4:00 PM
@JosephWright I would guess that this would be a chicken/egg situation.

@SeanAllred I might bring it up at the physical team meeting we are going to be able to have at TUG2015

@SeanAllred sure chat is chat but the latex sources have switched from rcs to cvs to svn, but I don't think there has ever been an real proposal to switch out of svn

@DavidCarlisle Indeed
@DavidCarlisle Would be more important if we had branches to worry about

@SeanAllred yes also I have just switched all my open source projects to github as google just killed off googlecode

@JosephWright you never feel the need for branches? Or rebasing?

4:01 PM
@DavidCarlisle ^^^ github support for subversion clients

What happens when you both work on stuff at the same time, and need to merge?
Or is there not enough overlap between you guys to worry about?

@FaheemMitha For siunitx I have two branches (dev/stable): other than that, not really

@FaheemMitha My guess is that since this is fundamentally a volunteer position, work doesn't happen as quickly. The need isn't really there.

@JosephWright oh

@FaheemMitha Rebasing still seems to me to be odd: a version control system should be like a lab book, no Tippex allowed!

4:02 PM
@SeanAllred Not sure I know what you mean.

@FaheemMitha Should never happen

@JosephWright The rebasing is done "off the books", before it hits the public repos.
@JosephWright why?

@FaheemMitha Still not really 'right'

@FaheemMitha For a long time, I never used git branches. I never really had the time to try out different ideas simultaneously. It was only during my thesis that I actually started using branches.

@JosephWright I think that is fine. :-)

4:03 PM
@FaheemMitha I'd always start again if someone else had changed something

@SeanAllred Ok.
@JosephWright Oh. And it doesn't happen often?

@FaheemMitha We usually agree who is handling what before anyone changes stuff
@FaheemMitha Not usually

I don't use branches in mercurial either. But then mostly it is just me.
@JosephWright I see.

@FaheemMitha Our approach is basically linear

@JosephWright yes, I see that.

4:04 PM
@SeanAllred no point really, if you switch you switch, no point in changing the underlying repository then trying to pretend you havent:-)

That's fine if it works for you. But it doesn't scale.

@DavidCarlisle Well, you did say the main advantage of using git would be GitHub :)

@SeanAllred Clients not a big issue: I think we are all using the CLI
@FaheemMitha I'm always surprised by this: don't big teams agree who is in charge of each area?

@JosephWright well, the CLI is a client – at least that's the program the linked article uses

@FaheemMitha the latex project isn't a scalable project It's never had active members in double figures.

4:06 PM
@JosephWright Well, complex projects don't necessarily subdivide so neatly. There is lots of scope for people to step on each others toes. They might be editing the same files while working on a different feature, or something.

@DavidCarlisle :-)

@DavidCarlisle Yes, you guys are a tech elite in close contact with each other. Not exactly the typical software development paradigm. :-)

@FaheemMitha Should agree that in advance and commit sequentially, but really I'm very surprised people try to change the same thing at the same time

@DavidCarlisle I just don't see how we can logically say that's a thing – I really believe that the 'project team' would grow if it was easier to contribute.

@SeanAllred Naive

4:07 PM
@JosephWright Perhaps.
@JosephWright But genuine.

@SeanAllred It's very easy to contribute: mail us a modified source. Doesn't happen

@JosephWright It's a coordination thing. I think the complexity goes up quite fast with n, where n is the number of people involved.

@JosephWright That's not easy to me.
It's not easy to see what changed, it's not easy to see why, it's not easy to discuss specific changes.
Though, I'm certainly really lazy when it comes to stuff like this. I admit that I'm used to doing things a particular way.

@SeanAllred Funny: to me it's probably easier than GitHub: it's clear you are not forking and are making a suggestion to someone

(As are you)

4:09 PM
It becomes increasingly difficult to keep track of what everyone is doing. After a certain point you can start spending as much or more time keeping track of what people are doing as geting work done.

@SeanAllred but if there are not the resources to manage a large distributed project, not allowing that growth is a positive thing.

@DavidCarlisle out of curiosity, what resources would be required?

