Denis Bitouzé <denis.bito...@univ-littoral.fr> writes: > several years ago, I already faced the following problem and, > unfortunately, it happened again yesterday, which made me lose quite > some time. > > Let me explain myself: I had a LaTeX file encoded in latin1 that > I wanted to encode in UTF-8. I used an external tool, in this case > `utrac`,
I think Emacs got upset because it saw you used an external tool -- hence the punishment 😉 > which confirmed the starting (latin1) and ending (UTF-8) > encoding. But, when I opened this file in Emacs with AUCTeX enabled, > the accented characters were wrong and it was only when I saw that the > file contained `usepackage[latin1]{inputenc}` that I understood where > the problem came from: changing it in `usepackage[utf8]{inputenc}` > solved it. I'm not even sure that AUCTeX has code to deal with a .tex file which already contains \usepackage[<enc>]{inputenc}. The only thing I'm aware of is when you type 'C-c C-m usepackage RET inputenc RET ENC RET' that AUCTeX changes the file encoding acc. to chose ENC. > So, here is my request: would it be possible that, for the detection of > the real encoding of the file, AUCTeX relies not on the `inputenc` > package option, but rather on the Emacs heuristics and that, in case of > discrepancy between the two, it issues a warning? Have a look at the variable `file-coding-system-alist'. If you don't want automatic conversion based on 'inputenc', remove the entry ("\\.\\(tex\\|ltx\\|dtx\\|drv\\)\\'" . latexenc-find-file-coding-system) Maybe that helps. Best, Arash