« first day (1709 days earlier)   

3:44 AM
@cfr I don't think the edit was warranted, but I would mark it as awkward phrasing if I was copyediting. Technically speaking, you're splitting the verb has | to be dealt with. It's more common (but still wrong) to split the infinite: has to | be dealt with. It would be best to say | has to be dealt with, but still – doesn't really warrant such a small edit.
 
 
2 hours later…
6:03 AM
@DavidCarlisle I found a bug and didn't get a cheque :) (Who knows, maybe it was a dupe.) I'd wager considering the minimal changes made in the last TeX update that there'll be very few given out again.
@DavidCarlisle @UlrikeFischer — yes, the best (only?!) thing to do is load T1 encoding locally. \newcommand\myfont[1]{{\fontencoding[T1]\fontfamily{emerald}\selectfont #1}} or whatever.
 
6:56 AM
@DavidCarlisle / @JosephWright — I'm toying with the idea of following breqn in unicode-math and making all maths symbols mathactive and have them expand to an appropriate \mathchardef'ed control sequence. The advantage being (a) I can support breqn, and (2) more generally, can add "hooks" to symbols. E.g., have a glossary of maths symbols that doesn't require macro input. Anyway, ridiculous idea?
It would also lead to far more efficient switching for \mathfrak and so on -- rather than iterate through dozens of \mathcode's, can just set a switch that branches inside the expansion of the characters. I haven't done any testing, but looking at \tracingall around \mathbf in unicode-math is a bit scary.
I once mentioned to Hans/Taco that having MathCodeTables like CatCodeTables in LuaTeX would be a really Nice Thing for this sort of thing, but it obviously didn't stick in their minds :)
 
7:14 AM
@WillRobertson one could naturally also create some EU1/EU2.fd-files for emerald. I didn't check if the enc-files would need much adjustment. But I don't think that one really need it. As the fonts are for decoration switching locally to T1 is quite ok.
 
7:30 AM
@WillRobertson Bother Hans at TUG2015
@UlrikeFischer You remind me of something for the LaTeX Team meeting :-)
 
7:58 AM
@WillRobertson Lots of GitHub spam I see (also for @DavidCarlisle from me!)
@WillRobertson I'm planning to put Git(Hub) on the team meeting agenda: OK?
 
8:11 AM
@DavidCarlisle lualatesupport to do: module loading (for the kernel?), whatsit stuff (I'm still not convinced), interaction/compat/fallback with luatex/luatexbase. The latter seems to need discussion so I plan to get the module loading done then write to LaTeX-L.
 
@JosephWright I don't see any (also lualatexsupport doc still errors unless I missed a checkin?)
 
@DavidCarlisle I haven't actually tested the checkins yet :-)
 
@JosephWright Very much so — I think it would be a Good Thing to use github. Although I understand the reluctance and the fact that we're outsourcing could be a cause for concern.
 
@JosephWright Ah, I believe we have a test harness, although you need to know some lua to drive it:-)
 
@JosephWright That's not fair to make me curious ... But you reminded me of a question I wanted to sent to the latex team: Are there any plans for a standarized "ascii input for unicode"? Imho it would be useful if there was some \unicode{20AC}. It would make the input in styles easier then having to use \"a, \thaiChoChang, \cyra, \textalpha, \euro, \texteuro, \hebfinaltsadi. And would make life easier for sorting applications like xindy or biber. (There are some bits in hyperref and ucs).
 
8:14 AM
@WillRobertson I have some arguments
@DavidCarlisle Certainly: will do that in the fullness of time
@UlrikeFischer You remind me I want to talk about the fact we currently have two Unicode encodings, and that is rather odd
 
@JosephWright I'm (heavily) procrastinating on my exam marking. Which is not a good thing!!
 
@JosephWright whatsits not at all sure they are useful, but probably we should do it anyway just do complete coverage rather than pass judgement:-) module loading, not sure quite what you mean here, over require?
 
