mm-with-unibyte-current-buffer broken in Emacs trunk

mm-with-unibyte-current-buffer broken in Emacs trunk

[Sascha Wilde]

Hi *,

with an current build of GNU Emacs from bzr I can not forward
(gnus-summary-mail-forward) mail any more.

The problem boils down to message-forward-make-body-mml calling
mm-with-unibyte-current-buffer which in turn tries to set the read only
variable enable-multibyte-characters.

According to this comment in mm-with-unibyte-current-buffer:

  ;; FIXME: (default-value 'enable-multibyte-characters) is read-only
  ;; so let-binding it is wrong.  The right fix is to not use this
  ;; macro at all any more, since it's been ill-defined from the start.

the problem is known, but I'd say not being able to forward messages
(and there might be other serious problems, too) is an regression, so
this _really_ needs fixing.

FWIW: simply removing the call to mm-with-unibyte-current-buffer from
message-forward-make-body-mml worked for me -- but I suspect there are
corner cases where this would cause problems.

cheers
sascha
-- 
Sascha Wilde : xedit? Das sieht zwar wie vi aus als könne es nix, aber im
             : Gegensatz zu vi kann es wirklich nix und nix ist noch geschönt!
             : (Michael Core in dafc)

Re: mm-with-unibyte-current-buffer broken in Emacs trunk

[Katsumi Yamaoka]

>>>>> Sascha Wilde wrote:
> with an current build of GNU Emacs from bzr I can not forward
> (gnus-summary-mail-forward) mail any more.

> The problem boils down to message-forward-make-body-mml calling
> mm-with-unibyte-current-buffer which in turn tries to set the read only
> variable enable-multibyte-characters.

[...]

The day before yesterday I just noticed the message forwarding
doesn't work, and fixed the code temporarily in Git and Bzr repos.

2010-05-10  Katsumi Yamaoka  <yamaoka@jpl.org>

	* message.el (message-forward-make-body-plain)
	(message-forward-make-body-mml):
	Don't use mm-with-unibyte-current-buffer.

But I have a plan to fix it further; that is to make the forwarding
codes assume that an original message to be forwarded is always
a multibyte string.  Actually an original message always comes from
only the " *Original Article*" buffer in which the multibyteness
is turned on, AFAIK.