Karl Eichwalder <keichwa@gmx.net> writes:

> Editing UTF-8 encoded mail messages does not seem to work reliably.
> I'll append such a message mailed to myself using lynx with "wide"
> ncurses support.
>
> The problem: Press e (gnus-summary-edit-article), then C-c C-c.  Saving
> back the message Gnus turns it into a base64 encoding and worse, it
> does not seem to preserve the UTF-8 charset.

You are seeing two different problems:

1. When you do a plain `e' Gnus, mangles the encoding.  This seems to be
   one of the usual MULE-related \201-problems.

   In this case it's caused by `mm-get-part' (in mm-decode.el), which
   uses the macro `mm-with-unibyte-current-buffer-mule4'.  This macro
   does *not* make the buffer unibyte temporarily.

   I don't understand the point of `mm-w-u-c-b-mule4' -- it doesn't
   work.  Why is it ever used?  I suspect that the use of it in several
   places could be causing some of the occasional \201-problems.

   `mm-with-unibyte-current-buffer' *does* work, though.  And if we use
   that macro instead, the problem is fixed (but maybe it breaks older
   Emacsen?).