@UlrikeFischer Might be worth checking the ConTeXt approach: I know they have an ordered set up
@DavidCarlisle See luatexbase-modutils
@DavidCarlisle I want to check on the LuaTeX list about the status of user whatsits
@WillRobertson Ours is done :-)
@WillRobertson I am in a pretty good mood at the moment: got my first substantial grant last week
 
@JosephWright yes looked at that before and wasn't convinced (but can't remember anything this morning) will look again
 
@DavidCarlisle Need at least to have some position on it :-)
 
8:17 AM
@JosephWright They would just say: "Why are you still mucking around with all these TeX macro solutions?" To be honest without an equivalent XeTeX approach it's of less interest to me.
@JosephWright Fantastic!! Erm, how on earth do you find the time??
@JosephWright Next up: grants for LaTeX3 work :)
Okay stop distracting me. Back later.
 
@JosephWright the main thing that module stuff adds over require is add a date checking feature to match \usepackage's date option, and in 20 years approximately no one has ever used that option. I'd just drop the whole thing.
 
@WillRobertson It's taken two submissions over two-and-a-half years
@DavidCarlisle I was thinking more the idea that modules should probably log their version as packages do: that is useful (version mismatches and the like)
 
@JosephWright Well as long as there is no kernel (where kernel includes fontenc) \ifxetex/\ifluatex I don't see how else you could load specific fontdefinitions through fd-files for xelatex and lualatex.
 
@DavidCarlisle With commit 6df70d2eeef83a2a0c2f339d0841a027aafccfdf the following test is all 'works for me'
% !TeX program = LuaLaTeX

% These bits are for the kernel
\RequirePackage{ltluatex}
\directlua{require("latex2e.lua")}

\documentclass{article}
\usepackage{lualatexsupport}

\newcommand*\showcs[1]{%
  \typeout{#1:
    @=\the\catcode`\@,
    space=\the\catcode32,
    Z=\the\catcode`\Z}%
}

\newcatcodetable\CatcodeTableExpl
\setcatcodetable\CatcodeTableExpl{%
  \catcodetable\csname catcodetable@latex\endcsname
  \catcode126 10 % tilde is a space char
  \catcode32  9  % space is ignored
  \catcode9   9  % tab also ignored
 
@JosephWright it's possibly useful but I suspect the lua side should use lua features. If all the lua code you are loading is luatex specific you can do as luatexregisternales/lualatexsupport do and just have a stub tex package that just has teh version information for \usepackage and then loads some lua. But for larger lua code the file inclusion mechanism and logging ought to be just lua facilities, so you can load off-the-shelf lua code
 
8:26 AM
@UlrikeFischer I was thinking you could just look for \directlua :-)
 
@JosephWright try lualatex luatexsupport.dtx
 
@DavidCarlisle Will fix
@DavidCarlisle But perhaps at least have a kernel-level latex.moduleinfo or similar to do the printing?
@DavidCarlisle Oh, I know what this is: I've written the package on the basis we will alter the kernel :-)
 
@JosephWright yes OK if some code wants to advertise itself with a log banner that seems OK but I don't think we should be offering a lua file management system.
@JosephWright yes:-)
 
@DavidCarlisle Remember you can commit to this too :-)
@DavidCarlisle I'll look over the code they have again and think about it
@DavidCarlisle We will still need to work out exactly how we see the transition working, and compatibility for commands is part of that
 
@JosephWright yes I wondered about fixing it last night, but fell asleep instead, and this morning been chatting to you here rather than write code. ...
@JosephWright agreed, it depends a bit whether they want to just tell people not to load lualatexbase in future or if they will want a lualatexbase that emulates everything that it did before (in which case could leave the module stuff to that) as see with etex.sty no easy answer what's best to do there
 
8:45 AM
@DavidCarlisle Yes, but made a bit more tricky as luaotfload also works with plain
@DavidCarlisle I think long-term we might be best looking for a clean break here
@DavidCarlisle Presumably @WillRobertson is in a good position to take a view on this
 
@JosephWright I don't think that fd-files should be cluttered with "if \directlua defined". Also if someone writes some special fd-file for xelatex he/she shouldn't be forced implicitly to adapt it to lualatex (I even don't know if it is always possible. I ran lately in a luaotfload problem with the deva script: github.com/lualatex/luaotfload/issues/277.)
 
