bug#3790: 23.0.95; C-x C-s in RMAIL buffer clobbers buffer-file-coding-system

[Eli Zaretskii]

> From: Glenn Morris <rgm@gnu.org>
> Cc: 3790@emacsbugs.donarmstrong.com
> Date: Thu, 09 Jul 2009 02:05:03 -0400
> 
> Eli Zaretskii wrote:
> 
> > After typing C-x C-s in an Rmail buffer showing some message,
> > buffer-file-coding-system becomes no-conversion, no matter what
> > encoding was used to decode and display the message.
> 
> I could not reproduce that with a chinese-iso-8bit-unix message.

It is data-dependent.  Do you see the problem with this message in the
message collection?  The presence of the line with BABYL OPTIONS below
is crucial, and "BABYL OPTIONS" need to be flushed all the way to the
left; a single blank before it makes the problem go away.  (See
auto-coding-regexp-alist for details.)

BABYL OPTIONS: -*- rmail -*-

> C-x C-s should just call rmail-write-region-annotate, which does not
> do very much.

No, it also calls write-region, to actually save the message
collection to disk.  And that calls select-safe-coding-system (via
Vselect_safe_coding_system_function), which in turn calls
find-auto-coding, which calls auto-coding-regexp-alist-lookup, which
sees the "BABYL OPTIONS" string below, and selects no-conversion for
writing the file.

I'm still not sure how no-conversion winds up in the mode line,
though.  Any ideas?