@DavidCarlisle -- You may remember that I objected to this when it was changed: tex.stackexchange.com/q/548602. Too late to do anything now.

@PhelypeOleinik -- The conversion has started. In the example I just posted, I just entered the url, but it reports the title. So it may be useless to add any more examples. But maybe time to start checking.

@barbarabeeton ? that is user error with or without the change. the change only got mentioned as he had an old release so needed to add the negative space as well as fixing the markup

@barbarabeeton ooh we sail the ocean blue! I know part of that musical thingy

@JosephWright @UlrikeFischer strangely the bbc doesn't blame you for bricking people's phones with colour space conversions bbc.co.uk/news/technology-52891650

@DavidCarlisle ooh

@DavidCarlisle at some time if will blame us for filling people's phone with pdf's ;-)

@UlrikeFischer that is definitely your fault:-)

@DavidCarlisle ooh

@DavidCarlisle ;-). I got a message from Thanh over Karl about a pdftex change. If it breaks pdftex one can blame too ;-)

@DavidCarlisle :)

@UlrikeFischer Change?

@JosephWright It is about the fatal error if a link ends on a different box listing. tug.org/pipermail/tex-live/2018-November/042759.html. It should be converted in a warning - the area is then perhaps wrong but you get at least a pdf. Karl seems to have added tests also for my other requests: tug.org/svn/pdftex/branches/stable/tests. If they make changes we need to convince luatex to add them too.

#ShutDownAcademia #ShutDownSTEM

@UlrikeFischer Ah, right, I see

If I use \begin{figure}[p], the figure goes on a separate page at the end of the document. Can I put it on a page right here, not at the very end?

@Szabolcs perhaps using h?

I would like to put the figure on its own, separate page. But I'd like that page not to be at the very end of the document. I would like it to be around the position where the \being{figure} is. Is this possible?

I usually go with htbp (here, top, bottom, page)

@Szabolcs Oh sorry, I got it wrong. :)

Is it even normal that it goes to the end? If not, could someone just tell me so I can avoid making a minimal example ... ?

This is the point where I would be quite upset with LaTeX if I didn't have to use MS Word just two weeks ago ...

@Szabolcs You could try with \renewcommand{\textfraction}{1.0} in the document preamble. As far as I know, pages with text and floats are not allowed then. All floats should go on a page of their own as soon as possible. Maybe this works for you.

But it's only some figures I need to put on their own page, not all of them

It is actually a begin{table}, not a figure. Sorry.

@Szabolcs it would normally go on the next page, you have something else forcing it to the end.

Can someone tell me if it is normal that as soon as I use p, it goes all the way to the end of the document? Reserving a page for it is such a simple thing, why does it have to be such a major pain in LateX? It's been 45 minutes now

@DavidCarlisle Thank you!

@AlexG if the float has [p] then \textfraction will not be consulted.

@Szabolcs latex keeps all your tables in order, so if one goes to the end then all the following ones will, so you need to check why the first one is not being placed (being too big for the page for example)

Basically, I'd like to force it to put my figures of choice together on the same page, not some figures it selects on its own.

@Szabolcs make \floatpagefraction bigger than the first figure, but if you want to force some figures to float together simply put them in the same figure environment.

@Szabolcs by default \floatpagefraction is 0.5 so if fig4 is > .5\textheight then page 4 will be shipped out without waiting for fig5

@DavidCarlisle @JosephWright if we can get primitives that do the same as the two xetex specials in the following example. How should they be named? Current suggestions are "\pdfnolink and \pdflink like xetex? or rather \pdfrunninglinkoff and \pdfrunninglinkon?"

\documentclass[a4paper]{article}
\usepackage{fancyhdr}
\pagestyle{fancy}
\cfoot{some longer text that should not be linked}
\usepackage{ifxetex}
 \ifxetex
  \cfoot{\special{pdf:nolink}some longer text that should not be
linked\special{pdf:link}}
 \fi
\usepackage{hyperref}

\begin{document}
\vspace*{48\baselineskip}
xxxxxxxxxxxxxxxxxxxxxxxxxxx xxxxxxxxxxxxxxxxxxxxxxxxxx
\href{xxxxx}{link link link link }
\end{document}

@DavidCarlisle, @UlrikeFischer Looking at the backend code again, I realise that I can gain performance in various places: when we made it all internal, we ruled out needing to worry about floating point values needing re-evaluating

@UlrikeFischer, @DavidCarlisle Expect some checkins later :)

@UlrikeFischer I'd go with \pdfnolink: as far as possible, similar names are good!