@FaheemMitha Never done a large project so can't comment

@SeanAllred time

The other great thing about DVCS's is the ability to commit locally. This is really useful. And then you can back up to your own private repos. Don't have to worry about losing your changes.

4:10 PM
@DavidCarlisle I can definitely sympathize with that.

Losing work is the stuff of nightmare for a sw developer.

Lots of open PRs on sx.el still.

@SeanAllred, @FaheemMitha One thing SVN offers that Git is violently against is that the sources themselves have the revision in ($Id): that's still important to us @JosephWright Mercurial still offers that. :-) And you can use tags. @FaheemMitha That certainly is useful if on the train or whatever 4:11 PM @JosephWright Yeah, I've noticed the sources use that a lot @FaheemMitha Yes, I know about Mercurial, but it's not built in so confuses people But things like gitinfo2 are a good compromise – I'm certain an equivalent system could be devised @JosephWright It's generally useful. Do you guys just have unsaved changes till you commit? And what about while you are working on it? What happens to intermediate changes? @FaheemMitha Memory stick or whatever @JosephWright So, you make lots of backups? 4:12 PM @FaheemMitha necessarily, they vanish, but in the few times I've worked with a CVS repository, I've created a local git one to track my changes @FaheemMitha Intermediate changes are my point about a lab book: each commit should be what it is, if you mess it up the log should say so @FaheemMitha No, not really @JosephWright I use temporary commits to save local state. When I'm done, I make a final commit. @FaheemMitha Very rarely do anything that does not go to the SVN pronto @JosephWright That sounds like a lot of pointless clutter. What's the point of documenting your mistakes? @FaheemMitha Lab book 4:13 PM @JosephWright Small changes then? If your system is very modular, that might work. @FaheemMitha no one is working live off the svn repository (ctan is for usable versions) so you can commit to the main svn branch as you work. @DavidCarlisle Ok. That makes for a lot of churn, though. Of course, maybe you don't make many mistakes. @FaheemMitha A lab book records what you did, not what you wanted to do: I see version control in the same way I might have 10-20 commits for one usable commit, depending on what I'm doing. I don't want to see all that crap. @FaheemMitha why think of svn commit as being git commit and ctan upload being git push, why is that churn? 4:15 PM @FaheemMitha Have your read our log ;-) @DavidCarlisle Well, for one thing, a git commit doesn't have to be final. @JosephWright Nope. @FaheemMitha 'Typo' comes up a lot @FaheemMitha It's good reading :) @FaheemMitha ask Joseph if ctan uploads are final... @JosephWright you should take a look at the new evolve Mercurial extension. It's state of the art. 4:16 PM @DavidCarlisle Oh don't It's where it's at, man. @FaheemMitha At some point you'll realise that no one is going to use mercurial for new projects @JosephWright my notion of commits it that you keep things clean. So later on people can easily see what you were doing. @DavidCarlisle Sure they do. Well, maybe not you guys... @FaheemMitha I've notice that in the Git workflow stuff, and have messed about a bit to try it out, but it still feels odd @JosephWright It's a different way of thinking 4:18 PM @FaheemMitha atlassian.com/git/tutorials for example i don't see the point of having lots of stuff saying - oh crap, I should have done this, but didn't, here's another try. Or, "scratch commit". @FaheemMitha I have a feeling neither of us has kept a lab book :) @FaheemMitha Why not? Tells other people what you were really up to. I'd imagine in a business situation people need to see that detail @JosephWright They absolutely don't. @JosephWright You people do TeX. Which nobody else can understand. I see this as straining at a gnat. To use biblical terms. 