@JosephWright Vafa mailed me a bug report from a bidi user, but in bidi he was stuffed really as etex.sty got loaded early enough that it switched out the new allocation then tikz, fontspec and bidi between them used up all the registers....
@JosephWright \newinsert bites again:-) the "old" way would be for the user to use \reserveinserts before loading the packages but it's not that clear to an end user that etex.sty is loaded at all and of course they can't use \reserveinserts unless they load etex.sty themselves first so just getting tkz-graph not to load etex is best,
@JosephWright Yes, I guess it may be discussed, but basically the OP included his answer (i.e. changing the path) into that "answer post", so it might qualify as an answer. If you deleted it again, you should rollback my rollback to the question though.
Do we have a proper test on whether we are in the preamble or not? Have a macro what can be used in preamble and body. But if used in the preamble it can trigger a strange bug in tikz unless I disable a feature in the macro when used in the preamble (this feature is only relevant when used in the document body).
@DavidCarlisle that seems to work, thanks. This is a strange bug between \SingleSpacing (which contains a \vskip at the end) and tikz where under some circumstances an \unskip is emitted. Now I'll only emit the vskip in the body instead and even add a starred version to remove it in the doc as well.
@DavidCarlisle — I'm in a bind about "designing" the ctan-post interface. This looks nice enough but I don't know whether to turn it into a supported part of l3build yet: github.com/wspr/breqn/blob/master/build.lua
@WillRobertson I suppose the last line translates as "if it's at all wrong, blame David" ?
@WillRobertson I see you've automated the version string which is nice, any thoughts about a good place to set the announcement text?
oh hang on what's currentchanges...
ohh OK I see @WillRobertson
@WillRobertson @JosephWright I suppose some documentation/examples could suggest some possible filter functions to go with changelog formats, leaving the default being as now to prompt at the terminal, or as Joseph suggested fire up $EDITOR ?
@WillRobertson I might experiment this evening as I'm probably going to have to upload most of these: github.com/tkz-sty so automating might make sense...
@DavidCarlisle — yeah in terms of announcements I'm too lazy to write anything, I just suck in the most recent changes.md text. I mean, I could also put in a blurb there, which would be good. I wouldn't (at this stage) look to enforce that part of the workflow on anyone, neither using changes.md to provide the version number.
originally (and internally) I had a ctan-post config table but when I pulled it in to l3build I exposed all the fields as top level variables (as all existing l3build settings are that way) but you built a table pulled the fields out to top level variables which then get out back into a table to pass to the post function. That may not be the optimal code path, but maybe it's OK...
@WillRobertson yes it's good to have examples of that though, so if I run a script to upload all 19 repos in rf-tex (or the 9 in tkz-sty) you can do it without it stopping and asking you to write something each time...
@DavidCarlisle That was one of my concerns, yes. As I wrote it it seemed more and more natural to have actual package metadata in a table at the top. If that seemed like a good idea in general, I would guess having a ctan table that maps some of those would seem to be better. OTOH I like your simplicity of lots of simple variables.
@DavidCarlisle @WillRobertson The original settings in l3build are 'flat' largely as that made sense given the original aims. I think I'd still do that now: I don't fancy having to write build.sourcefiledir or whatever. But for something like the complexity of CTAN upload, a single table is likely the right approach
@WillRobertson I think a table is probably best: we can do ctan.foo = ctan.foo or <name> somewhere in the upload code
@WillRobertson, @DavidCarlisle One of the team meeting?
@DavidCarlisle, @WillRobertson What I would like to hammer out is the interface for l3build and uploads, i.e. do we want l3build upload or l3build ctan --upload or ...
@JosephWright while driving in I decided that l3build ctan --upload would probably be best as otherwise the existing ctan command seems strangely named. But I don't want the default to upload so having it do what it does now unless --validate or --upload option given seems good to me
@DavidCarlisle Hopefully with a pause in there so it does the (lengthy) build of the zip file, then pauses and says "go check the actual zip file that you're happy" then continues with the upload. (Similar to your current interface of asking to proceed.)
@DavidCarlisle Yes, though we could simply rename ctan to something else. I'd imagined having a flag in the build.lua that is false as standard but can be forced using --upload ...
@daleif Actually that's not an example of quirks about xparse, but it's really a feature: with the *-variant, a space is added at the end of the paragraph instead of \par.
@DavidCarlisle But we don't have such a separation between build ctan and build check — if we wanted to decouple the building and the upload here it wouldn't make as much sense to keep build and check together. Well, I guess that's why we're talking about it :)
@WillRobertson you mean after waiting hour and a quarter for oberdiek bundle to build and ctan upload rejects the upload because Heiko's description text is longer than 3000 characters, I should build it all again so I can write a smaller description?
@JosephWright @WillRobertson (although despite that I do always build from scratch, I don't trust zip files that are just lying around) so possibly you are right.
@WillRobertson if we use make instead of l3build then if no sources have changed, rebuilding is a no-op (or is that a politically incorrect observation)
@JosephWright I suspect that apart from us and Heiko build times for latex packages are not that great so it's probably not so much of an issue as it seems
@JosephWright but talking of oberdiek bundle reminds me it isn't currently built via l3build but it's still possible to use the lua upload to post to ctan, so it can be useful to have an upload only option (even if the defaults try to steer people to rebuilding)
@DavidCarlisle ^^^^^ woow, do you even help the last day of the year? That's awesome :). In 2012 I didn't know Word lol, and you earned a gold badge... good job (and for the rest congrats too!)
@manooooh I might have posted that day or I might not:-) (you could check the records) After you post a few answers you earn rep and the occasional badge most days even if you do not use the site.
Hi all I have MiKTeX Version 2.9.6643 at my computer so I can I am generating pdf reports and everything works fine. Today I installed the new version of MiKTeX (2.9.6813) and all the text instead of having black font colour now is white. The tex file is exactly the same with not change. At my co...
@DavidCarlisle on other news, the luaotfload import scripts seems to work, I had to adapt them, as the set of files is different today but now it looks good. I guess in one or two days (if nothing unterrupts me or some bad error appears) I can sent you something to test. Side remark: Today I had to install the 64bit luatex to be able to handle a cjk font, the first time that 32/64bit made a difference for me ;-)
@DavidCarlisle and @daleif -- good question. the first thing to do, i think, would be to identify the items that can be moved from one to the other, without requiring any "workaround" code, i.e., items which simply act directly on the existing amsmath definitions. i think i may be able to identify many of these from the ams bugs list, and will try to do so very soon.
@JosephWright The 64bit-problem was triggered by this question tex.stackexchange.com/questions/451442/…. It probably involves tagging or at least marked content literals and as the OP uses xelatex it needs the interface to \pdfliteral etc ;-).
@UlrikeFischer I'm using Alexander Grahn's code in pdfbase plus what Till did in pgf, and trying to find a common interface suggestion. Plus pushing for some new hooks in 2e: that's part of a more general overhaul
@UlrikeFischer I should ask if you can pop in to one of the team meetings ;)
I found a new way to make bibtex get mad at me! Have @article{Mine in your bib file. Then have both \cite{mine} and \cite{Mine} in your document. It REALLY doesn't like that type of typo
Which seems odd to me that it doesn't just go 'yeah, no idea what that citation is' like if I made a typo like \cite{Minee} when I meant \cite{Mine}
Also, recommendations for a command line tool that works like online2pdf.com ? Or is there one?