T-F-S
commented
1 year ago Once again parbox=false problems... :-(
I think it's wrong to leave
\if@newlisttrue inside a tcolorbox, no matter/tcb/parboxis true or false. But I'm unsure if adding\global\@newlistfalseto\tcb@lrboxis the right way.
Currently, I see no better way as to add \global\@newlistfalse to \tcb@lrbox and its breakable counterpart \tcb@vbox. Also, I would add \let\par\@@par. This gives the following:
\documentclass{article}
\usepackage[breakable]{tcolorbox}
\makeatletter
% lrbox with integrated minipage
% #1 box
% #2 width
% #3 color (for color stack)
\def\tcb@lrbox#1#2#3{%
\edef\reserved@a{%
\endgroup
\setbox#1\hbox{%
\begingroup\aftergroup}%
\def\noexpand\@currenvir{\@currenvir}%
\def\noexpand\@currenvline{\on@line}}%
\reserved@a
\@endpefalse%
\global\@newlistfalse% +++++++++++++
\let\par\@@par% +++++++++++++
\let\tcbbreak\par%
\csname tcb@parbox@use@\kvtcb@parbox\endcsname%
\tcb@minipage{#2}%
\color{#3}%
\tcb@hyph@fix\ignorespaces}
\let\endtcb@lrbox=\endminipage
\let\tcb@savebox=\tcb@lrbox
\let\endtcb@savebox=\endtcb@lrbox
\def\tcb@vbox#1#2#3{%
\edef\reserved@a{%
\endgroup%
\setbox#1\vbox{\hsize=#2%
\begingroup\aftergroup}%
\def\noexpand\@currenvir{\@currenvir}%
\def\noexpand\@currenvline{\on@line}}%
\reserved@a%
\@endpefalse%
\global\@newlistfalse% +++++++++++++
\let\par\@@par% +++++++++++++
\let\tcbbreak\tcb@@break%
\iftcb@usecolorstack%
\pdfcolSwitchStack{tcb@breakable}%
\fi%
\color@begingroup%
\textwidth\hsize%
\columnwidth\hsize%
\csname tcb@parboxrestore@\kvtcb@parbox\endcsname%
\def\@mpfn{mpfootnote}%
\def\thempfn{\thempfootnote}%
\c@mpfootnote\z@%
\let\@footnotetext\@mpfootnotetext%
\let\@listdepth\@mplistdepth \@mplistdepth\z@%
\@setminipage%
\color{#3}%
\tcb@hyph@fix%
\let\tcb@drawcolorbox\tcb@drawcolorbox@standalone%
\let\FN@pp@footnote\@empty% disable perpage mode of 'footmisc' package
}
\makeatother
\begin{document}
\makeatletter
\def\@par@inlist{% set by \@setpar in \@trivlist
\if@newlist
\advance\par@deathcycles \@ne
\ifnum \par@deathcycles >\@m
\@noitemerr
{\@@par}%
\fi
\else
{\@@par}%
\fi}
\def\debugInfo{ %
{\ttfamily
<@newlist = \if@newlist T\else F\fi,
par \ifx\par\@par@inlist= \else!= \fi par@inlist>%
}%
}
\makeatother
\LaTeXe: \fmtversion,
tcolorbox: \UseName{tcb@version}
\begin{enumerate}
\item abc \par def \par ghi\debugInfo
\end{enumerate}
\begin{enumerate}
\item
\begin{minipage}[t]{10cm}
minipage, abc \par def \par ghi\debugInfo
\end{minipage}
\end{enumerate}
\begin{enumerate}
\item \parbox[t]{10cm}{parbox, abc \par def \par ghi\debugInfo}
\end{enumerate}
\begin{enumerate}
\item
\begin{tcolorbox}[title={\texttt{parbox=true}}]
abc \par def \par ghi\debugInfo
\end{tcolorbox}
\end{enumerate}
\begin{enumerate}
\item
\begin{tcolorbox}[title={\texttt{parbox=false}, first in a list}, parbox=false]
abc \par def \par ghi\debugInfo
\end{tcolorbox}
\begin{tcolorbox}[title={\texttt{parbox=false}, second in a list}, parbox=false]
abc \par def \par ghi\debugInfo
\end{tcolorbox}
\end{enumerate}
\end{document}parbox=false is a patchwork which I am not so very happy with... OK, fingers crossed that no serious side effects happen.
