An issue has been identified with this patch for which I wrote a fix.

I ran many tests and combination with this additional patch (see
attachement) and I could not find any issue left.

The reason I missed this obvious bug was my configuration using my
`org-msg' module which has an advice on
`mml-expand-html-into-multipart-related' to workaround the limitation
that my previous patch addresses.

The issue originates from the fact that the
`mml-expand-html-into-multipart-related' function returns a list of
list for an untouched part while it returns a simple list for expanded
multipart.  The recursive engine of
`mml-expand-all-html-into-multipart-related' is not handling
this.  Considering that the `mml-expand-html-into-multipart-related'
function is only used by mml as an internal function, IMO the best
approach is to make it consistent on its potential return values
structure. The patch I am submitting here makes
`mml-expand-html-into-multipart-related' returns a simple list
regardless of if it is a part or a multipart.

On Wed, Feb 12, 2020 at 11:18 AM Jérémy Compostella
<jeremy.compostella@gmail.com> wrote:
>
> Hi,
>
> I would really appreciate if I could receive a feedback on this patch.  I can't send GPG HTML email because of this limitation and this is is also blocking an issue reported of my org-msg module.
>
> I have attached the patch here as I have fixed a typo in the commit headline.
>
> Let me know if I need to address this patch to another mailing list.
>
> Regards,
> Jeremy
> .
>
> On Tue, Jan 21, 2020 at 4:01 PM Jeremy Compostella <jeremy.compostella@gmail.com> wrote:
>>
>> If the text/html part is not the unique part of the email,
>> mml-generate-mime does not expand it to a related multipart. This
>> break uses cases like PGP signing of an HTML email including images.
>>
>> For instance, if you compose an email with the <multipart sign=pgpmime>
>> and a <part type="text/html" disposition=inline>, the second part is
>> not expanded into a multipart related and the images are not included.
>>
>> Signed-off-by: Jeremy Compostella <jeremy.compostella@gmail.com>
>> ---
>>  lisp/gnus/mml.el | 19 ++++++++++++++-----
>>  1 file changed, 14 insertions(+), 5 deletions(-)
>>
>> diff --git a/lisp/gnus/mml.el b/lisp/gnus/mml.el
>> index cdd8f3d3a5..3d86c5bc40 100644
>> --- a/lisp/gnus/mml.el
>> +++ b/lisp/gnus/mml.el
>> @@ -487,11 +487,8 @@ type detected."
>>                  (= (length cont) 1)
>>                  content-type)
>>         (setcdr (assq 'type (cdr (car cont))) content-type))
>> -      (when (and (consp (car cont))
>> -                (= (length cont) 1)
>> -                (fboundp 'libxml-parse-html-region)
>> -                (equal (cdr (assq 'type (car cont))) "text/html"))
>> -       (setq cont (mml-expand-html-into-multipart-related (car cont))))
>> +      (when (fboundp 'libxml-parse-html-region)
>> +       (setq cont (mapcar 'mml-expand-all-html-into-multipart-related cont)))
>>        (prog1
>>           (with-temp-buffer
>>             (set-buffer-multibyte nil)
>> @@ -510,6 +507,18 @@ type detected."
>>             (buffer-string))
>>         (setq message-options options)))))
>>
>> +(defun mml-expand-all-html-into-multipart-related (cont)
>> +  (cond ((and (eq (car cont) 'part)
>> +             (equal (cdr (assq 'type cont)) "text/html"))
>> +        (mml-expand-html-into-multipart-related cont))
>> +       ((eq (car cont) 'multipart)
>> +        (let ((cur (cdr cont)))
>> +          (while (consp cur)
>> +            (setcar cur (mml-expand-all-html-into-multipart-related (car cur)))
>> +            (setf cur (cdr cur))))
>> +        cont)
>> +       (t cont)))
>> +
>>  (defun mml-expand-html-into-multipart-related (cont)
>>    (let ((new-parts nil)
>>         (cid 1))
>> --
>> 2.24.1
>>
>
>
> --
> « Si debugger, c'est supprimer des bugs, alors programmer ne peut être que les ajouter » - Edsger Dijkstra



-- 
« Si debugger, c'est supprimer des bugs, alors programmer ne peut être
que les ajouter » - Edsger Dijkstra