Re: Coding system detection: Emacs 23 vs. 22

[Kenichi Handa <handa@m17n.org>]

In article <jwvejahgdsv.fsf-monnier+emacs@gnu.org>, Stefan Monnier <monnier@iro.umontreal.ca> writes:

> > when opening the attached file (see attachment `ae.el.gz'), thee
> > coding system detection of Emacs 23 and 22 give different results:

> > - Emacs 22 (both, Emacs 21.1 and the current Emacs 22 branch) says
> >   emacs-mule-unix.  (The file was created with Emacs 22)

> > - Emacs 23, says iso-latin-1-unix

> > The real problem with this is that my Gnus score files of Emacs 22 no
> > longer work with Emacs 23 if the score entry (subject, author, ...)
> > contains a non-ascii character.

In Emacs 23, the charset definition for iso-8859-1 is taken
from glibc's locale data
(/usr/share/i18n/charmaps/ISO-8859-1.gz), and it contains
0x80..0x9F.  So, in the environment that prefer iso-latin-1,
most files are detected as iso-latin-1.

> > Expected behavior:

> > Emacs 23 should detect that the attached file is not iso-latin-1 (even
> > `file' detects this) but emacs-mule.

> I suspect this is related to bug #16
> http://emacsbugs.donarmstrong.com/cgi-bin/bugreport.cgi?bug=16

If gnus writes score files in emacs-mule, it must read that
file by explicitly specifying emacs-mule.  Or, it must write
score files with coding: tag.

The detection of emacs-mule is not 100% correct even in
Emacs 22.  If the problem didn't appear in Emacs 22, it is
just because of good luck.

---
Kenichi Handa
handa@ni.aist.go.jp