bug#81648: 32.0.50; Running LaTeX on files with non-ASCII chars in the name

Eli Zaretskii <[email protected]> writes:

>> From: Arash Esbati <[email protected]>
>> Cc: [email protected],  [email protected]
>> Date: Tue, 18 Aug 2026 15:19:01 +0200
>> 
>> I'm not familiar enough with the internals of tex-mode.el and what
>> happens after hitting "C-c C-c RET", but after having another look, I
>> think that the current invokation:
>> 
>>   pdflatex  \\nonstopmode\\input \ä.tex
>> 
>> should be:
>> 
>>   pdflatex  \\nonstopmode\\input \\detokenize{ä}.tex
>
> I know next to nothing about LaTeX, so can you please explain what
> does detokenize do and why do you think it is needed in this case?

I'm also not familiar with LaTeX's internals, so please take this with a
big grain of salt.

Currently, the invokation does:

  pdflatex \\nonstopmode\\input \ä.tex

I'm not sure how the UTF-8 support of LaTeX is implemented, but I think
it turns ä into an active char like ~ (for unbreakable space), i.e., you
use a TeX macro without a backslash.

Since the invocation above uses \input without braces, the TeX primitive
\input is used which doesn't grok the input.  \detokenize helps now
which is defined:

  \detokenize <general text>.

  The expansion is a list of character tokens representing the token
  list <balanced text>.  As with the lists of character tokens produced
  by TeX’s \the and e-TeX’s \readline, these tokens have category 12
  (‘other’), except that the character code 32 gets category 10
  (‘space’).

And with that, \input understands which file to open.  The difference in
a shell is:

  pdflatex \\nonstopmode\\input ä.tex

throws an error.

  pdflatex \\nonstopmode\\input \\detokenize{ä}.tex

works, and

  pdflatex \\nonstopmode\\input{ä}.tex

works as well.  \detokenize in action seems to be this:

produced with this code:

--8<---------------cut here---------------start------------->8---
\documentclass[12pt,a4paper]{article}

\usepackage[T1]{fontenc}
\usepackage{lmodern}

\def\ä{Foo}

\begin{document}
\noindent
\begin{tabular}{lcl}
  \verb|\detokenize{ä}| & $\mapsto$ & \texttt{\detokenize{ä}} \\
  \verb|\detokenize{a}| & $\mapsto$ & \texttt{\detokenize{a}} \\
  \verb|\detokenize{\foo}| & $\mapsto$ &  \texttt{\detokenize{\foo}} \\
  \verb|\ä| & $\mapsto$ & \ä             
\end{tabular}

\end{document}
--8<---------------cut here---------------end--------------->8---

IIRC we had to deal with something similar within AUCTeX and \detokenize
was the solution.

HTH.  Best, Arash