Changed criteria for deciding if articles have been read?

Changed criteria for deciding if articles have been read?

[Lloyd Zusman]

I've been away from the group for a month or so, and I haven't done a
CVS update in that long.

I just now did the CVS update, and when I run the resultant gnus
build, I see many old articles in almost all my groups showing up as
unread.

The weird thing about this is that not all old articles show up as
unread ... only maybe the past 3-5 weeks' worth.

If I switch to my older version of Oort, the unread-ness of the
articles once again appears correct.

Are recent Oort versions doing something different with marks, or
something?

Forgive me if this has been discussed during the past few weeks, but
due to my absence, I've gotten way behind in all my mailing list and
newsgroup reading, and I still haven't been able to catch up on
everything I've missed.

Feel free to email me privately if this has recently been discussed
and you don't want to clutter the group with redundant info.

Thanks in advance.

-- 
 Lloyd Zusman
 ljz@asfast.com

Re: Changed criteria for deciding if articles have been read?

[Christoph Conrad]

    > Are recent Oort versions doing something different with marks,
    > or something?

(setq nnml-marks-is-evil     t
      nnfolder-marks-is-evil t)

Re: Changed criteria for deciding if articles have been read?

[Simon Josefsson]

Lloyd Zusman <ljz@asfast.com> writes:

> I just now did the CVS update, and when I run the resultant gnus
> build, I see many old articles in almost all my groups showing up as
> unread.
>
> The weird thing about this is that not all old articles show up as
> unread ... only maybe the past 3-5 weeks' worth.
>
> If I switch to my older version of Oort, the unread-ness of the
> articles once again appears correct.
>
> Are recent Oort versions doing something different with marks, or
> something?

Yes, Oort also keep marks in a .mrk/.marks file in nnfolder/nnml
groups.  Some said M-x nnml-generate-nov-databases RET fixed the
problem in the transition.  But maybe you could try debugging this, if
you can reproduce it by switching between versions, so that others
don't have to do M-x nnml.. RET.

For starters, in one group that triggers this, could you cut'n'paste
the content of `G E' (both new and old version) and the contents of
the .mrk/.marks file (new version)?

Re: Changed criteria for deciding if articles have been read?

[Lloyd Zusman]

Simon Josefsson <jas@extundo.com> writes:

> Lloyd Zusman <ljz@asfast.com> writes:
> 
>> [ ... ]
>>
>> Are recent Oort versions doing something different with marks, or
>> something?
> 
> Yes, Oort also keep marks in a .mrk/.marks file in nnfolder/nnml
> groups.  Some said M-x nnml-generate-nov-databases RET fixed the
> problem in the transition.  But maybe you could try debugging this, if
> you can reproduce it by switching between versions, so that others
> don't have to do M-x nnml.. RET.
> 
> For starters, in one group that triggers this, could you cut'n'paste
> the content of `G E' (both new and old version) and the contents of
> the .mrk/.marks file (new version)?

I did the `G E' for both Oort versions, and you can see the results
below.

After running the new version, no `.mrk/.marks' had been created
anywhere, but I didn't do the `nnml-generate-nov-databases'.  Would
I have to do that before any .mrk/.marks appear?


Results of `G E' in older Oort version
(group buffer shows no unread messages for this group):

  ;;; Editing the group info for `nnml+email:gnus'.
  ;; Type `C-c C-c' after you've finished editing.

  ("nnml+email:gnus" 2
   ((1 . 27024))
   ((reply 26994 27004)
    (tick 27000 27023))
   "nnml:email"
   ((to-address . "ding@gnus.org")
    (to-list . "ding@gnus.org")
    (total-expire . t)
    (expiry-wait . 30.0)))

Results of `G E' in very recent Oort version:
(group buffer shows 591 unread messages for this group):

  ;;; Editing the group info for `nnml+email:gnus'.
  ;; Type `C-c C-c' after you've finished editing.

  ("nnml+email:gnus" 2
   ((1 . 26433))
   ((reply 24924 25041 25273 25295 25306
    (25310 . 25311)
     25329 25337)
    (tick 26017 26020))
   "nnml:email"
   ((to-address . "ding@gnus.org")
    (to-list . "ding@gnus.org")
    (total-expire . t)
    (expiry-wait . 30.0)))


-- 
 Lloyd Zusman
 ljz@asfast.com

Re: Changed criteria for deciding if articles have been read?

[Lloyd Zusman]

Lloyd Zusman <ljz@asfast.com> writes:

> Simon Josefsson <jas@extundo.com> writes:
> 
>> [ ... ]
> 
> After running the new version, no `.mrk/.marks' had been created
> anywhere, but I didn't do the `nnml-generate-nov-databases'.  Would
> I have to do that before any .mrk/.marks appear?

Aha!  I discovered that there *are* indeed a bunch of `.marks' files,
each of which appear in their respective group directory (however, not
under any directory named `.mrk').  For example, for my `gnus' nnml
group, the following file exists:

   ~/Mail/gnus/.marks

It was last created/modified on Aug. 19, which is the last time I
attempted to run a new version of Oort (but I didn't continue using
it, due to an unrelated problem, and I went back to the older version
which is still the version I'm using now).  No doubt this Aug. 19
version of Oort created these `.marks' files when it started up.

The `.marks' file I mentioned above contains this data (it's all on a
single line in the file, but I'm wrapping it here):


  ((read (1 . 26024))
   (reply 24924 25041 25273 25295 25306 (25310 . 25311) 25329 25337)
   (tick 26017 26020))

Funny how it just _happens_ to contain the same info that `G E' shows
below for the "recent Oort version" case.  :)

I then renamed all my `.marks' files in all of my group directories,
and I once again started up the recent version of Oort.  Without those
old `.marks' files, the unread-ness of my articles is now once again
correct.  And then, after exiting the recent version of Oort, there
was now a zero-length `.marks' file in that same group directory.