4:19 PM @SeanAllred For performance reviews, surely? @JosephWright People are looking at a log for information,usually, not to see what typos you made when. @JosephWright Nope. For performance reviews, it's generally a matter of how complex the solution was and how long it took you. And the cleaner and clearer it is, the better. In industry, nobody (should) care about your mistakes – they should care about what you've done right (now, within reason…) Of course, if there are only 3 people working on it at any given them, and you are all talking to each other all the time anyway, then it may not really matter very much. 4:21 PM @FaheemMitha yes some do, but google didn't kill goolecode to push people on to mercurial, they left a strong hint to use github, microsoft didn't opensource the .NET runtime on to mercurial (or even their own codeplex) they used github the w3c HTML test suite was on mercurial but now it' on github, there is sort of a theme... How many people are currently active on this repos, anyway? @FaheemMitha, @SeanAllred I don't think you'd get on with the rules for the LaTeX2e sources, which predate using SVN and are 'step the version for each code change' @DavidCarlisle Sure, it is dominant. No question about that. @FaheemMitha LaTeX2e: Frank, David, Me; LaTeX3: Frank, Bruno, me (perhaps Will) But there are plenty of people using Mercurial. 4:22 PM @JosephWright I don't want to seem dismissive, but I don't know why that rule is still in place. @DavidCarlisle BitBucket have moved from Mercurial to Git by default ... @JosephWright So 4 people. Not that many. @FaheemMitha Certainly @JosephWright unfortunately. @FaheemMitha Mercurial isn't going to happen :) 4:23 PM @SeanAllred Isn't going to happen for what/who? If you mean for LaTeX, I'm just using it as an example. Not trying to push it. @FaheemMitha It's never going to be the large-scale dvcs. Doesn't make any difference to me, anyway. @SeanAllred large-scale dvcs? what does that mean? @FaheemMitha At the moment, git is too big to fail. I'm sure another will come in time that will replace git, but it will not be hg @FaheemMitha there are plenty of people using CVS (I was using it a few minutes ago:-) but it would be a strange choice for a new project, given the way the tide is flowing @FaheemMitha, @SeanAllred Don't get me wrong: I'd be happy to move to Git 4:24 PM @SeanAllred agreed. @DavidCarlisle true @FaheemMitha and true of mercurial:-) @JosephWright I just wish I could impress how beneficial it could be for the LaTeX project as a whole Not just in how the sources are managed, but in how public the tech is Yes, git seems pretty dominant for the forseeable future. Nobody is disagreeing with that. But it is also a community project. As long as there are people using it, and people working on it, it will continue. @DavidCarlisle not true of mercurial. CVS is basically dead. Development is at zero. On the other hand, Mercurial is being very actively developed. @SeanAllred step the version? So that the log file shows which version a user is using @DavidCarlisle But that can be automated with a git hook – after all, only usable versions are sent to CTAN (ideally ;)) 4:30 PM @JosephWright what version is that? the version counter? @FaheemMitha \listfiles output @SeanAllred we also have \changes entries for a typeset change log, which are currently date and version based. Of course there are alternatives but that is an unbroken 25 year change log, and not something to just drop as it doesn't fit a natural distributed git flow.. @SeanAllred ? @DavidCarlisle why can't you import it into git? @DavidCarlisle Why doesn't it fit into the natural git flow? I use such changelogs personally. @DavidCarlisle DVCS is almost a superset of centralized VCS. There are a few things DVCS can't do that centralized systems can, but they are pretty marginal. Everything else can be emulated. E.g. locking, which makes no sense for DVCS. 4:34 PM @SeanAllred the \changes entry has a filedate and version number, which implies that the changed file has a (new) date and version, which was what you were implying ought to be dropped in favour of an automated hook on push @DavidCarlisle Certainly doesn't need to be. @FaheemMitha of course, as I said above, but if you are not using the decentralised features, using svn as it is designed is simpler than using git in a way that emulates svn @SeanAllred doesn't need to be what? @DavidCarlisle I doubt you make a \changes entry for typos. Doesn't need to be dropped. @SeanAllred \changes{v2.3j}{1994/11/10}{fixed a few typos} @DavidCarlisle No reason that can't be kept. I'll try to devise an example system (for fun) when I'm off work For now, I have to study for an exam :) 4:40 PM @DavidCarlisle Well, some of the decentralized features can be useful, as I attempted to explain above. I sense you are not convinced, though. 