@UlrikeFischer Well that's the question, really. We end up with two .fd files because we have two encodings, and that's because of the different syntax for the two engines. When we (the team) talked about this last there was a feeling we could provide a suitable wrapper for use in an .fd file that would take engine-independent input and expand to the correct thing. Needs one or two new commands to be added.
 
@UlrikeFischer but it should be possible to have fd files (if they are needed at all) without explicit test so long as there is a cross-platform way to refer to the font name
 
9:09 AM
@JosephWright @DavidCarlisle: and there again we are at the "if the kernel provides ...". (But such a wrapper for font call don't solve the problem that there is no garanty that fonts work in both engines -- imho xelatex is still better for the indic scripts -- and no garanty that similar font settings call can/should be used for both engines.)
 
@UlrikeFischer That's not the same as saying that we should have two encodings
 
9:23 AM
@UlrikeFischer When I looked over TL we could find no examples where a merge of EU1 and EU2 could not be covered by a small set of new kernel commands
 
@UlrikeFischer yes but that's a general (but different) problem. As the engines develop and possibly diverge they may have different font shaping, different hyphenation or anything else, but marking that by having two names for the same (Unicode) font encoding doesn't seem optimal, as the font and its encoding is the one thing definitely shared.
 
9:49 AM
@JosephWright @DavidCarlisle: But I didn't say this ;-). I only said that one needs first some kernel tools to be able to get around the differences in the engines. But you can't simply exchange xelatex mappings with lualatex feature files, so there also must be some way to indicate "this font setup is only for xelatex". Using the encoding axis to this purpose doesn't sound very logical but it worked rather good. There has been worser misuses of an nfss-axis ;-)
 
10:29 AM
@DavidCarlisle On user whatsits: we don't need a TeX count here at all
@DavidCarlisle Do you think we need to provide the Lua-side interfaces for allocating attributes, etc.?
 
10:39 AM
@JosephWright don't you need some global counter, somewhere?
@JosephWright I don't think so, not initially at least.
 
@DavidCarlisle I pulled down the infinite loop in tex.stackexchange.com/questions/253807/… to this here: -- Damn how do I insert code?
\documentclass{article}
\usepackage{pgffor}
\foreach\x in {1,...,177} %
{\newcount\a}

\makeatletter
\newcount\c@MaxMatrixCols \c@MaxMatrixCols=10
\makeatletter
\usepackage{etex}

\begin{document}
\[
 \array{*\c@MaxMatrixCols c}
  a
 \endarray
\]
\end{document}
 
10:55 AM
@DavidCarlisle Only one at the Lua level, which we could provide
 
@DavidCarlisle Ulrike's string of \newcount commands seems to allocate \count255 and also the \count registers already allocated by \newinsert. It seems wrong, doesn't it?
 
@UlrikeFischer Formatting fixed :-)
 
@DavidCarlisle If I add \message{\meaning\a} after \newcount\a, I get
[...]
\count241 \count242 \count243 \count244 \count245 \count246 \count247
\count248 \count249 \count250 \count251 \count252 \count253 \count254
\count255 \count256 \count257 \count258 \count259 \count260 \count261
[...]
 
@egreg, @UlrikeFischer Well etex is loaded in completely the wrong place so ...
 
@JosephWright But the allocations are done before loading etex.
 
10:59 AM
@egreg Doesn't matter, etex is still loaded in the wrong place
@egreg A document needing the extended pool needed \RequirePackage{etex} before \documentclass
 
@JosephWright I don't want that \newcount\something makes \something equal to \count255. Do you?
 
@egreg Why not?
@egreg \count255 isn't reserved
 
@JosephWright Yes, it is: \count@. But also \count254 is reserved because the kernel does some \newinsertcommands.
 
Does anybody else have no idea of what they are talking about? :)
 
@egreg Ah, I meant by the engine (was forgetting 2e does that)
@egreg @DavidCarlisle probably should know that :-)
 
11:03 AM
@JosephWright Letting LaTeX allocate \count255 will break several packages that use \count@ for scratchwork. Not to mention the breakage to the \insert mechanism.
 
@egreg Like I say, I'd forgotten about \count@ but then I didn't write the allocator code
@egreg If @DavidCarlisle sorts it I can fix on CTAN today I think (seems urgent)
 
@JosephWright yes but I thought it might look more consistent to do them all in tex, but don't mind
@egreg that isn't supposed to happen, did I mess up I checked that at one point... will check again...
 
