gst@gnosys.UUCP (Gary S. Trujillo) (10/04/89)
Recently, the following was posted: |Path: gnosys!spdcc!bloom-beacon!shelby!mcnc!gatech!emory!skeeve!arnold |From: arnold@skeeve.UUCP (Arnold D. Robbins) |Newsgroups: unix-pc.sources |Subject: Patch #18 to B News 2.11 |Message-ID: <253@skeeve.UUCP> |Date: 17 Sep 89 04:08:30 GMT |Distribution: unix-pc |Organization: The Basement Computer, Atlanta GA |Lines: 2254 | |Well, believe it or not, B News is NOT dead! (Merely on heavy life support.) |Despite the release of C News and the imminent release of B News 3.0 |(TMN-Netnews), Rick Adams has released the following patch to B News. | |Skeeve is still at 2.11.14. I haven't decide yet which news I'm going |to switch to, or when. At the moment, it ain't broke so I ain't fixing it. |------------------------------- |Path: emory!gatech!udel!wuarchive!cs.utexas.edu!uunet!rick |From: rick@uunet.UU.NET (Rick Adams) |Newsgroups: news.software.b |Subject: patch 18 for B news 2.11 |Message-ID: <67503@uunet.UU.NET> |Date: 12 Sep 89 23:10:59 GMT |Organization: UUNET Communications Services, Falls Church, VA |Lines: 2232 | |Description: | This is patch 18 for news 2.11. It addresses the following problems: | | Small changes to work around various compiler bugs. | fixed batch to return to BATCHDIR before renaming work file. | checkgroups no longer mails you a message if everything's ok | Fixed active file to handle a maximum of 9,999,999 messages | to avoid wraparound problem at current 99,999... Well, I naively patched my sources, which were at patchlevel 17, and got a very corrupted situation in my /usr/spool/news/* directories. Here's the sort of thing I found: 2 -rw-r--r-- 2 news 1022 Oct 3 15:34 41414 5 -rw-r--r-- 1 news 2093 Oct 3 15:33 414 3 -rw-r--r-- 1 news 1330 Oct 3 15:33 4 4 -rw-r--r-- 1 news 1580 Oct 3 15:33 31414 3 -rw-r--r-- 1 news 1490 Oct 3 15:33 314 2 -rw-r--r-- 1 news 808 Oct 3 15:33 3 2 -rw-r--r-- 1 news 530 Oct 3 15:33 21414 2 -rw-r--r-- 3 news 958 Oct 3 15:32 214 2 -rw-r--r-- 1 news 704 Oct 3 15:32 2 4 -rw-r--r-- 1 news 1651 Oct 3 15:31 5714 D'ya see a pattern?? Looks like it's taking the last two digits of the last valid article filename, and plays games with them. Worse yet, rn goes crazy when there's such a wide article-sequence spread (runs out of memory on my 2 meg machine, in fact). Here's what we find in /usr/lib/news/errlog: Oct 3 15:32 <1989Oct2.175907.8836@alberta.uucp> inews: found bad ngsize 359695 ng misc.jobs.misc, setting to 1 Oct 3 15:32 <195@temvax.UUCP> inews: found bad ngsize 292827 ng misc.jobs.offered, setting to 1 Oct 3 15:33 <Oct.2.17.13.00.1989.27704@topaz.rutgers.edu> inews: found bad ngsize 2969695 ng misc.jobs.misc, setting to 1 Oct 3 15:33 <4826@cps3xx.UUCP> inews: found bad ngsize 2141413 ng comp.unix.xenix, setting to 1 Oct 3 15:33 <ROB.89Oct2163903@violet.berkeley.edu> inews: found bad ngsize 2434342 ng soc.men, setting to 1 Oct 3 15:33 <ROB.89Oct2163903@violet.berkeley.edu> inews: found bad ngsize 3969695 ng misc.jobs.misc, setting to 1 Oct 3 15:33 <37@gutarman.UUCP> inews: found bad ngsize 3141413 ng comp.unix.xenix, setting to 1 Oct 3 15:34 <216@hutch.UUCP> inews: found bad ngsize 4969695 ng misc.jobs.misc, setting to 1 Oct 3 15:34 <334@mlfarm.UUCP> inews: found bad ngsize 4141413 ng comp.unix.xenix, setting to 1 Oct 3 15:35 <13915@well.UUCP> inews: found bad ngsize 2282827 ng misc.jobs.offered, setting to 1 I have promptly fallen back to patchlevel 17! -- Gary S. Trujillo gst@gnosys.svle.ma.us Somerville, Massachusetts {wjh12,spdcc,ima,cdp}!gnosys!gst
