Re: Customizing coding priority

[Sven Bretfeld]

Eli Zaretskii writes:
 > > The last line outside the citation just containes three German umlauts
 > > (I use this example for testing). The broken characters in the
 > > citation are umlauts too. The headers of these replies say text/plain;
 > > iso-2022-jp was the Content-type. I've tried C-x RET iso-8859-15 as
 > > well as Peter's suggestion M-x prefer-coding-system iso-8859-15. But
 > > the problem is still exactly the same. What else can it be?
 > 
 > How did you insert those umlauts, exactly?  And what does Emacs
 > display if you go to one of those characters and type "C-u C-x ="?

I've already tried this today. It's quite interesting. The umlauts in
the quotations are displayed in a font different from the rest
(somewhat smaller and bolder). These are said to be part of the
iso-8859-15 charset when I type C-u C-x = (just as expected). All
other characters are plain ascii (ASCII (ISO646 IRV)). It is different
outside of quotations. When I add my own text after a quotation all
umlauts I type belong to latin-iso8859-1. When I send this text it is
destructed again. After a while I found out that, if I yank umlauts
from the quotation to my own text (instead of typing them myself) and
send this message to myself, it stays intact and is displayed fine,
albeit the Content-type header says "charset unknown".

To my understanding this means that Emacs is unable to translate
umlauts belonging to iso-8859-15 to the default charset I use for
umlauts when writing replies, i.e. iso-8859-1. Therefore the umlauts
in quotations are kept untranslated in the coding system of the
original sender, i.e. iso-8859-15. This mixture of iso-8859-15 in
quotations and iso-8859-1 in the text typed by myself seems to disturb
Emacs so that it encodes the entire umlauts of my reply as iso-2022-jp
when sending it.

I concluded that I have to avoid iso-8859-1 completely in replies and
use umlauts belonging to iso-8859-15 also for my own part of the
text. But this doesn't work for some reason. I've tried to change the
coding system and the input method of the reply buffer manually to
iso-8859-15 (using C-x f RET and C-x RET C-\) but the umlauts I type
always come as iso-8859-1.

By the way, I've also tried to start Emacs with an almost empty .emacs
file. The problem remaines. So it seems not to depend on any user
specific configuration.

It's really strange. Can it be a simple bug in vm? But it seems to
work for other people.

Thanks for your help

Sven