Re: *Group* buffer disaster fix

[Lloyd Zusman <ljz@asfast.com>]

Hrvoje Niksic <hniksic@srce.hr> writes:

> [ ... ]

> > Why do we need to do a `kill-buffer' in this routine at all when
> > `(mapcar 'mm-destroy-part gnus-article-mime-handles)' is already
> > being done within `gnus-summary-exit' and `gnus-group-exit-hook'?
> 
> Because we want the "*mm*" buffer to go away as soon as we're done
> with it.  Also, the code killing the buffer was already there -- no
> `+' before it.  I only added the unwind-protect bit.

I know that you didn't add the `kill-buffer'.

I was wondering if the `mapcar' that gets invoked later might break
something, since the reference to the ` *mm*' buffer will be put into
the mime handle alist for future deletion, even though that buffer
gets killed here.

However, I just now figured out that this is probably not a problem,
since `mm-destroy-part' only does a `kill-buffer' if `buffer-live-p'
succeeds on the buffer referenced in this alist.  Hence, in this case,
we're safe to do the `kill-buffer' on the ` *mm*' buffer *and* to add
a reference to this killed buffer to the alist.

Now, as long as we're sure that no other code tries to do something to
this killed `* mm*' buffer in the alist prior to the `mapcar' call,
then we're OK.

-- 
 Lloyd Zusman
 ljz@asfast.com