@DavidCarlisle The whatsit one is very different from all of the others as there is no TeX access to the concept at all
 
@JosephWright well i assumed whoever was defining the whatsit would provide the tex access \def\myfunkyspecial#1{\directlua...
 
@JosephWright @DavidCarlisle Shouldn't \usepackage{etex} do nothing with the new kernel?
 
11:07 AM
@DavidCarlisle Ah, good point
@egreg We've been through this: 'fraid not
@egreg etex does more than just extend the allocators
 
@egreg that would be a possibility but then all the stuff we didn't copy exactly breaks, local allocation, names for interaction modes \reserveinserts etc. the idea was that if you use \usepackage{etex} you get exactly what you get before so compatibility is increased, but you don't get the new fearures
 
@DavidCarlisle Provided you load etex in the right place, of course ;-)
 
@JosephWright details details (and provided other packages don't decide not to use \reserveinserts{32})
 
@DavidCarlisle Indeed
@DavidCarlisle There was a good reason for that
 
@JosephWright yes same as good reason for adding extended allocation in the kernel and not making etex a no-op, but getting the exact balance still proves general tex rule that any change breaks something
 
11:22 AM
@DavidCarlisle Quite
 
@JosephWright in the original question etex was loaded by a package (tkz-berge) in the midst of a bunch of other packages. It took me some time to find the correct number of count allocations to get rid of all the packages. -- And I still don't understand why it then breaks the *\c@MaxMatrixCols in the array (using e.g. *{10} instead works fine).
 
@egreg oh I see it's etex (which redefines the meaning of the counts in count10 etc to be one out) I thought you meant the new kernel allocation was wrong (@UlrikeFischer, @JosephWright) I'll see if I can make etex.sty recover better in new formats)
@UlrikeFischer it isn't an infinite loop, just that matrix cols is a bit big:
> 32768.
l.14 \showthe\c@MaxMatrixCols
@UlrikeFischer, @JosephWright, @egreg, the count register has been allocated twice so been reset....
 
11:38 AM
@DavidCarlisle well a large enough number is as good as infinite ;-).
 
@DavidCarlisle The problem is probably in that \foreach works in a group, and the allocation macros do their job by local assignments. Indeed, if I use a standard \loop instead of \foreach, the reserved registers are jumped over.
 
@JosephWright Thanks. And how do one correctly insert longer examples? It was suddenly there but I don't know what I did. Is there somewhere a training chat room?
 
@egreg: It's quite amusing to see you and David discussing real TeX witchcraft instead of poking each other. :)
 
@UlrikeFischer You simply paste the code in the bottom box; if the text contains returns, a button “fixed font” appears on the right. Press it.
 
@egreg A red button, which makes it even more interesting. :)
 
11:49 AM
@PauloCereda I wanted to be politically correct, so I didn't mention the color. :P
 
@egreg ooh :)
@SeanAllred This. <3
 
@PauloCereda <3
By the way, TAOCP is now [was? has been?] available as an ebook: redd.it/3ca40f
 
My TeX level is, look ma, I know how to produce bold text! /writes {\bf quack} deliberately
@SeanAllred ooh indeed, I haven't seen the result, but it's surely a great acquisition. :)
 
@PauloCereda That's about the sum of it for me, too :)
Honestly some of the stuff in the LaTeX kernel is just your run-of-the-mill witchcraft to me. You know, the kind of stuff you throw water on and hiss at.
 
11:53 AM
@SeanAllred :)
user image
2
 
12:12 PM
@PauloCereda normal service will be resumed soon
 
@DavidCarlisle LOL
 
@UlrikeFischer to the user, but not to working out what's happening
@PauloCereda my Java program now has 162 lines (and less than half of them are import...) , does that make me a real Java programmer?
 
@DavidCarlisle 162 lines? You should be considered a Java guru for writing very short codes. :)
 
@PauloCereda I'm not sure I really understand all of them:-)
@UlrikeFischer, @egreg @JosephWright If you take Ulrike's document above and use TL2014 then you get a different error:
l.4 {\newcount\a}

?
! No room for a new \count .
 
