Maybe Pan should convert all strings to utf-8 when it reads them in,
for example, the article headers could be converted in nntp.c when
they're downloaded.  Then these mismatch issues would be moot.

This problem also exists in filter-phrase.

<charles> but IIRC we are going to store .msg files in utf8 format, so
we don't need to convert after fread() unless we're
          feelin paranoid
<charles> We need to convert to utf-8 in acache_set_message()
<charles> If gmime has an option for utf-8 output, we should just
create a GMimeMessage there, dump its output to file, and
          then destroy the GMimeMessage
<charles> that way GMime can handle for us esoteric cases of different
mime attachments having different charsets
<fejj> hmmm, gmime doesn't have a way to write content in UTF-8
<charles> if you don't want to do that, Pan could just stick a decode
in a foreach function for each part
<fejj> you could do a g_mime_objct_foreach() and get the data-wrapper
and then get the stream, create a filtered-stream to
          convert to UTF-8, then re-set the stream into the
data-wrapper and set the MIME part's charset to UTF-8
<fejj> that should work
<charles> okay
<charles> sounds good
<fejj> although if the content isn't in the charset the mime part says
it is and/or if there are problems
          converting from %charset to UTF-8 when writing, you may get
corrupted data

Punting remaining bugs to 0.12.1

This is fixed for 0.12.1.