5:00 PM @FaheemMitha Still on this :-) (Have been away for a few minutes) @JosephWright ? @SeanAllred We have a mirror of the SVN for LaTeX3 on GitHub, originally Will's idea but generally quite handy (notice how I never point anyone to the SVN web interface) @FaheemMitha Discussion still seemed to be going on @JosephWright Oh. No, I don't think so. That was my last comment. And I don't get the sense I am changing any minds around here. Not that I expected to. @FaheemMitha I think it's fair to point out I have no proper programming or CS background, only what I've learned 'on the job', so may well miss things that are useful/standard. I know how I use version control (Git nowadays for my own code, Mercurial in the past, SVN for the team stuff): there may well be other approaches. @FaheemMitha You are missing the fact that I'd welcome a move from SVN to Git :-) But at always, it is interesting to see that things one takes for granted are not shared by others. @JosephWright Yes, but you already did. :-) 5:04 PM @FaheemMitha Ah @FaheemMitha Will Robertson would be very keen but I think we both know we'd need to convince all of the team and that would not be easy Now if I could have made David exclaim - how could I have been so blind! the scales have fallen from my eyes! etc. etc. Well, that would be something. @JosephWright From what you describe of your workflow, it doesn't sound like it is critical, and if what you have is working well for you, I don't see a problem. @FaheemMitha David is not the worry :-) As they say, if it ain't broken, don't fix it. @FaheemMitha That's at least in part true @JosephWright well, he clearly doesn't see the point of DVCS. At least in the context of the LaTeX 3 project. I personally think the point of DVCSs is obvious. Though I realised long ago not everyone sees it like that. 5:07 PM @FaheemMitha It would be nice to be able to accept pull requests on GitHub, for example @FaheemMitha Remember the canonical nature of the LaTeX sources is important: that's a big issue After I first saw a DVCS (Mercurial) it took me like 5 minutes to see the point of it. Or so it seems in hindsight. Might have been longer in reality. @JosephWright But don't other projects have that requirement? As @SeanAllred said, you can simply bless one repos/branch as the canonical version. @FaheemMitha Yes, but only in a 'convention' sense: with SVN we have root ownership of the master in a technical sense @JosephWright It's an enforceable convention. Practically, I don't see a difference. @FaheemMitha You can download a TeX file and use it without compiling the source: that's not true with anything that builds binaries. So there is a lot of concern in making sure if some user picks up a file the file itself can be traced back to a specific version. So, who are the big DVCS holdouts in your group? Or would you prefer not to say? 5:12 PM @FaheemMitha It's more a feeling: we've never actually discussed it @JosephWright oh @JosephWright I'm not clear how centralised vs distributed makes a difference here. @FaheemMitha The question is, as @DavidCarlisle says, where are the benefits for our workflow: a 'sales pitch' is going to need something tangible @JosephWright Then maybe you should put together a sales pitch. :-) Or a transition plan. Assuming you have the time and energy. Are Github pull requests a real possibility, btw? @FaheemMitha SVN has a globally unique checkin serial number and (by convention) an incremental version number. You can arrange for SVN to add an $Id\$ line to the sources (expl3, for example) or always increment the version on every checkin (LaTeX2e). The former plan fails with a DVCS, the latter is certainly more tricky (although still workable)
@FaheemMitha We've had some pull requests for l3build, which at present I have to add by hand. I'd probably rework some bits of them in any case, but it would be better to accept the merges then modify after-the-fact.

