Peculiar treatment of edited messages

Peculiar treatment of edited messages

[Ted Stern]

In Pgnus 0.96, Emacs 20.3, I sometimes lose my email after editing my
messages.  That is, my cursor sits on a message description in the *Summary*
buffer, I hit 'e', and do some changes to the article.  When finished, I hit
'C-c C-c'.  But then Gnus puts a 'G' on the message summary, indicating that
it is about to be purged from the group.  When I exit the group and re-enter
it using 'C-u SPC', the edited message has disappeared, although other
messages are unaffected.

If I use 'more' or some other editor to actually look through the nnfolder
corresponding to the nnfolder group, the message is still there.  But Gnus no
longer acknowledges this -- how can I get Gnus to re-read all headers in the
nnfolder file?

Oddly enough, there was a message that disappeared this way 2 months ago in my
mail.misc folder, but it recently reappeared.  I have no idea what I did to
bring it back, although 2 months ago I was using pgnus 0.95 vs. 0.96.

Restarting Gnus doesn't seem to help.

Any ideas?  One thing I am thinking is that the email could be getting screwed
up because my write-file-hooks contains functions to remove trailing blanks
and trailing empty newlines.

Ted
-- 
 Tera Computer Company                                   http://www.tera.com
 411 First Avenue South, Suite 600        Direct 206/701-2182, Main 701-2000
 Seattle, WA 98104-2860                   Fax1   206/701-2205, Fax2 701-2500

Re: Peculiar treatment of edited messages

[Dan Christensen]

Ted Stern <stern@tera.com> writes:
> In Pgnus 0.96, Emacs 20.3, I sometimes lose my email after editing my
> messages.  That is, my cursor sits on a message description in the *Summary*
> buffer, I hit 'e', and do some changes to the article.  When finished, I hit
> 'C-c C-c'.  But then Gnus puts a 'G' on the message summary, indicating that
> it is about to be purged from the group.  When I exit the group and re-enter
> it using 'C-u SPC', the edited message has disappeared, although other
> messages are unaffected.
> 
> If I use 'more' or some other editor to actually look through the nnfolder
> corresponding to the nnfolder group, the message is still there.  But Gnus no
> longer acknowledges this -- how can I get Gnus to re-read all headers in the
> nnfolder file?

I've had similar things happen, although if I remember correctly the
articles actually vanished from the nnfolder file as well, and I had
to get them from the Incoming directory.  

Dan

Re: Peculiar treatment of edited messages

[Dan Christensen]

Dan Christensen <jdc@chow.mat.jhu.edu> writes:
> Ted Stern <stern@tera.com> writes:
> > In Pgnus 0.96, Emacs 20.3, I sometimes lose my email after editing my
> > messages.  That is, my cursor sits on a message description in the *Summary*
> > buffer, I hit 'e', and do some changes to the article.  When finished, I hit
> > 'C-c C-c'.  But then Gnus puts a 'G' on the message summary, indicating that
> > it is about to be purged from the group.  When I exit the group and re-enter
> > it using 'C-u SPC', the edited message has disappeared, although other
> > messages are unaffected.
> > 
> > If I use 'more' or some other editor to actually look through the nnfolder
> > corresponding to the nnfolder group, the message is still there.  But Gnus no
> > longer acknowledges this -- how can I get Gnus to re-read all headers in the
> > nnfolder file?
> 
> I've had similar things happen, although if I remember correctly the
> articles actually vanished from the nnfolder file as well, and I had
> to get them from the Incoming directory.  

This happened to me again today, so I can give a bit more info.  I
had point in the *Article* buffer when I hit `e';  not sure if that
is relevant, but it might be.  What happened is that the headers of
the article in my nnfolder file got replaced by just the headers I
have displayed in the *Article* buffer:

> From nobody Sun Oct 24 14:00:54 1999
> From: James McClure <mcclure@math.purdue.edu>
> Subject: Re: operad actions on Tot
> To: jdc@jhu.edu (Dan Christensen)
> Date: Sat, 23 Oct 1999 21:12:58 -0500 (EST)
> In-reply-to: <87ogel9k4o.fsf@chow.mat.jhu.edu>
> Message-id: <199910240213.VAA19485@cauchy.math.purdue.edu>

(I have In-reply-to and Message-id in my gnus-visible-headers.)

