with xelatex was only because I had no content (I always forget that xelatex needs something there.

@UlrikeFischer I think the goto one is bust ... workingon it

@JosephWright ok. Then I leave it to you - not necessary that we both do it (but this means that I will have to do my ironing, sigh).

@UlrikeFischer Fixed I hope

6:54 PM
@JosephWright I get with dvips an undefined error:
! Undefined control sequence.
<argument> \__driver_pdf_link_begin_aux:n
should be \__driver_pdf_link_begin_aux:nw I guess.

@UlrikeFischer Fixed

7:13 PM
@JosephWright looks ok.

@UlrikeFischer Good
@UlrikeFischer Building for CTAN

@JosephWright ooh

@PauloCereda It's all developer stuff

@JosephWright :)

7:30 PM
@AlanMunn no problem. Didn't see it, I'm on vacation. What did you write, though?

7:58 PM

@thymaro Just that my comments weren't aimed at the OP (which I thought you were.)

2 hours later…
9:39 PM
Please have a look to my answer for an moderncv question and the printing of an image. The upvoted answer (not mine) is wrong, because moderncv needs to add the image inside the header of the cv, and not via \includegraphics in the cv code. My answer reflects that but was never upvoted. Is it worth to get an upvote?
0

It seems (you gave us no compilable code to test!) you are using a style of moderncv that does not include an image to the cv. To see all posibilities moderncv offers, please see my answer to this question. Style banking does not include an image. If you do not want to use one of the styles in...

@DavidCarlisle today I encountered a jpeg/JFIF file which couldn't be included by graphics without additionally loading the pict2e package. Who is to blame?

@Kurt to be honest that looks more of a comment than an answer, since it just references previous answers
@Skillmon blame who you like, so long as you don't blame me.
@Skillmon can't happen actually:-) (which doesn't mean that it didn't happen)

@DavidCarlisle I can assure you, it happened. \documentclass{article}\usepackage{graphics}\begin{document}\includegraphics{foo}\end{document} failed with no BoundingBox error, additionally using pict2e worked.

@Skillmon never believe the OP:-)

@DavidCarlisle I tried myself, so I shouldn't believe myself?

9:44 PM
@Skillmon pdflatex or latex/dvips ?

@DavidCarlisle pdf/lua/xelatex.
@DavidCarlisle didn't try dvips
@DavidCarlisle ok, now I'm confused. the minimal example above doesn't throw the error. But the actual document does (no BoundingBox...). Will investigate further.

@Skillmon there was an issue a year or so ago where jpg had inconsistent exif metadata which meant that different parts of the toolchain took different size information and things blew up...
@Skillmon check which .def file is being loaded in each case

@DavidCarlisle Yes, now you can compare the outcomes of TikZ codes to non-TikZ ones. ;-)

@DavidCarlisle I tried compiling on my local TL2018 frozen, which threw the error. Currently I try on Overleaf with TL2017.

@Skillmon send me a zip if you want (no promises but I might look)

9:52 PM
@Skillmon how is the real name of your graphic?

@UlrikeFischer don't tell me it has spaces:-)

@UlrikeFischer nothing special, no spaces, no non-ASCII characters.
@DavidCarlisle found the issue
@DavidCarlisle it is unbelievable how some classes come into existence.

@Skillmon forcing dvips back end?

@UlrikeFischer I've sent to CTAN

@DavidCarlisle yep.

9:55 PM
@Skillmon I thought so:-)

@DavidCarlisle Last touched on 1995-11-21.

@JosephWright I saw it on the team list.

@UlrikeFischer :)

@Skillmon not so old then

@DavidCarlisle having this very interesting comment:

9:56 PM
@UlrikeFischer I'm looking forward to thoughts of the senior people on the atbegshi question (@DavidCarlisle ...)

%%% This is not a clean and optimal LaTeX2e class port yet. I am working
%%% on that but this class file has been used for a PhD thesis that did
%%% pass a preliminary inspection by the graduate office.
%%%
%%% I would appreciate any help/comments and improvements. The goal is
%%% to make this an official Rice thesis style for LaTeX2e that can
%%% exist in parallel with the old LaTeX thesis style.
@DavidCarlisle ^^^^

@Skillmon Definitely dvips then

@JosephWright yes, just found the \RequirePackage[dvips]{graphicx} line in the class.