@DavidCarlisle Like I said, etex is in the wrong place :-)
 
12:21 PM
It's always been teh case that you need to load etex.sty before you run out of classic registers...
... so before you just ran out, but now you start using extended registers but the different etex allocatation means that you get multiple allocations (here the latex cmax column counter is being over-written by etex's internal local allocation count. Given that it gave an error before, and gives an error now, i am minded to call it a feature
@JosephWright yes but question is whether we should detect that and try to recover or just let it go.
 
@DavidCarlisle I assume that the original problem didn't show the same: @UlrikeFischer?
 
@DavidCarlisle Who does? :)
 
@JosephWright no it just worked before, I'll go back to that question and see exactly what it's doing should be easy now @UlrikeFischer has isolated main issue.
 
@JosephWright Maybe! I'd say that if we're seriously adding XeTeX/LuaTeX support to the kernel, we should add at the very least the standard \ifpdftex/\ifxetex/\ifluatex switches (perhaps under the names \ifengineXYZ?) … and on the topic of .fd files, I'm unsure whether we should continue with the bifurcated definition for unicode (EU1/EU2) or define some wrapper commands to unify them. ((Also need to pull xunicode.sty into fontenc in some form or another...))
 
@WillRobertson One for TUG2015 discussions
 
12:31 PM
There are two problems. The one is the allocation of count 256 when using `\foreach`, the other is the wrong value of \c@MaxMatrixCols. The second problem can also be reproduced with a normal `\loop` but with a different length
\newcount\loopcnt
\loop \ifnum \loopcnt<147
\newcount\a
\advance\loopcnt by 1
\repeat
 
@UlrikeFischer Ah, you've discovered that you can't have a reply to someone and a multi-line code block in the same chat message :-)
@WillRobertson From memory I proposed an approach inside .fd files where we provide an 'engine-aware' command that does the work
 
@UlrikeFischer both of them are cause by bad luck:
\documentclass{article}
\makeatletter

\usepackage{biblatex}
\usepackage{hyperref}
\usepackage{exercise}
\usepackage{mathtools}
\usepackage{tikz}
\showthe\count10
\showthe\insc@unt
\usepackage{tkz-berge}

\showthe\c@MaxMatrixCols
\show\c@MaxMatrixCols
\begin{document}


\[
\begin{matrix}
  a
\end{matrix}
\]
\end{document}
 
@JosephWright One lives and learns ...
 
@JosephWright But like @UlrikeFischer says, perhaps that's shortsighted when we consider we might want separate setup for the same font in the two engines. OTOH… it is also a bad idea, probably, to have a font with the same name have different behaviour in the two engines!
 
That particular combination of packages makes the \count allocator \count10 be exactly the value of \insc@unt where the allocation flips to the extended range so it just manages to avoid a no room error in 2014 but still gets clashing allocation schemes in 2015. I think probably etex.sty can detect this....
 
12:35 PM
@WillRobertson I guess my feeling is it should be possible to have one description of how to use a font for the 'user' and translate that properly into engine directives. Otherwise it suggests that there is a flaw in the engine font loading logic or similar.
@WillRobertson Certainly I don't like having two encodings. We could argue for engine-specific .fd files where that was explicit
@WillRobertson If you look at most of the current EU1/EU2 cases they are mechanical variations only
 
Not quite what I said above, but still think etex can probably avoid this
 
@DavidCarlisle In my tests a range of value (147-150 in the example with \loop) lead to a wrong value of \c@MaxMatrixCols.
 
@JosephWright Well, the vast majority of fonts loaded with EU1/EU2 are (probably) dynamically created in fontspec… One argument for keeping them separate is that the syntax between the engines is coincidentally, mostly, similar. But what would we do, for example, about Graphite fonts that can only be loaded in XeLaTeX?
 
@WillRobertson I'd say 'tough': there are lots of other engine limitations where things just fail if you use the wrong engine
 
@WillRobertson but it complicates lots of other code eg if you want to set up hyphenation patterns per font encoding you have to special case merge EUx to be the same thing
 
12:43 PM
@DavidCarlisle — BTW now have new \setmathfontface in unicode-math repository. Allows the creation of new \mathXYZ commands. Basically a wrapper around \DeclareMathAlphabet and \SetMathAlphabet.
 
