Yoshiki Hayashi <t90553(a)mail.ecc.u-tokyo.ac.jp&gt; writes: I've got no reply, so I tried to fix it myself. ;-) The bug is in determine_real_coding_system. It doesn't handle 8bit character correctly. Although I have not fixed the real bug, I fixed some bug with this patch and it makes XEmacs crashes less often. What this patch fix is: 1. Search if -*- exists. 2. Stop searching at the first \r or \n 3. Don't treat encoding: or .*coding: as 'coding: cookie' The problems left are: 1. It doesn't check if closing -*- exists or not. 2. EOL detection is not working well when you don't specifiy it explicitly. (i.e. coding: iso-2022-jp assumes LF as EOL char.) 3. It doesn't fix crash. You can still crash it with files like this. 1999-10-26 Yoshiki Hayashi <t90553(a)mail.ecc.u-tokyo.ac.jp&gt; * file-coding.c (determine_real_coding_system): Check if '-*-' exists. End searching for coding: cookie at the end of line. Check whether character before coding: is space, tab or ';'. ;;; -*- mode: emacs-lisp; syntax: elisp; coding: ß -*- -- Yoshiki Hayashi