Severe xover corruption problems on INN2.2.1

Hans Lambermont Hans.Lambermont at garg.mpn.cp.philips.com
Mon Oct 11 09:47:38 UTC 1999


Katsuhiro Kondou wrote:

> In article <19991008135856.B26554 at garg.mpn.cp.philips.com>,
> 	Hans Lambermont <Hans.Lambermont at garg.mpn.cp.philips.com> wrote;
> } Yes, it appears my problems are restricted to nl.* and news.* but with
> } the sole use of cycbuffs for these groups ... ?
> That's very weird, since the problem is not relevant to cycbuff
> but uniover.

But because of my makehistory run this uniover info came from my
cycbuffs. And all articles with corrupt xover info have correct headers
in the cycbuffs.

> According to your another mail, if the problem happens to those,
> other newsgroups whose name start with 'n' overview could be
> corrupted.

Correct. As all n* newsgroups are in just one uniover file
(/news/uniover/13/overview) i just had a look at it and immediately
found errors in it (see below). But not all n* groups have xover errors,
fi, ne.jobs has > 1000 articles in it here and show no xover errors ...

Re: Can a good catholic be saved?       Timothy Consodine <timothyc1 at yahoo.com> 
28 Sep 1999 13:05:59 +0200      <37DD998A.BC40BFAC at yahoo.com>   <7ropo6$rh4$1 at nn
rp1.deja.com> <7sodo1$5qc$1 at zonnetje.nl.uu.net> <slrn7v169h.187.pino+nl_newsgrou
ps at mud.stack.nl> <37F09AA1.68679F7F at xs4all.nl>          22      Xref: news.eur.c
is.philips.com nl.burgerrechten:54797 nl.juridisch:25666 nl.newsgroups:68709 nl.
politiek:171918 nl.wetenschap:78160
SV: since you suck so much      "OLAV N<D8>TTVEIT" <olav.nottveit at c2i.net>      
Tue, 28 Sep 1999 16:05:29 EST   <441415-25681dsi$160528.920000 at news.nac.net>    
<7rhavf$gik$1 at nnrp1.deja.com>           32      Xref: news.eur.cis.philips.com n
csc.general:410
Annihilator New Binary-EMP Report binnew.19990915004019 @@NCM   Annihilator-1 at er
lenstar.demon.co.uk     Tue, 14 Sep 1999 19:40:20 -0500 <nocem-binnew-1999091500
4020$0754 at news-out.riddles.org.uk>      <37EF77E9.1F8ECA59 at gmx.de> <7sq35b$isu$1
@fu-berlin.de> <37fcb949.5927168 at news.stgl.sel.alcatel.de> <7sqfgh$p6$1 at fu-berli
n.de>           86      Xref: news.eur.cis.philips.com news.li

tba.pbz>        1932    20      Xref: news.eur.cis.philips.com news.admin.net-ab
use.usenet:337763

09520.25716-100000 at hepp.uio.no> 1392    11      Xref: news.eur.cis.philips.com n
o.alt.frustrasjoner:91803

snip

Last two entries are clearly wrong. So probably my uniover problem is
limited to this file. If i'm correct this uniover format is removed in
2.3, so it is not very useful to create an 'uniover format checker'.
Also a makehistory run for just one uniover file would be handy, if not
for the removal of the uniover in 2.3 ... :-/

Note that most of this corrupt uniover file was generated by my
makehistory run. So simply running it again won't help. I think my
makehistory run is hindered by the same bug that caused the uniover
corruption in the first place.

Another thing that might be interesting; the xover estimate of the
number of articles in a group is way off. F.i.
group nb.biz
211 10 930 970 nb.biz
has only 3 articles in it, not 10.

Any ideas ? I'm reluctant to upgrade to 2.3, although it might solve our
problem. Do you have an idea on how to find the real problem ? (our
number of user complaints is rising :(

And thanks to you all for your time of course ;)

regards,
   Hans Lambermont


More information about the inn-workers mailing list