Is anyone actively using these? I don't use them myself, which means
I'm not very qualified to criticise, but it does seem broken.
When Mutt is deciding what charset to use for a message you are
sending, it calls mutt_iconv_open with M_ICONV_HOOK_FROM. When it then
really sends the message, it calls fgetconv_open with M_ICONV_HOOK_TO.
This is obviously wrong and causes bad behaviour in the following
case:
charset=utf-8
send_charset=us-ascii:iso-8859-1:utf-8
charset-hook utf-8 utf-8
charset-hook . gb2312
The aim of these settings is to treat all messages as being in GB2312
unless they are labelled as UTF-8. This is the sort of thing someone
who receives a lot of broken Chinese e-mails might want to do, with
the significant difference that such a person is probably using a
GB2312 terminal instead of a UTF-8 one like me.
With these settings, when I send a message containing a '£', Mutt
correctly chooses iso-8859-1, but then misconverts "£" to "??".
Stupid.
Now, I don't really understand why charset-hooks are being used AT ALL
for sending messages, now that iconv-hook is available, so I am
suggesting the attached patch, hoping that people who use charset-hook
can tell me if it makes sense.
Note that this patch makes M_ICONV_HOOK_TO obsolete. If the patch is
correct, further tidying up should be done.
Edmund
--- mutt.orig/sendlib.c Sat Apr 20 20:40:39 2002
+++ mutt/sendlib.c Fri May 31 09:47:50 2002
@@ -496,7 +496,7 @@
if (a->type == TYPETEXT && (!a->noconv))
fc = fgetconv_open (fpin, Charset,
mutt_get_body_charset (send_charset, sizeof (send_charset),
a),
- M_ICONV_HOOK_TO);
+ 0/*M_ICONV_HOOK_none*/);
else
fc = fgetconv_open (fpin, 0, 0, 0);
@@ -682,7 +682,7 @@
CONTENT_STATE *states;
size_t *score;
- cd1 = mutt_iconv_open ("UTF-8", fromcode, M_ICONV_HOOK_FROM);
+ cd1 = mutt_iconv_open ("UTF-8", fromcode, 0/*M_ICONV_HOOK_none*/);
if (cd1 == (iconv_t)(-1))
return -1;