@Skillmon Bit of a clue :)

@Skillmon but above you use article. Did you have a local, changed version?

9:59 PM
13 mins ago, by Skillmon
@DavidCarlisle didn't try dvips
@Skillmon so user error: you used everything except the one thing it works for:-)
@UlrikeFischer never believe the OP

@UlrikeFischer that was my bad. I had played a lot with it and thought I also tried with article.

@JosephWright I'm quite for a copy of atbegshi in the kernel. I need it too.

@DavidCarlisle even worse: Support error, I'm not the user in this case.

@DavidCarlisle I didn't believe him, that's why I asked for the real name ;-)

@UlrikeFischer Yeah, I'm just thinking if we need all of it

10:01 PM
@UlrikeFischer @DavidCarlisle nobody believes me :(

@DavidCarlisle wouldn't jpg fail with dvips too?

@Skillmon doden't matter, main thing is IT'S NOT MY FAULT
6

@UlrikeFischer I'll still need a L3 version for generic mode, or we say 'load atbegshi if you want generic PDF resource support'

@DavidCarlisle of course not!

@UlrikeFischer yes:-)

10:03 PM
@UlrikeFischer Mainly down to 'will FMi got for it': @DavidCarlisle (or me) can make the original atbegshi a no-op for newer kernels

@JosephWright I think we should do something, but as we were discussing at the start of the year, that kind of hook should be built in not package based.

@DavidCarlisle Yes
@DavidCarlisle Exactly: if Frank goes for it, like @UlrikeFischer I think we just pick up (almost) all of the package in the format

@JosephWright that and disentangling from the rest of Heiko's universe infwarerr, ltxcmds, ...

@DavidCarlisle We (I) can't do PDF resources without it
@DavidCarlisle That's easy, really

@JosephWright yes, but it means a few decisions about how much you exactly mimic current package behaviour

10:09 PM
Good night, everybody! (sorry for posting this)

@JosephWright then there are all the patches for (eg) memoir, can they be pushed upstream

@DavidCarlisle Do it in the format, don't have to mess around as we know we can access \shipout properly
@DavidCarlisle For me, it's only about the 'extra' functions for graphics: I want (need) \AtBeginShipout, \AtNextShipout, \AtFirstShipout (admittedly, latter is more-or-less \AtBeginDvi)

@JosephWright ah OK must admit I didn't look exactly what the memoir block was doing. Adding the hooks to shipout seems basically a good plan

@DavidCarlisle I have a feeling we will end up integrating a lot of heiko's stuff, either into the kernel or into expl3 (on the latter, of course, we've already grabbed quite a bit, and the hyperref stuff really will hover up a massive amount)
@UlrikeFischer I wonder how long before someone (@AlexG?) finds some bugs :)

@JosephWright ;-). I will have to start to write test files. And there is loads of stuff in hyperref that in the end will have to move.

10:17 PM
@UlrikeFischer Sure
@UlrikeFischer Going to \RequirePackage{expl3} in hyperref? (May as well do, really)

@JosephWright I need \AtBeginShipout and the \AtBeginShipoutBox.

@UlrikeFischer I think this is going to be what takes us to 'LaTeX2e*' (first asked about on team list several years ago)
@UlrikeFischer At, the box one: yes, also needed

@JosephWright well my custom driver has it. And if we change something in hyperref itself I wouldn't have any qualms to load it.

@DavidCarlisle better now?

@UlrikeFischer We just should: I'm seriously thinking about finishing l3image and saying 'lets just take over graphicx too'

10:23 PM
How did you know about the atbegshi question more than 20 minutes before it appeared?

@marmot :)

@JosephWright sure. Make a latex-expl and we try it out ;-).

@UlrikeFischer Going to need a new l3build feature: custom format building as a general thing

@JosephWright for a start running fmtutil-sys with a local config will work too.

@Kurt yep:-)

10:31 PM
@UlrikeFischer Just a question really of taking the format-building code and making it into a general target, though we will end up needing a decent number of files ...

@DavidCarlisle Thanks!

10:54 PM
@JosephWright are you thinking about the test suite? But if the file are in their own branch wouldn't it work out-of-the-box?