@JosephWright Oh, I see. You want the version number embedded in the source of each file?

5:16 PM
@FaheemMitha For LaTeX2e I don't think we'll do anything: SVN works, development is very restricted and needs a lot of understanding. For LaTeX3 things are different.

Hi! What do things like [1 mean in the log file?

@FaheemMitha At present, yes: see expl3 sources

@JosephWright ok

@Szabolcs Page shipped out with something happening
@FaheemMitha We have the GitHub mirror already: other than dealing with version strings, a switch would be trivial at a technical level

@JosephWright ok

5:19 PM
@FaheemMitha I do see the business that @SeanAllred alludes to about being 'accessible': GitHub is certainly a better place to report issues than Gnats (the LaTeX2e bug tracker)

@JosephWright yes, github is popular.

@FaheemMitha Probably we (or at least some of us) would pick up branches and the like for features if we switched: would be a cultural change so not immediate

@JosephWright ok. you can do branches in svn, but it is a nutty system. different directories - merging must be a nightmare.

@FaheemMitha I know about branches in SVN: I do not want to try that! (siunitx went from no version control to SVN with two branches. It's fine if you only want dev/stable and to merge rarely.)

@JosephWright agreed. you don't want to try that.
2

5:22 PM
@FaheemMitha :-)
@FaheemMitha I might be wrong about the Git business: I think though this is one for a face-to-face discussion, which handily will be possible at TUG2015 for quite a large part of the team

@FaheemMitha Likely objections from other team members

@JosephWright oh, you mean they might not object?

@FaheemMitha Certainly possible as it's never been discussed in any substantive way

I' going to paste this here as well, since I'm available in chat to clarify anything that needs clarification ...
0

When running pdflatex -interaction=nonstopmode I am looking for: Ways to detect whether there were errors solely based on the log files (assume we don't have access to the process exit code). Extract a user-friendly error message from the log file. Error messages seem to always start with a !...

5:26 PM
@JosephWright ok. Well, I think DVCS certainly offers some advantages. Though I'm no fan of Git. terrible UI.

@Szabolcs auctex or any of the tex IDE are doing this, can't you just borrow the code from there?

@FaheemMitha As I've said, I think in the TeX world GitHub and thus Git is the most sensible choice

@DavidCarlisle I know you mentioned auctex David, but I was hoping someone who already knows the solution (and doesn't need to refer to auctex) would be able to answer ... then I wouldn't need to try to understand emacs lisp ...

@JosephWright sure. Wasn't arguing against that. Just saying I dislike the Git UI.

@FaheemMitha Far enough: I probably would go for Mercurial on that basis too
@StefanKottwitz Thanks for e-mail re. servers: need time to digest and pursue UK-TUG committee (and solutions in the UK)

5:32 PM
@JosephWright Well, we agree on that, at least. :-)

Hello @PhilMolyneux

@Joseph Hi - Just discovered the chat part of StackExchange following cfr's course earlier this week

@PhilMolyneux Cool: was going to ask if you saw the e-mail
@PhilMolyneux Obviously 'yes'
@FaheemMitha, @SeanAllred Perhaps I'll try to get some key arguments together (with Will) and make the case later in the year, at least for LaTeX3

@Joseph - since cfr wants to do a second course we should put up the other course materials

@PhilMolyneux Yes
@PhilMolyneux We got stalled (again)!
@PhilMolyneux @NicolaTalbot might have some useful input, I suspect
@PhilMolyneux We've got some sections, just need some content!
@FaheemMitha I hope you'll like the fact that the UK-TUG training stuff is in a DVCS, even though it's Git: github.com/uktug/latex-beginners-course :-)

5:42 PM
@JosephWright DVCS is good. :-)
The last one seems to have been in 2013. Are you having one this year?

@FaheemMitha Possibly
@FaheemMitha I probably need to raise this with the committee and the 'team'

@JosephWright Ok. 10 pounds is very cheap, though.

@FaheemMitha The team: @NicolaTalbot, @PhilMolyneux, me and a guy called Jay if he's well
@FaheemMitha Heavy subsidy
@FaheemMitha Venue cost us £300 originally, but now £600 as we no longer get a discount :-(
@FaheemMitha For students, even that can be too much as they have to pay travel

5:58 PM
@JosephWright true, but you should try and filter it in some way. So you don't just get anyone. How many days is this course for?

@FaheemMitha Just the one
@FaheemMitha We did want to get people to come!
@FaheemMitha I'll have to see what the committee want to do

@JosephWright sure, I just meant try to select people who might benefit, rather than some random person who won't know or care.
You could set up a little test on the web or something...
If they don't know anything, they probably aren't going to get much from a one day course.

