[2.2.3] lowmark (was: [2.3] lowmark)

[Bettina Fink]

/me wrote:

>>> The problem with the mysterious returning "1584" is on a trad
>>> spool 2.2.3 with "traditional" overview, not on a 2.3.x, see 
>>> my last mail for details ... And I thought Russ and you are not

[...]

Ok, found at least this one. The mysteriously returning incorrect
lowmarks on 2.2.3 (trad spool, trad overview) seem to be history 
file based:

news at wuff:~ > grep de.newusers.infos log/expire.lastlowmark
de.newusers.infos 1587

1587 is false, the actual range in this group is 1779 - 1796.

I can correct the lowmark in active for 24 hours doing a "ctlinnd
renumber de.newusers.infos", but the next news.daily will set it
back to the incorrect value, even when deleting expire.lastlowmark.

I've searched the whole news system for "1587" and "de.newusers.infos"
and found it in history:

news at wuff:~ > grep de.newusers.infos db/history  | grep 1587
<de-newusers-infos/einleitung/20000628-1 at krell.zikzak.de>       962143201~970178402~962143202   de.newusers.infos/1587 de.answers/1765

news at wuff:~ > grep de.newusers.infos db/history  | grep 1588
<de-newusers-infos/einleitung/20000629-1 at krell.zikzak.de>       962229602~970264802~962229603   de.newusers.infos/1588 de.answers/1767

news at wuff:~ > grep de.newusers.infos db/history  | grep 1589
<de-newusers-infos/de-alt-newsgruppen/20000630-1 at krell.zikzak.de>       962316003~970351201~962316003   de.newusers.infos/1589


And so on ...

news at wuff:~ > grep -c de.newusers.infos db/history
234

All these articles (minus the handful actual articles in spool)
were superseded and their ID's should have been purged from history
a long time ago (/remember/ is 9).

news at wuff:~ > grep de.newusers.infos etc/expire.ctl 
de.newusers.infos:M:1:35:70

It seems that ID's for cancelled / superseded articles stay in
history for the complete period of 70 days (all articles had 
Expires headers with values "posting date + 3 months", so it's
the purge value from expire.ctl (70 days) that matches).

The articles are gone from spool, the overview data is gone, but
the ID's are still in the history. That's not nice because that
causes incorrect lowmark values and pumps up the history file
size ...

As I said, that's 2.2.3 (which I like a lot especially for smaller
boxes and my machines at home).

Bye,
Bettina