modeline bug, reprised

modeline bug, reprised

[Karl Kleinpaste]

I appear finally to have a uniformly reproducible circumstance for
getting the "topmost" buffer below all Gnus-related buffers to have
its modeline screwed up.

The fundamental problem is that a fresh *Summary* modeline has its
glyph and article number/count indicators erroneously dropped on the
wrong buffer's modeline.

If:
1. One's current group is not on the last group in *Group*, and
2. One has entered that group, and
3. One has an *Article* buffer showing (i.e., you have read an article
   in the current group), and
4. One finishes the last article in the group and moves to the next by
   a keystroke which moves one there directly, such as `n'...
Then:
On arriving in the next group, that topmost buffer's modeline will
have been damaged, AND the new *Summary* buffer's modeline will not
have been properly set.  (I had never noticed the latter effect until
I did this.)

I took to using `C-x 5 2' to create an auxiliary frame in which to
hold that topmost (and only non-Gnus) buffer.  (*scratch*, of course.)
I have variables set for auto-selection of the next group without
immediately seeing the 1st article:
	  gnus-auto-select-first nil
	  gnus-auto-select-next 'quietly

If I go to the next group with `n' or `k' while an *Article* buffer is
showing, then the modeline on *scratch* gets destroyed and shows
    --**-[glyph] *scratch* [0] {9 more} (Lisp Interaction)----L1--All--
when the next group is displayed.  Further, the newly-constructed
*Summary* buffer has a modeline showing
    -----XEmacs: *Summary group.name* (Summary)----L1--All--
Note that *Summary* shows "XEmacs:" and is missing both the glyph and
the article indicators that have erroneously appeared in the *scratch*
modeline.

When an article is then selected in the new group, the *Summary*
modeline is corrected, so that it shows
    -----[glyph] *Summary g.n* [12345] {8 more} (Summary)----L1--All--

This does not happen if any of the following are true:
1. One does not get to the next group directly with `n' or similar.
   (That is, quitting out of one group and then entering the next does
    not induce the bug.)
2. No *Article* is showing. (???  Maybe this has to do with how many
   buffers get disposed of, when the new *Summary* is built?)

So the problem/fix has to do with making sure that the right buffer's
modeline is being messed with, when the glyph and article indicators
are being added.

The modified flag ("**") on *scratch* is wrong, as hitting an "undo"
there tells me that no more undo info is available (which is correct,
since I haven't touched *scratch*).

I'm in XEmacs 19.16 under Linux with q0.17.  This has been happening
since about q0.15.  I'm pretty sure 0.14 did not do this.

Is this enough info with which to reproduce/fix the bug?

--karl

Re: modeline bug, reprised

[Lars Magne Ingebrigtsen]

Karl Kleinpaste <karl@jprc.com> writes:

> I took to using `C-x 5 2' to create an auxiliary frame in which to
> hold that topmost (and only non-Gnus) buffer.  (*scratch*, of course.)
> I have variables set for auto-selection of the next group without
> immediately seeing the 1st article:
> 	  gnus-auto-select-first nil
> 	  gnus-auto-select-next 'quietly

[...]

> Is this enough info with which to reproduce/fix the bug?

I think I've located and fixed this bug.  Could you check after
Quassia Gnus v0.19 has been released to see whether things start
working?

-- 
(domestic pets only, the antidote for overdose, milk.)
  larsi@gnus.org * Lars Magne Ingebrigtsen