6:20 PM
@FaheemMitha magit is magical. I recommend it as a git client even for those that don't use emacs.

@SeanAllred the emacs git interface?

@Szabolcs vvv
@FaheemMitha Oh yes, the best.

@SeanAllred That's probably meant to be CL .

@JosephWright If you send me what you think may be of concern with the transition, I'll try to put together solution sketches. You know where to find my email :)
@FaheemMitha Posh.

@SeanAllred Victoria Beckham?

6:24 PM
@FaheemMitha ? 'posh' as in 'nonsense'

@SeanAllred joke, attempt at.

@SeanAllred not a meaning I'm familiar with:-)

@DavidCarlisle I more-or-less made it up this week.
@DavidCarlisle Does it have a standing meaning?

@SeanAllred port out, starboard home, the upper classes

(it sorta morphed from 'psh' as an onomatopoeia)
@DavidCarlisle That sort of makes sense: urbandictionary.com/define.php?term=posh Maybe I just picked it up from someone. Funny how language can act as a virus.

6:27 PM
@SeanAllred it would never be used in that sense in the UK

@FaheemMitha me = vvv

I have \surroundwithmdframed[linewidth=2pt, backgroundcolor=LightGray]{tabular}. Can I get a lighter gray? And what colors does mdframed recognize?

@DavidCarlisle Because the illness has not yet jumped the pond ;)
@FaheemMitha Anything you can use with xcolor

@SeanAllred that's obviously ST, and that might be the android Data. But that's just a guess.
Whoever he is. he doesn't look happy.

@FaheemMitha 'tis indeed

6:28 PM
@FaheemMitha color=LightGray!50!white

the all-powerful mixin

@yo' Thanks. Increase or decrease 50 to make it lighter?

@FaheemMitha I'm never sure, try it :)

@yo' ok

@FaheemMitha Always best to ask the system

6:31 PM
@SeanAllred you are probably thinking of pish, which indeed has the meaning you evidently intended, albeit it is an archaic exclamation.

@FaheemMitha Ah, that works :D I like old words
@FaheemMitha I feel like I'm bringing sexy back

Though the original meaning of "pish" was just an expression intended to express or contempt or disgust. So not meaning rubbish, exactly.

Oh. :(

@SeanAllred or 'bosh' which does in fact mean 'nonsense'. :-)

@JosephWright b/p aren't too dissimilar

6:37 PM
@PaulGessler tosh does.
I was not aware of bosh.
nope, not finding that meaning.
tosh is an old-fashioned word meaning nonsense. Almost certainly of English origin.
I take that back, apparently bosh means that too.
Apparently both bosh and tosh also mean line breaks. Bringing this discussion neatly back on topic.

@FaheemMitha I was told the only things strictly on-topic in this chatroom were ducks and cricket. :-)

@SeanAllred oh, yes, those too! :-)

6:55 PM
@PaulGessler sometimes football too.

Hi guys! Is there 64-bit version for TeX Live?

@Cortizol which OS do you use?

@JosephWright @DavidCarlisle is ltnews22 really still supposed to say May 2015?

@PaulGessler It's for my professor. I think windows 7.

@Cortizol 64-bit binaries are not distributed for Windows by default
@Cortizol but see tex.stackexchange.com/a/227731/21344 if you really want them for some reason. Not a lot of benefit to getting them as far as I can tell.

7:07 PM
@Cortizol Why would you need 64-bit, anyway?

@SeanAllred The same question is wondering my professor, but in his office faculty changed all "old" computers with new one. So, it wasn't his idea.
@PaulGessler Thanks. I will see that.

@Cortizol Don't get me wrong -- 64-bit is very useful in many applications (especially graphical ones), but TeX isn't such an application

@Cortizol you can run 32-bit TeX Live with no problems at all on 64-bit Windows.

You can easily run 32-bit programs on 64-bit machines.

Oh, if that is case than splendid! I will install 32-bit.