@DavidCarlisle On the LuaTeX modutils thing, do you see any value in provides_module and having a table of module info (_cf._ \ver@...`)
 
@DavidCarlisle That's a good argument!!
 
Isn't chat useful :-)
 
@WillRobertson the range of "available system fonts" depends on the search paths and the kinds of fonts the engines can read, that is likely to be different in principle.
@JosephWright as I say my guess is no, as I think people will want to load random bits of lua code which won't have that data so it's not really like loading tex code where ypu can say it should have a \Provides... line to play a full part in the system
@WillRobertson hmm I guess I should fork or clone or something that and play locally...
 
@DavidCarlisle I don't know if I would wish that upon you… :)
@DavidCarlisle How about I ping you again when I'm closer to an actual release
 
12:48 PM
@WillRobertson unlikley to do anything before darmstadt anyway, between day job and luatex stuff and possibly sorting out etex allocation:-)
 
@DavidCarlisle good to hear :)
oops, getting late again. early morning tomorrow. have fun without me! night.
 
@cfr :-)
 
@JosephWright Imho neither EU1 nor EU2 are encodings like the other encodings. They don't describe a concrete set of symbols supported by the font, the enc.def-files don't describe slot positions. I always saw them as "engine instances" of the general unicode.
 
1:04 PM
@JosephWright @UlrikeFischer Oh now finally I understand the MaxMatrixCols issue. It's etoolbox.... 2014 gives no error as etoolbox is loaded early, which loads etex.sty early so all allocation uses the etex scheme, in 2015 etoolbox doesn't load etex, so the kernel scheme is used and allocates counters in the 265-276 range then etex.sty is loaded and over-writes them. 270-276 getting set to 32768.
 
Inspired from a question on the site: is there a callback or similar method for hooking into just math?
 
So at the very least etex.sty should detect the counter has already gone past 255 and then give an error saying to put \RequirePackage{etex} at the start of the document, or remove etex package.
@daleif mlist_to_hlist but that probably isn't what the user wanted
 
@DavidCarlisle sounds like a good plan. (perhaps add a comment in the error that this is be cause of 2015 changes to the kernel, such that people understand why it might have worked in earlier TLs)
 
@daleif I could make the message say it's because @JosephWright changed etoolbox :-)
2
 
@DavidCarlisle sure blame the poor guy for everything
 
1:08 PM
@daleif I was thinking the same: no :-(
 
@daleif seems like a good plan to me
 
\directlua{
  for i,_ in pairs(callback.list()) do
    texio.write_nl("- " .. i)
  end
}
\bye
@daleif ^^^^
@DavidCarlisle I was asked to (that's why I've ended up with etoolbox)
 
@JosephWright probably asked by me, but that doesn't invalidate the truth of the above statement!
 
@JosephWright brain storage capacity exceeded. Too many languages to keep track of.
2
 
@JosephWright so if etex.sty detects count10 (etc) already past 255 then it knows another extended allocation scheme must be in place, does it give a warning and stop, give a warning but carry on and define its scheme, give an error rather than a warning?
 
cfr
1:12 PM
@JosephWright ??
 
@cfr Reply to your comment on the meta thread
 
@DavidCarlisle My example doesn't load etoolbox, and even pgffor can be removed. So only article and etex - Joseph can't be blamed here ;-).
 
@UlrikeFischer Couldn't we blame him anyway? :)
 
cfr
@JosephWright Ah, OK. That was my best guess but it was not really clear. The system confuses me ;).
 
@UlrikeFischer no but your example gives an error in tl2014 so no compatibility issues there.
 
