This turned out to be a bug in Horde_Prefs, not IMP. Fixed in Horde 
HEAD and 3.1.4.

Of course there is. The preference value is converted from browser 
charset to backend charset when saving the pref and vice versa when 
loading the pref. The text should always be encoded in   
NLS::getCharset().

I applied the fix of Jan Schneider (11/28/2006) to my test 
installation (imp-4.1.5 stable)

replacing

"$msg = $this->_expandAttribution($prefs->getValue('attrib_text'), 
$from, $h) . "\n\n" . $msg;"

with

"$msg = 
$this->_expandAttribution(String::convertCharset($prefs->getValue('attrib_text'),NLS::getCharset(), $old_part->getCharset()), $from, $h) . "\n\n" . 
$msg;"

line 355 in ./imp/lib/Compose.php



Even after applying this fix, the bug is still remaining in some case 
in the stable version (imp-4.1.3).

My default charset is "iso-8859-1".

When I am replying to a message written in charset  "iso-8859-1", it 
is correct :

"a écrit :"

But when I am replying to a message written in charset "utf8", I still 
see the bug :

"a écrit :"

I think what needs to be done is to save the browser charset at the 
time we save 'attrib_text' so we know how to convert the text when 
creating the message (i.e. via a 'attrib_text_charset' implicit pref 
set in a prefs callback).  As of right now, there is no way of knowing 
what charset 'attrib_text' is when we try to convert (definitely no 
guarantee NLS::getCharset() is the correct charset).

This still needs to be done in HEAD. I looked at it, but the place to 
get the old part's charset is moved around in HEAD, and I'm not clear 
why we're converting to the old part's charset instead of to the 
current charset anyway. Jan or Michael?

It has been fixed in imp/lib/Compose.php version 1.107.2.39 (FRAMEWORK_3)



I think if remains in Imp H3 4.2-cvs.

      imp/lib/Compose.php version 1.268



I tested with Horde-3.1.3 , imp H3 4.2-cvs

Thank you.

But I don't know what to do to fix my own installation (which file to 
patch ?).

We are currently using "imp3" in production and I can wait for the 
next  "imp" stable version.

Today, I sent  a comment  on ticket  #3676 and I see that you 
reassigned it . I think that I don't have to open an other ticket for 
that.

I am having at time a lot of  difficulty  using "horde_share"  to 
convert a public sql address book , I will post to the turba mailing 
list.

Regards

Fixed.

I notice no other place  where  I see UTF-8 chars in translated strings.

Is this the only place where you see UTF-8 chars in translated strings?

When replying to a message by clicking on "Répondre"(Reply),   A   
"Reply:Re: ..." window  is opening.

In the "Text" paragraph (box),  there is a sentence introducing the 
original post.

In English it looks like :

Quoting  "Name of the sender" <"email adress of the sender">

In  french this sentence is bad written

It is:

"Name of the sender" <"email adress of the sender">  a écrit :

but it should be :

"Name of the sender" <"email adress of the sender"> a écrit :

Sorry, I don't follow. Can you try to explain again in other words?

(Using "french" language and  charset "iso-8859-1")

when responding to a message, the end of the line introducing the 
original message is the coding of  'a écrit :' in "utf8"

it is written :

**** <****@****> a écrit :

if should have been written :

**** <****@****> a écrit :



Remark : the coding of : 'a écrit :' in "utf8" is ' a écrit :'

as '   ' is the  "NO-BREAK SPACE"