gst@gnosys.UUCP (Gary S. Trujillo) (10/04/89)
In article <392@gnosys.UUCP> gst@gnosys.UUCP (Gary S. Trujillo) writes: > ...I naively patched my sources, which were at patchlevel 17, and got > a very corrupted situation in my /usr/spool/news/* directories... > the sort of thing I found: Before anyone asks, I thought I should say that I did follow the instructions below: | make update now checks to see if you have defined LINES large | enough to support your existing active file | |Fix: | cd src and make install;make update | | YOU MUST "make update" or your active file will be in the wrong | format -- Gary S. Trujillo gst@gnosys.svle.ma.us Somerville, Massachusetts {wjh12,spdcc,ima,cdp}!gnosys!gst
gst@gnosys.svle.ma.us (Gary S. Trujillo) (10/05/89)
In article <1676@mtunb.ATT.COM> jcm@mtunb.UUCP (was-John McMillan) writes: > In article <392@gnosys.UUCP> gst@gnosys.UUCP (Gary S. Trujillo) writes: > >Recently, the following was posted: > : > >... Worse yet, rn goes crazy > >when there's such a wide article-sequence spread (runs out of memory on my > >2 meg machine, in fact). > > There is little relationship between having 2 Meg of RAM and > running out of memory: Virtual Memory means not having to say > "You're outta RAM" -- you're just out of Virtual Address [VA] space > or out of SWAP space... Yeah, yeah. I just left out the details - that's all. I realize that what got exhausted was swap space. Still, my point was that there was a problem with the situation I encountered. BTW, the story has a happy ending: turns out I accidentally let an old (pre-patch 18) version of rnews deal with a new batch of news. It didn't know how to deal with the new-format active file. Thanks to Gordon Burditt for putting me on the track of the problem: |Date: Wed, 4 Oct 89 10:35:05 cdt |From: ima!harvard!husc6!wuarchive!sneaky!gordon (Gordon Burditt) |Subject: Re: B news 2.11 patch 18 considered harmful |In-Reply-To: <392@gnosys.UUCP> |Organization: Gordon Burditt | |This is the sort of thing I saw while installing patchlevel 18 when I |accidentally let loose the OLD rnews with the NEW active file format. |(A batch sort of snuck in during the installation, and rnews is a copy, |not link, to inews, because they are on different file systems, so |it was a little slower getting both changed). | |Once I cleaned up the mess in the active file, it's been working fine. | | Gordon L. Burditt | ...!texbell!sneaky!gordon Same here. Thanks, Gordon! -- Gary S. Trujillo gst@gnosys.svle.ma.us Somerville, Massachusetts {wjh12,spdcc,ima,cdp}!gnosys!gst
miw@bunyip.cc.uq.oz (Mark Williams) (10/05/89)
gst@gnosys.UUCP (Gary S. Trujillo) writes: >Well, I naively patched my sources, which were at patchlevel 17, and got >a very corrupted situation in my /usr/spool/news/* directories. Here's >the sort of thing I found: [horrendous situation caused by a corrupt active file] I had this problem, too. I got around it by blowing away my active file entirely, running expire -r, and then putting a current checkgroups through the thing. The make update script failed because a regular expression was too long for ULTRIX 3.1 sed. I also had to back out one of the patches to ifuncs.c Apart from that, everything ran smoothly. :-) Mark Williams