Re: little nits with 5.2.15

[Christopher Davis <ckd@loiosh.kei.com>]

[Ah, aren't interactions between various packages fun?]

SLB> == Steven L Baur <steve@miranova.com>

 ckd> - point seems to jump around in the summary buffer when selecting
 ckd>   the next article; this didn't happen with 5.0.15 (on beta21, but I
 ckd>   don't think anything in that area of XEmacs changed).

 SLB> I don't notice this.  Does the behavior change if you add (setq
 SLB> gnus-auto-center-summary nil) to your .gnus?

Yes.  The jumping around goes away if I do that.

I should be more specific about the jumping around: it basically goes
something like the following paragraph. (I might have the order a little
off, it's hard to be sure what happens exactly when on X.  It's easier to
see on a slow tty.)

The cursor moves to the next line, near the beginning (one character in, I
think).  The "R" mark gets added.  The new message is displayed.  The
cursor moves to the colon.  The buffer moves up a line.

It's REALLY distracting.  (Even worse on a tty :)

 SLB> This kind of sounds like the XEmacs off-by-1 auto-center bug, but I
 SLB> thought it was fixed.  I discovered how much I hated auto-centering
 SLB> so I have left it turned off in my environment.

I do like auto-centering (though it seems to be working differently than I
remember it).  I also thought the OBOB was fixed, and in any case I
shouldn't be seeing it, because as a workaround when it WAS a problem I
turned off the horizontal scrollbar and the modeline shadow, so I don't
get any clipped lines (and wasn't that what was triggering it?).

The auto-centering behavior I remember, when used with a 5-line summary
buffer, always auto-centered on the second line from the top, giving you
one line of "look-behind" and three of "look-ahead".  It did *not* do what
Gnus 5.2 is doing, which is to auto-center on the second line from the top
unless there would be blank lines in the window.  (I *want* the blank
lines, that's how I can tell I'm at the end of the group without having to
look at the modelines.)

 ckd> - [tm bug?] Is there a way to get message-mode to add my .signature
 ckd>   after I hit C-c C-c but before tm-edit MIME-fies the message?
 ckd>   tm-edit doesn't put in the proper signature separator. (This
 ckd>   should only happen in news buffers; mail shouldn't get a
 ckd>   .signature unless I ask for it.)

 SLB> Examine the hook order carefully on message-send-hook, and arrange to
 SLB> have your .signature inserter put ahead of mime-editor/maybe-translate.

 SLB> There's message insert-signature, and message-news-p, and
 SLB> message-mail-p for deciding when to do what.  The message setup
 SLB> hooks are broken though, so you can't customize your .signature by
 SLB> hook alone based on News -vs- Mail (unless this has very recently
 SLB> been fixed).

Well, that I can fix.  Something like this (as yet untested :) should
work.  (tm should maybe add mime-editor/maybe-translate to the end of
message-send-hook, rather than the beginning...then I wouldn't need to do
this trickery.)

(defun ckd-maybe-add-signature ()
  "Maybe add a .signature before posting."
  (if (message-news-p)
      (message-insert-signature t)))

;; slight trickiness here (order-dependent hook setting)
(add-hook 'mime-setup-load-hook '(lambda ()
				   (add-hook 'message-send-hook
					     'ckd-maybe-add-signature)))

 ckd> - the signature separator, like mail-header-separator, should be
 ckd>   added to paragraph-separate when in message-mode so that filling
 ckd>   the last paragraph of a message doesn't fill your .signature with
 ckd>   it.

 SLB> Yup.  That is annoying.

Easily fixed, though; just add "-- $\\|" to the concat in message.el.

 ckd> - W w is broken somehow; it's filling things without paying any
 ckd>   attention to paragraph separators. I am using filladapt, but it
 ckd>   didn't do this before.

 SLB> Filladapt is partially broken on 19.14.

Well, okay, but that's not causing my problem; the replacement of
"gnus-article-word-wrap" with "gnus-article-fill-cited-article" is causing
the problem.  (M-x gnus-article-word-wrap Does The Right Thing; W w, or
gnus-article-fill-cited-article, is breaking stuff.)

I'm not 
-- 
Christopher Davis <ckd@kei.com> <URL: http://www.kei.com/homepages/ckd/ >
"I conclude that the CDA is unconstitutional and that the First Amendment
 denies Congress the power to regulate protected speech on the Internet."
                               -- Judge Stewart Dalzell in _ACLU v. Reno_