nested minipage highjacks the footnote

[u-fischer]

Reference: https://tex.stackexchange.com/q/504688/2388 (here with todonotes)

A nested minipage grabs footnotes from the outer minipage:

\RequirePackage{latexbug}
\documentclass[12pt]{article}
\begin{document}
\section{nested}
\begin{minipage}{6cm}
some text\footnote{footnote A}

\hspace*{3cm}%
\begin{minipage}{3cm}
 nested
\end{minipage}

some text \footnote{footnote B}
\end{minipage}
\end{document}

Log file (required) and possibly PDF file

test-utf8.log

This can be surprising for the user if the inner minipage is actually hidden, e.g. in a tikzpicture:

\documentclass[12pt]{article}
\usepackage{tikz}

\begin{document}
\begin{minipage}{6cm}
some text\footnote{footnote A}

\hspace*{3cm} \begin{tikzpicture}
 \node[draw=red,text width=3cm,align=left]
 {nested};
 \end{tikzpicture}

some text \footnote{footnote B}
\end{minipage}
\end{document}

[PhelypeOleinik]

Not only the footnote goes to the wrong place, but also the inner minipage inherits the width of the outer one, resulting in an overfull box.

Adding \setbox\@mpfootins\box\voidb@x to the initialisation of the minipage to ensure that the insertion box starts empty (thus, only footnotes added to the current minipage are seen) seems to solve the problem:

\makeatletter
\def\@iiiminipage#1#2[#3]#4{%
  \leavevmode
  \@pboxswfalse
  \setlength\@tempdima{#4}%
  \def\@mpargs{{#1}{#2}[#3]{#4}}%
  \setbox\@tempboxa\vbox\bgroup
    \color@begingroup
      \hsize\@tempdima
      \textwidth\hsize \columnwidth\hsize
      \@parboxrestore
      \def\@mpfn{mpfootnote}\def\thempfn{\thempfootnote}\c@mpfootnote\z@
      \let\@footnotetext\@mpfootnotetext
      \let\@listdepth\@mplistdepth \@mplistdepth\z@
      \setbox\@mpfootins\box\voidb@x% <=================================
      \@minipagerestore
      \@setminipage}
\makeatother

\RequirePackage{latexbug}
\documentclass[12pt]{article}
\begin{document}
\section{nested}
\begin{minipage}{6cm}
some text\footnote{footnote A}

\hspace*{3cm}%
\begin{minipage}{3cm}
 nested
\end{minipage}

some text \footnote{footnote B}
\end{minipage}
\end{document}

---

The outer minipage starts with \@mpfootins empty, and then the footnote A adds stuff to it. When the inner minipage ends, it checks if \@mpfootins is empty, which is not because of footnote A, then it typesets the footnote at the bottom of the inner minipage. Thus initialising every minipage with an empty \@mpfootins should avoid that one minipage sees the contents of the other.

[u-fischer]

@PhelypeOleinik your code solves my example, but if the inner minipage has also a footnote one looses the first:

% patch as a above ....
\documentclass[12pt]{article}
\begin{document}
\section{nested}
\begin{minipage}{6cm}
some text\footnote{footnote A}

\hspace*{3cm}%
\begin{minipage}{3cm}
 nested\footnote{footnote C}
\end{minipage}

some text \footnote{footnote B}
\end{minipage}
\end{document}

[PhelypeOleinik]

@u-fischer Oh :/

The code for \@mpfootnotetext globally sets \@mpfootins, so my local \setbox\@mpfootins\box\voidb@x won't work. I suspect that the reason for that assignment to be global is that both footnotes A and C are labeled "a", then things get confusing.

Changing \global\setbox\@mpfootins to \setbox\@mpfootins in the definition of \@mpfootnotetext, plus the patch above seem to solve the second problem as well. However, having the global assignment will probably require something more elaborate to detect the nested minipages...

[car222222]

Maybe nested minipages were never supported (at least not with footnotes everywhere!) so this is simply a documentation problem?

But a package extending coverage to “recursive minipages” may be a good idea. Good for making pages that contain reduced copies of themselves, I guess.

[blefloch]

Maybe replace

\setbox\@mpfootins\box\voidb@x

in Phelype's code by

\setbox\mybox\box\@mpfootins

(with a better name for the box than \mybox, of course) and in the \end{minipage} code do

\global\setbox\@mpfootins\box\mybox

[FrankMittelbach]

This is a documented restriction: LaTeX Book page 106, so it works as designed. That does not mean that we can't change it. However, I'm not sure that it is correct that a tikz picture is internally using a minipage instead of a parbox ... is that supposed to support inner footnotes?

[car222222]

I should have RTFM :-).

[u-fischer]

@FrankMittelbach I don't know why tikz or todonotes (which triggered the issue) use minipage. It could be footnotes, but perhaps simply as environments are easier to handle in begin/end code. I never saw a question about the side-effect before and even if one know in theory that a single mp-footnote counter and insert means that nesting could be problematic it is easy to forget that in practice.

[car222222]

So it would be a good idea to at least issue a warning when a nested minipages us spotted. Not that anyone will notice a warning, of course.

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity.

[FrankMittelbach]

the above commit is a red herring it should have referenced #186

[stale[bot]]

This issue has been automatically marked as stale because it has not had recent activity.

[FrankMittelbach]

I have now added a check and a warning if minipages are nested.