So ... for those people who are transitioning from an older to a more
recent version of Oort, and who see this same inconsistency concerning
unread articles, I recommend that you search in your group directories
for any `.marks' files that are laying around, and delete them.

Does this solve the problem, or does it just illustrate another aspect
of a bug?


> Results of `G E' in older Oort version
> (group buffer shows no unread messages for this group):
> 
>   ;;; Editing the group info for `nnml+email:gnus'.
>   ;; Type `C-c C-c' after you've finished editing.
> 
>   ("nnml+email:gnus" 2
>    ((1 . 27024))
>    ((reply 26994 27004)
>     (tick 27000 27023))
>    "nnml:email"
>    ((to-address . "ding@gnus.org")
>     (to-list . "ding@gnus.org")
>     (total-expire . t)
>     (expiry-wait . 30.0)))
> 
> Results of `G E' in very recent Oort version:
> (group buffer shows 591 unread messages for this group):
> 
>   ;;; Editing the group info for `nnml+email:gnus'.
>   ;; Type `C-c C-c' after you've finished editing.
> 
>   ("nnml+email:gnus" 2
>    ((1 . 26433))
>    ((reply 24924 25041 25273 25295 25306
>     (25310 . 25311)
>      25329 25337)
>     (tick 26017 26020))
>    "nnml:email"
>    ((to-address . "ding@gnus.org")
>     (to-list . "ding@gnus.org")
>     (total-expire . t)
>     (expiry-wait . 30.0)))
> 
> 
> -- 
>  Lloyd Zusman
>  ljz@asfast.com
> 
> 

-- 
 Lloyd Zusman
 ljz@asfast.com

Re: Changed criteria for deciding if articles have been read?

[Kai Großjohann]

Lloyd Zusman <ljz@asfast.com> writes:

> Aha!  I discovered that there *are* indeed a bunch of `.marks' files,
> each of which appear in their respective group directory (however, not
> under any directory named `.mrk').

Simon meant "*.mrk" files (for nnfolder) or ".marks" files (for
nnml).  I think he assumed that you had been following the discussion
on this, where he first announced to have produced .marks files for
nnml, then said what the file names for nnfolder were.

kai
-- 
Abort this operation?   [OK]  [Cancel]

Re: Changed criteria for deciding if articles have been read?

[Lloyd Zusman]

Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:

> Lloyd Zusman <ljz@asfast.com> writes:
>
>> Aha!  I discovered that there *are* indeed a bunch of `.marks' files,
>> each of which appear in their respective group directory (however, not
>> under any directory named `.mrk').
>
> Simon meant "*.mrk" files (for nnfolder) or ".marks" files (for
> nnml).  I think he assumed that you had been following the discussion
> on this, where he first announced to have produced .marks files for
> nnml, then said what the file names for nnfolder were.

Well, as I mentioned in my earlier post, I still haven't caught up
with all the older discussion.

But in any case, given what I did with the .marks files to take care
of my problem, I don't believe that we need anything besides that to
help people in their transitions through the recent versions of Oort.

Or should there still be something else?


> kai
> -- 
> Abort this operation?   [OK]  [Cancel]


-- 
 Lloyd Zusman
 ljz@asfast.com

Re: Changed criteria for deciding if articles have been read?

[Kai Großjohann]

Lloyd Zusman <ljz@asfast.com> writes:

> But in any case, given what I did with the .marks files to take care
> of my problem, I don't believe that we need anything besides that to
> help people in their transitions through the recent versions of Oort.
> 
> Or should there still be something else?

FWIW, I didn't have to do anything at all.  Just upgraded my Gnus.

kai
-- 
Abort this operation?   [OK]  [Cancel]

Re: Changed criteria for deciding if articles have been read?

[Simon Josefsson]

Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:

> Lloyd Zusman <ljz@asfast.com> writes:
>
>> But in any case, given what I did with the .marks files to take care
>> of my problem, I don't believe that we need anything besides that to
>> help people in their transitions through the recent versions of Oort.
>> 
>> Or should there still be something else?
>
> FWIW, I didn't have to do anything at all.  Just upgraded my Gnus.

Lloyd's analysis is correct, but he points a potential problem: If you
use a Gnus version with .mrk/.marks, then downgrade Gnus to a older
version and read alot of mail, and then upgrade Gnus again, you will
lose those marks added when you used the old Gnus.  That is not nice.

But I wonder if up-down-up-grading isn't already considered harmful.
What happens if you downgrade Gnus, adds some articles to a nnfolder,
then upgrades to a Gnus that uses NOV nnfolders for instance?  There
might be other examples as well.  I remember many times when starting
old Emacsen and starting Gnus 5.3 (or something) which completely
destroyed my .newsrc.eld beyond repair..  (E.g the topics structure
was lost, I think.)

Re: Changed criteria for deciding if articles have been read?

[Dan Christensen]

Simon Josefsson <jas@extundo.com> writes:

> Lloyd's analysis is correct, but he points a potential problem: If you
> use a Gnus version with .mrk/.marks, then downgrade Gnus to a older
> version and read alot of mail, and then upgrade Gnus again, you will
> lose those marks added when you used the old Gnus.  That is not nice.

Could the marks code look at gnus-newsrc-file-version, and if that
is before the first version that correctly supported .marks files,
remove any .marks files it finds?  Or maybe just display a warning
message?

-- 
Dan Christensen
jdc+news@uwo.ca