1:27 PM
\let\cheerio\bye
 
@UlrikeFischer (@PauloCereda, bout time you did some testing) @JosephWright @egreg Ulrike's example works if you add this to top of etex.sty, original example doesn't but get more sensible warning
\@tempswafalse
\ifnum\count10>\@cclv\@tempswatrue\else
\ifnum\count11>\@cclv\@tempswatrue\else
\ifnum\count12>\@cclv\@tempswatrue\else
\ifnum\count13>\@cclv\@tempswatrue\else
\ifnum\count14>\@cclv\@tempswatrue\else
\ifnum\count15>\@cclv\@tempswatrue
\fi\fi\fi\fi\fi\fi

\if@tempswa
\PackageWarningNoLine{etex}{Extended allocation already in use.\MessageBreak
Aborting etex.sty.\MessageBreak
To force etex package to load, add\MessageBreak
\string\RequirePackage{etex}\MessageBreak
at the start of the document}
 
@DavidCarlisle <3 I was feeling outcast.
:)
 
@DavidCarlisle I'd go for an \ifnum and not set the temp switch, but that's just me
 
@JosephWright yes perhaps, but is warning and \endinput the right thing to do?
 
@DavidCarlisle Perhaps go for an error: people do tend to ignore warnings
 
1:44 PM
@JosephWright always ignore, but most documents (eg ulrikes) just work
 
@DavidCarlisle If we want to squash the problem longer-term perhaps we have to accept a bit of error-reporting now
 
@JosephWright Ulrike's problem is that etex is loaded by another package. An error message would be wrong, in this case.
 
@egreg Not sure it would: the advice would still be correct
 
@JosephWright Users loading tkz-whatever will be puzzled when receiving an error message they have no way to solve.
 
@egreg I'm obviously missing something: what's wrong with the advice from the message?
 
1:51 PM
@JosephWright Here's the document that started the debate:
 
@JosephWright basically that they now have to react to some package they have never heard of in order to make their document run. A document that ran fine on 2014 kernel
 
\documentclass{article}

\usepackage{biblatex}
\usepackage{hyperref}
\usepackage{exercise}
\usepackage{mathtools}
\usepackage{tkz-berge}

\begin{document}
\[
\begin{matrix}
  a
\end{matrix}
\]
\end{document}
@JosephWright Suppose you modify etex.sty to raise an error instead of the warning suggested by @DavidCarlisle: what should a user do?
 
@LaRiFaRi Welcome to goLaTeX :-)
 
@egreg Yes, they've updated their packages and kernel and are being told to make a change. Either we give a warning and they ignore it or we give an error to force them to act.
@egreg Exactly what the message says: add \RequirePackage{etex} before \documentclass
 
@JosephWright They don't know that they need to do that. Of course the best solution would be to update tkz-berge.
 
1:56 PM
@egreg Problem is tkz-graph, no?
 
@JosephWright normally I'd take a hard line as well, but in this case if you get a warning then you might get a later error eg \locbox was undefined, but in that case you were warned, but in the majority of cases the document will just work and in fact etex stopping is better than forcing it to load (so \extrafloats etc work)
 
@DavidCarlisle Ah, that is true
 
@JosephWright The fact is that loading etex or not is quite a deep technical issue and the normal user will be dumbfounded by that error message: they expect packages to work.
 
@egreg I guess someone should contact the author
 
@JosephWright I'll do it I've been contacting a lot of authors recently (when I can reach them)
 
1:59 PM
@egreg Yes I know but unless you go for no change at all ever (cf. plain) then you have to deal with some issues sometimes
 
@daleif No, with Davids changes their document should normally run fine. They only will get into problems if for some reason they really need etex.
@DavidCarlisle I would probably make it a bit clearer that it is not necessary to load etex. "You probably don't need etex and can safely ignore this message. If you want to force etex to load use ..."
 
@UlrikeFischer perhaps, although I don't think saying ignore warning is good long term policy I'd rather push people into getting packages updated.
 