7:23 PM
@Cortizol you could build it yourself.

has this type of question ever been answered? tex.stackexchange.com/questions/236688/…
I think it is (though perhaps not well asked) quite a puzzling little detail of LaTeX's inner working

@FrankMittelbach it knows to be PITA. I'm not sure it has been asked before.
I'm always puzzled why \vbox to \textheight{...} doesn't fit in one page.

@yo' yes nice one isn't it but it fits on the second or kind of

@FrankMittelbach well, it's pushed into the second page

@yo' sure because of the legal breakpoint on page 1
but that is still surprising
and yes I know too

7:39 PM
@FrankMittelbach yeah

@FrankMittelbach I think to have an answer for a similar problem.

@egreg the answer for that one is simply the mark at the beginning of a LaTeX document but it si kind of weird if you run into it unaware

@FrankMittelbach Here it is
5

I can reproduce your problem if the figure is too high to fit in the remaining part of the page after the chapter style. There's nothing you can do about it other than reducing the figure height or letting it float. Here's an explanation. After a chapter title, LaTeX disallows a page break. The...

@egreg well it is not the vertical space as such it is the \mark before it that initializes the running header info

@FrankMittelbach Is a \mark a legal break point?

7:51 PM
@FrankMittelbach hmm I guess 1st Jan would have been more consistent..

@egreg not it is not so what I'm saying is rubbish

@FrankMittelbach :)

@egreg but neither is the any space vanishing afterwards ...
its the break point vanishing

@FrankMittelbach The only legal break point is the space or penalty before \chapter; this vanishes at the break; then on the second page no break point is available.

8:07 PM
@egreg chapter uses \vspace* and that doesn't vanish -- it is the \topskip space that produces the problem and because of the mark or the \write-{} that one becomes a breakpoint
\begin{document}
\vbox to 1.2\textheight{}
\end{document}
@egrep while this doesn't show the problem:
\begin{document}
xxx
\eject
\vbox to 1.2\textheight{}
\end{document}
@DavidCarlisle so? change it Jan then?

@FrankMittelbach Yes, that's the vertical glue that is the only feasible break point.

@FrankMittelbach yes I guess so

@egreg so it is because of the way LaTeX uses \write not \mark (I wonder if that got changed or whether I only dreamed it)
@DavidCarlisle ok then I'm going to update the project pages
@DavidCarlisle well the combined one on CTAN will still say May for a while but ...
one can't expect us to get those things right ... can one?
@DavidCarlisle arrg inside it gives /06/01 as an example ... not really wrong but ...
@DavidCarlisle well maybe ok even though there will never ever be a 06/01 format now

8:28 PM
@FrankMittelbach I'm not sure it makes a difference: if the \topskip glue is preceded by either a \mark or a \write it becomes a feasible break point.

@egreg technically no but the explanation is slightly different then. anyway I wonder if there shouldn't be a nobreak following that \write in \document

@FrankMittelbach That might be good, so the first \topskip is not a break point (unless something else comes along, of course).

@FrankMittelbach I was just looking at latex2.09 earlier:
% \begin{macro}{\mark}
%    \LaTeX~2.09 initialized an empty mark.  Who knows, someone may have
%    relied on it:
%    \begin{macrocode}
\mark{{}{}}
%    \end{macrocode}

@DavidCarlisle so that's what I remembered ... old fond memories :-)

@FrankMittelbach that was an example of a post latex package update date

8:35 PM
@DavidCarlisle yes I eventually realized
I'm slow these days (and rusty)

@FrankMittelbach but we're not as old as @egreg

@DavidCarlisle doesn't help if he is less rusty

@FrankMittelbach :)

@FrankMittelbach perhaps but we can probably do an update before tL pretest done.

Hi, I noticed my question from over half a year ago is still untouched: tex.stackexchange.com/questions/202242/…. So can I conclude from this that it's hard to do / typographically not sound / or otherwise?
Okay, untouched is a pretty big word.

8:56 PM
@1010011010 I'd probably start from @FrankMittelbach's answer here
25

The multicol environment is not designed to support column floats. The concept of balancing makes this next to impossible to automatically provide correct results in the general case and therefore I decided not to extend multicolin this direction for 2e. For example, with multicol you can change...