@JosephWright I'm always worried if new primitives are too short and to obvious. \pdflink has a few hits for example in texinfo.tex and cwebmac.tex

@UlrikeFischer Hmm

@barbarabeeton twitter.com/In_Otter_News2/status/1270106190270337025

@UlrikeFischer should we fix hyperref not to lose breakpoints?

@DavidCarlisle if you know where we are loosing them. Is it the whatsit of the anchor?

@UlrikeFischer I was just wondering whether to look when I saw your comment, yes it will be the whatsit but it could/should probably add a penalty0 or something if there was a breakpoint there, I may look later.

@DavidCarlisle A simpler example is this. But the question is naturally if it is a good idea to issue the \refstepcounter there. The anchor is then probably not where you want it.

\documentclass{article}

\usepackage{hyperref}  %
\textheight 3cm

\begin{document}
\begin{minipage}{5cm}
\rule{1cm}{2cm}
\end{minipage}

\refstepcounter{section}
\begin{minipage}{5cm}
\rule{1cm}{2cm}
\end{minipage}

\end{document}

@UlrikeFischer but nasser's example is an [H] float so the refstepcounter is inside the box isn't it?

@DavidCarlisle No it is outside. algorithm starts with some init code which does a \refstepcounter {algocfline} and then processes the figure (but there could be more counter inside the figure, I didn't check).

@UlrikeFischer ah OK, so we could blame algorithm (but might be easier to get hyperref to check...)

@UlrikeFischer \Hy@RestoreLastskip ->\protect \nobreak

@DavidCarlisle but only in \ifhmode or not?

@DavidCarlisle I didn't found that place yet. Where is it?

Ah inside \Hy@SaveLastskip. And where is that used?

in some cases we could replace the skip=0 test by testing that the etex last node type is a skip then explicit skips of 0pt would be seen, but here there is no skip as it is after the \par from the first minipage but before the next para starts so before \parskip is added, adding a breakpoint.

@UlrikeFischer tracingall says

@UlrikeFischer if the previous node is a non discardable item it should just add itself immediatedly after that then it won't change the breakpoints, by adding \nobreak there it is preventing the glue that is added later making a breakpoint.

Yes. So why is there an explicit \nobreak if \ifdim\lastskip=\z@?

@UlrikeFischer because it's wrong, I think.

@DavidCarlisle the dtx says

%    Inserting a \cmd{\special} command to set a
%    destination destroys the \cmd{\lastskip} value.

@UlrikeFischer I'll look later but I think with etex it can be made better as you know there is no glue there and you are immediately after the first minipage box, so you know that \lastskip would do nothing useful anyway.

@UlrikeFischer I know it's a bit modern to be suggesting using etex primitives.

@DavidCarlisle ;-). Any comment on the naming for the ever more modern primitive to all a link to stop?

@UlrikeFischer so the usage is \pdfstartlink ... stufff \newwthing no link here \newthingb resume link \pdfendlink ??

@DavidCarlisle yes. (I haven't asked yet what will happen, if there is a \pdfstartlink inside the \newthing stuff, with xelatex one gets nonsense. But even if nesting don't work it would be already quite an improvement.).

@UlrikeFischer actually the existing ones end in link so maybe pdfstoplink \pdfresumelink

@DavidCarlisle sounds good.

@UlrikeFischer oh edited, meant to give a second suggestion..

@DavidCarlisle yes that's fine to.

@DavidCarlisle I did run the hyperref tests with your change. There are three failures because of the missing penalties but all looks ok.

@UlrikeFischer was short of time but I want to think about it a bit more, it probably should be testing any discardable item not just glue, but even in the glue case I don't currently see why the nobreak is right