Path: utzoo!utgpu!news-server.csri.toronto.edu!clyde.concordia.ca!thunder.mcrcim.mcgill.edu!snorkelwacker.mit.edu!apple!julius.cs.uiuc.edu!rpi!uupsi!unvax.union.edu!boehme
From: boehme@unvax.union.edu (Eric M. Boehm)
Newsgroups: news.software.b
Subject: Expire is mangling the history database
Summary: expire is corrupting the dbz history database, but mkhistory fixes
Keywords: expire cnews dbz nntp
Message-ID: <1991Jan7.021551.16323@unvax.union.edu>
Date: 7 Jan 91 02:15:51 GMT
Organization: Union College, Schenectady, NY
Lines: 30

I am running cnews (patchdate 15-Dec-1990) with nntp.1.5.10 on a
DECsystem 5000/Model 200 running Ultrix 4.1. When doexpire runs (and
thus expire), the history database is corrupted.

This is my diagnosis because:
1. After expire rebuilds the history database, nntp fails on the next
connect after only 1 article.
2. Examining the coredump for nntpd shows that a Segmentation
violation occurred when trying to lookup the article in the database.
3. If I run "mkhistory" the problem goes away (that is, there are no
more core dumps and news runs okay).

For the time being, I have modified the doexpire script to run expire
with "-r" and run "mkhistory" to rebuild the database.

Is this normal behavior? Is anyone else having the same problem? One
possible explanation is that I selected dbz when building cnews but
selected ndbm for nntpd. I also selected fast stdio for cnews. I had
to copy /usr/include/stdio.h to cnews/libstdio so that rdwr.c would
compile.

I don't think that the dbz is a problem for nntpd since it works okay
after mkhistory does its task.

Thanks in advance for your assistance.

-- 
Eric M. Boehm
boehme@unvax.Union.EDU
BOEHME@UNION.BITNET