muzimuzhi
(originally reported as part of https://tex.stackexchange.com/q/695327, the TeX-SX question mentioned in #244)
If the content of a list (environment) starts with a
parbox=falsetcolorbox, then the\paris ignored in this tcolorbox.Analysis
Inside a list,
\paris redefined as (set by\@setparin\@trivlist)Hence
\parwill be silently ignored when\if@newlistis true. Normally\if@newlistis set to false globally when the first paragraph inside a list starts, by the special\everyparset by\@item.To fix #171,
\noindentis moved from\tcb@parbox@use@false(actually\tcb@parbox@false@settings) to a latter place, at the beginning of hooks likebefore upper, see commit 20805e947fa5453f2ada5a26f457c537515062df (shipped with v5.1.0). Unfortunately the\everypartoken list use by these two\noindents are different, and this resulted in a regression.In
\tcb@lrbox,\kvtcb@parboxis false, line 1026 (see code snippet below) expands to\tcb@parbox@use@falsewhich indirectly contains\noindent. When such\noindentis executed, the\everyparin use is still the one set by outer list, hence\if@newlistis set false.\noindentanymore, and line 1027 will expends to (sth starting with)\minipage, which will set\everyparto just\@minipagefalse\everypar{}, in\@setminipageat the very end of\@iiiminipage. Then, when thebefore upperhook executes\noindent, the\everyparused at that time never sets\if@newlistfalse. Therefore\pars used in a such tcolorbox will be ignored.https://github.com/T-F-S/tcolorbox/blob/6d98db4263c612279cca47efde22d1ef8404e573/tex/latex/tcolorbox/tcolorbox.sty#L1012-L1029
An extended example showing some debug info
```tex \documentclass{article} \usepackage{tcolorbox} \begin{document} \makeatletter \def\@par@inlist{% set by \@setpar in \@trivlist \if@newlist \advance\par@deathcycles \@ne \ifnum \par@deathcycles >\@m \@noitemerr {\@@par}% \fi \else {\@@par}% \fi} \def\debugInfo{ % {\ttfamily <@newlist = \if@newlist T\else F\fi, par \ifx\par\@par@inlist= \else!= \fi par@inlist>% }% } \makeatother \LaTeXe: \fmtversion, tcolorbox: \UseName{tcb@version} \begin{enumerate} \item abc \par def \par ghi\debugInfo \end{enumerate} \begin{enumerate} \item \begin{minipage}[t]{10cm} minipage, abc \par def \par ghi\debugInfo \end{minipage} \end{enumerate} \begin{enumerate} \item \parbox[t]{10cm}{parbox, abc \par def \par ghi\debugInfo} \end{enumerate} \begin{enumerate} \item \begin{tcolorbox}[title={\texttt{parbox=true}}] abc \par def \par ghi\debugInfo \end{tcolorbox} \end{enumerate} \begin{enumerate} \item \begin{tcolorbox}[title={\texttt{parbox=false}, first in a list}, parbox=false] abc \par def \par ghi\debugInfo \end{tcolorbox} \begin{tcolorbox}[title={\texttt{parbox=false}, second in a list}, parbox=false] abc \par def \par ghi\debugInfo \end{tcolorbox} \end{enumerate} \end{document} ```
I think it's wrong to leave
\if@newlisttrue inside a tcolorbox, no matter/tcb/parboxis true or false. But I'm unsure if adding\global\@newlistfalseto\tcb@lrboxis the right way.For a
minipageas the first element inside a list,\leavevmodeat the beginning of\@iiiminipagestarts a new paragraph, hence sets\if@newlistfalse\let\par\@@parin\@arrayparboxrestore(the latter is then in\@parboxrestore) restores\parto its normal definition outside a list.