@DavidCarlisle Then add "if you want to get rid of this warning inform the package maintainer that he/she should do .... ". Then the users can do some of the pushing ;-)
 
@JosephWright @clemens @StefanKottwitz If anybody could make the images smaller? Just had a real crappy cup of coffee with somebody (well, the coffee was fine, the talk was ...) and now my motivation to do anything is approaching zero. And it is warm. And monday. Die Woche geht echt gut los. latex-community.org/forum/…
 
Hmm, many files in TL15 that contain {etex}
 
2:21 PM
addlines, bigfoot, catchfilebetweentags, chemexec, cryptocode, datatool, dtk.cls, eledmac, elocalloc, esami, etextools, interfaces-base, keycommand, memhangul-ucs, oblivoir-base.cls, leaflet.cls, linegoal, ltxnew, memoir.cls, msu-thesis.cls, pdfcolmk, pythontex, schule, sesamanuel.cls, sesamanuel, songs, thm-kv, ttlps.def, tkz-base, tkz-doc.cls, tkz-graph, tkz-kiviat, tkz-linknodes, tkz-tab, umthesis.cls, xifthen (extension =.sty if not mentioned).
Plus autonum, changelayout, chemschemex, concepts, cryptocode, dataref, getfiledate, mycv_base.def, ribbonproofs that load etextools. I don't count those that load etoolbox, because it loads etex only with older kernels.
 
2:38 PM
@egreg We are already getting several of those updated
 
2:51 PM
Thanks for comments all: I decided to keep it pretty much as I suggested, will push to ctan later today unless someone screams no. git commit: github.com/davidcarlisle/dpctex/commit/…
 
@DavidCarlisle AAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAA
<3
 
@DavidCarlisle Speaking of recent bugs, did you look at latex-project.org/cgi-bin/ltxbugs2html?pr=latex/…?
 
@PauloCereda thank you for the helpful comment and in depth testing
 
@DavidCarlisle I could even put the \let thingy inside a language conditional, so we could also have \let\ciao\bye for Italian. Advanced TeX. :)
 
@JosephWright ouch no (spam filtered for some reason:(
 
2:57 PM
@Johannes_B Thank you. Are you the moderator there?
 
@LaRiFaRi Si :-)
 
@Johannes_B muy bien muy bien. Busy man.
 
@Johannes_B Thank you for the pointer! I asked the user to remove the images and to use code as text instead.
@LaRiFaRi Yes, Johannes is the most active moderator and most helpful LaTeX supporter there
 
@StefanKottwitz Thanks. Was a crappy day today. I think i follow the example of @arthur today and get drunk :-)
 
@StefanKottwitz Just have seen that. And found your name somewhere in the impressum as well. Well, if that's so, I registered to the site. But actually I have enough to do here... we will see.
 
3:08 PM
@Johannes_B Good for you :-)
 
@StefanKottwitz Without @ulrike, @esdd and Besserwisser, i would be completely helpless.
 
@LaRiFaRi Cool! We need helpers there. Just click: Unanswered posts (direct link here, or menu top right there) - besides the currently active threads there's a lot older to reanimate or to complete with an answer.
 
The discussion about etex has been fruitful, bringing me another vote for tex.stackexchange.com/questions/38607/no-room-for-a-new-dimen :)
 
@JosephWright PL2 ? (Petra will kill me)
@egreg no justice in the world
 
Apparently I'm the maintainer of yet another package I will never use. And someone found a bug.
 
3:19 PM
@AlanMunn That's better than being the maintainer of a package that you are trying to get nobody to use (see above discussion:)
 
3:29 PM
@DavidCarlisle Yes, I suspect so
 
@DavidCarlisle :)
 
@JosephWright did you see the comment (Frank I guess) above the bit that's wrong? vvvv
%    \end{macrocode}
%   Next set of docstrip guards are a bit weird, essentially
%   |\@addtotoporbot| ends up inside the kernel and the
%   \texttt{fltrace} package and |\@addtotoporbot| shows up in the
%   \texttt{flafter} package. Guess that could have been done a bit
%   more obvious :-)
quite prophetic given that I then messed up exactly that point while trying to weave in the latexrelease guards in addition to 2ekernel,flafter and fltrace....
 
@DavidCarlisle :-)
 
3:54 PM
Felt angry, annoyed, exhausted and sad... soundcloud.com/beeftheatre/rage-against-the-machine... feeling pumped again
 
4:32 PM
@StefanKottwitz hey, that's another homepage :-) One new homepage is enough for today.
 
5:03 PM
How could @JosephWright or @PauloCereda have forgotten to post this, given current events. :)
"Socrates has scored! The Greeks are going mad, the Greeks are going mad! (...) The Germans are disputing it."
 
@AlanMunn ooh :)
 

« first day (1709 days earlier)