Re: [gmime-devel] Accessing attachments including inline-parts.

[Steve Kemp <steve steve org uk>]

On Sun May 25, 2014 at 18:51:22 -0400, Jeffrey Stedfast wrote:

> Right - your current code will only work with GMimeParts and won't
> work with GMimeMessageParts because GMimeMessageParts do not subclass
> GMimePart and so g_mime_part_*() functions will fail.

  Your explanations always make these things seem so easy.  I'm
 enormously in your debt, once again.

>         if (GMIME_IS_MESSAGE_PART (part))
>         {
>             GMimeMessage *msg = g_mime_message_part_get_message (GMIME_MESSAGE_PART (part));
>
>             g_mime_object_write_to_stream (GMIME_OBJECT (msg), mem);
>         }
>         else
>         {
>             GMimeDataWrapper *content = g_mime_part_get_content_object (GMIME_PART (part));
>
>             g_mime_data_wrapper_write_to_stream (content, mem);
>         }

  This is the key that I was missing, and now I've got something that
 is working well, so far as I can see, for random mails and the
 problematic one that was reported to me.

  (Subject to bugs porting this 'demo' code into the real application,
  of course.)

>         // by setting the owner to FALSE, it means unreffing the memory stream won't
>         // free the GByteArray data.
>         g_mime_stream_mem_set_owner (GMIME_STREAM_MEM (mem), FALSE);

  That's a good thing to see too, thanks.

>         // get the GByteArray data
>         adata = g_mime_stream_mem_get_byte_array (GMIME_STREAM_MEM (mem));

  Here we need a "->data" added, rather than just the start of the
  GByteArray, but yeah I've got that now.

> I'd recommend replacing the malloc & memcpy logic in CAttachment with
> just using the GByteArray to avoid unnecessary overhead of
> duplicating the content in memory, but that's up to you.

  That'll have to happen, yes.  For the moment I tend to use malloc
 sparingly and new/delete a few places.  But mostly I just reparse
 things on the heap, so I haven't had to choose a definitive
 memory-strategy.

  (This will change though, if I'm having to reparse multi-Mb PPT
 files, or similar.)

> Hopefully my code snippet above helps get you closer.

  Definitely.

> There's one last thing you'll need to be aware of with the code
> snippet I pasted above, though, which is that the iterator will
> descend into attached messages, so you'll be extracting both the
> attached messages *and* each of the attachments they contain with the
> logic in my snippet above.

  I think that I'll document that and call it a feature...

> >   My advice to anybody writing a mail client is to stick with mutt .. ;)

> Been down that road and scarily enough, I'm half considering going
> down that road again. Lord help me.

  Well you obviously understand the MIME parts pretty well.  The other
 half of the job is deciding what user-interface to have, and people
 get very attached to those.  Me?  I always liked mutt-patched with the
 sidebar, and when that broke my client was born.  (Although it looks
 different the feeling is still the same:  Folders -> Indexes ->
 Messages.)



Steve
-- 
http://www.steve.org.uk/