Note that the "From " line is incorrect, and the various other
headers that Gnus makes use of (like X-Gnus-Mail-Source, Xref,
and X-Gnus-Article-Number) are all missing.  So as far as Gnus
can tell this article doesn't exist.

Dan

Re: Peculiar treatment of edited messages

[Shenghuo ZHU]

>>>>> "Dan" == Dan Christensen <jdc@jhu.edu> writes:

Dan> Dan Christensen <jdc@chow.mat.jhu.edu> writes:

>> Ted Stern <stern@tera.com> writes:

>>> In Pgnus 0.96, Emacs 20.3, I sometimes lose my email after editing
>>> my messages.  That is, my cursor sits on a message description in
>>> the *Summary* buffer, I hit 'e', and do some changes to the
>>> article.  When finished, I hit 'C-c C-c'.  But then Gnus puts a

[...]


>> I've had similar things happen, although if I remember correctly
>> the articles actually vanished from the nnfolder file as well, and
>> I had to get them from the Incoming directory.

Dan> This happened to me again today, so I can give a bit more info.
Dan> I had point in the *Article* buffer when I hit `e'; not sure if
Dan> that is relevant, but it might be.  What happened is that the
Dan> headers of the article in my nnfolder file got replaced by just
Dan> the headers I have displayed in the *Article* buffer:


[...]

Dan> Note that the "From " line is incorrect, and the various other
Dan> headers that Gnus makes use of (like X-Gnus-Mail-Source, Xref,
Dan> and X-Gnus-Article-Number) are all missing.  So as far as Gnus
Dan> can tell this article doesn't exist.

Right. Hitting `e' in *Article* buffer causes this problem. Besides,
it will not decode MIME either.

Different functions are called for hitting `e' in *Article* buffer and
*Summary* buffer. Therefore, I am wondering whether we should remove
function `gnus-article-edit' and use `gnus-summary-edit-article'
instead.

As a hot-fix, evaluating the following lines may help.

(gnus-define-keys gnus-article-mode-map 
        "e" gnus-article-read-summary-keys)

I am not sure what caused Ted's bug. Maybe the separating blank line.

To bring back those missing articles, I tried to move all articles in
the group to a temporary group, then rescan the group, and repeat
until no articles show in the group.  It was long time ago. I also am
not sure whether it works for Ted's situation.

-- 
Shenghuo ZHU

Re: Peculiar treatment of edited messages

[Shenghuo ZHU]

>>>>> "Ted" == Ted Stern <stern@tera.com> writes:

>>>>> "ZSH" == Shenghuo ZHU <zsh@cs.rochester.edu> writes:

[...]


ZSH> To bring back those missing articles, I tried to move all
ZSH> articles in the group to a temporary group, then rescan the
ZSH> group, and repeat

Ted>                                               ^^^^^^^^^^^^^^^^
Ted> ???  will this work even if the previously edited files no longer show
Ted> in the summary buffer?

[...]

In Gnus, there are some codes to rebuild the nnfolder-article-marker.
I am not sure when they are executed. Here are some comments. When you
move all articles with marker from the group, the articles without
marker tends to be at the end of the file (folder). Then the codes are
executed.

,--------
| ;; As long as we trust that the user will only insert unmarked mail
| ;; at the end, go to the end and search backwards for the last
| ;; marker.  Find the start of that message, and begin to search for
| ;; unmarked messages from there.
`--------

BTW, it seems to need some codes (nnoo-request-check-consistence) to
automatically check and fix the consistence of back-ends, just like
fsck does on file-systems.  We often got problems like expiration
problem (some expirable articles are never removed. similar problem in
agent), consistence in cache, missing articles, bad overview or active
files, unwanted directories. Any more?

-- 
Shenghuo

Re: Peculiar treatment of edited messages

[Lars Magne Ingebrigtsen]

Shenghuo ZHU <zsh@cs.rochester.edu> writes:

> Different functions are called for hitting `e' in *Article* buffer and
> *Summary* buffer. Therefore, I am wondering whether we should remove
> function `gnus-article-edit' and use `gnus-summary-edit-article'
> instead.

Yup.  Fix in Pterodactyl Gnus v0.98.1.

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