*scratch* buffer modeline wrong after exit

*scratch* buffer modeline wrong after exit

[Karl Kleinpaste]

I'm sending this note on Lars' request/suggestion, after a `C-c C-b'
bug report of mine proved not to be reliably reproducible.

In q0.15 and q0.16, I am irregularly encountering the bizarre effect
that, upon exit from Gnus, my modeline looks like this:

--** [gnus glyph] *scratch* [0] {1 more}   (Lisp Interaction)-----L1--All---

Note the *Summary*-style bits that remain in a *scratch* buffer.  Note
also that the buffer claims to be modified, though attempting to
execute `undo' fails, with a complaint that no more undo info is
available.

If I start killing buffers with `C-x C-k', eventually *scratch* is
reconstituted and has a normal modeline, lacking the gnus glyph and
summary counters.

As for the circumstance which leads to this error, I am not sure
precisely what it is that does this.  I have tried:
1. Both M-x gnus and M-x gnus-no-server.
2. Exiting immediately after start.
3. Reading entirely through a couple groups until they show no unread
articles.
4. Reading partially through a group so that some unread articles
remain.
5. Sending (and not sending) replies to articles being seen.

So far, I have no reliably reproducible circumstance.  It seems to
happen (guestimating) about 1-in-5 exits from Gnus.

Is anyone else seeing this in the most recent release or two?

My configuration:
XEmacs 19.16 under Linux (recent RH).
I change gnus-{group,summary,tree}-line-format, but have no
customizations regarding modelines.
All my customization is in .emacs; I don't use .gnus at all.

Clues, or at least corroborating failures, would be most welcome.

--karl

Re: *scratch* buffer modeline wrong after exit

[Dan Christensen]

Karl Kleinpaste <karl@jprc.com> writes:
> I'm sending this note on Lars' request/suggestion, after a `C-c C-b'
> bug report of mine proved not to be reliably reproducible.
> 
> In q0.15 and q0.16, I am irregularly encountering the bizarre effect
> that, upon exit from Gnus, my modeline looks like this:
> 
> --** [gnus glyph] *scratch* [0] {1 more}   (Lisp Interaction)-----L1--All---
> 
> Note the *Summary*-style bits that remain in a *scratch* buffer.

I've had something similar happen to me once.  I was editing my
.gnus file and at some point I noticed that its modeline had
incorporated Gnusian features.  This was with something in the
q0.14 to q0.16 range.  I don't recall the circumstances, although
I think it did occur just after exiting Gnus.

Dan

Re: *scratch* buffer modeline wrong after exit

[Karl Kleinpaste]

An addendum to my previous observations:

The extraneous modeline bits get dumped on whatever buffer is lying
around to be picked up as "current" when Gnus exits.  It doesn't seem
to matter at all whether it's *scratch*, .emacs, foo.c, or anything else.

Also, maybe it's my imagination, but it seems to me that there's an
inordinately long pause at exit, between the saving of .newsrc.eld and
the removal of the *Group* buffer, returning to whatever buffer is
underneath, now with a screwed modeline.  But I can't say I noticed
this pause much until this morning, so I can't say whether this is new
in 0.15 or 0.16, versus previous behavior from 0.14 (when the modeline
damage was not occurring).

--karl

Re: *scratch* buffer modeline wrong after exit

[Lars Magne Ingebrigtsen]

Karl Kleinpaste <karl@jprc.com> writes:

> The extraneous modeline bits get dumped on whatever buffer is lying
> around to be picked up as "current" when Gnus exits.  It doesn't seem
> to matter at all whether it's *scratch*, .emacs, foo.c, or anything else.

Hm.  `gnus-group-set-mode-line' (which is the only function that
meddles with the group mode line, I think) does `set-buffer' to
gnus-group-buffer before doing anything.  So this is rather odd.

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