Desmond Coughlan | 1 Feb 13:56 2003

Re: Have There Been Changes in leafnode ?

Le Wed, Jan 29, 2003 at 10:48:46PM +0100, Matthias Andree a écrit ... 

{ snip }

> Well, you can derive it from the unix time (available also in Perl, for
> example) if you don't post more than once a second.

{ snip }

I'll find out tomorrow if it works.  I just redid all the headers with
perl.

Thanks. :-)

D.

--

-- 
Desmond Coughlan          |'Io non mori, e non rimasi vivo'
desmond@... 
http://www.zeouane.org

Lukasz Rzanek | 1 Feb 14:38 2003
Picon

Re: Fetching - problem

Użytkownik Matthias Andree napisał:
> Please download 1.9.32.rel or 1.9.33.devel3 and run rpm -tb 
> leafnode-1.9.32.rel.tar.bz2 (or 1.9.33.devel3.tar.bz2) on the
> tarball,
Upgrading wasn't a problem, except this is not my computer and I wanted
to keep up with official repository. I'm not a newbie, so I know how to
make my own package...

But the problem with those servers still exists. I got something like
this in log files:
Feb  1 14:35:01 szopka fetchnews[1788]: connected to 81.3.74.23:119,
reply: 200
Feb  1 14:35:01 szopka fetchnews[1788]: news.atozedsoftware.com: 0
articles posted
Feb  1 14:35:01 szopka fetchnews[1788]: Reading server info from
/var/spool/news/leaf.node/news.atozedsoftware.com
Feb  1 14:35:01 szopka fetchnews[1788]: Merging in
/var/spool/news/leaf.node/news.atozedsoftware.com~ from previous run
Feb  1 14:35:02 szopka fetchnews[1788]: atozedsoftware.intraweb.v5beta:
considering articles 1 - 3120
Feb  1 14:35:03 szopka fetchnews[1788]: Unable to parse

And it freezes.

I know this server is running on some windows software and it have
problems with Path (found 2 paths in many of posts from there), so I
think the problem is with the posts itselfs, not the server.

Anyway, I discovered that many windows servers will act like this, so I
think it could be considered as a bug :)
(Continue reading)

Robert | 1 Feb 14:51 2003
Picon
Picon

broken newsserver seeming to give more problems?

Since upgrading to leafnode-2.0.0.alpha20030128a I've noticed that when
one of my newsservers is having an off day, the water marks stay wrong
until another article is received for that group

I don't think I've seen this in earlier versions.

To expand slightly - I get

news.cache.ntlworld.com: connected (200), banner: "200 ntl NNTP news cache (feedback to nntptrial-feedback@...)"
news.cache.ntlworld.com: checking for new newsgroups
Reading new newsgroups failed
alt.comp.periphs: considering 53 articles 81368 - 81420, using XOVER
Unknown reply to XOVER command: 411 No such newsgroup
<sudoscript>: line 11:  3226 Broken pipe sudo -u news /usr/local/sbin/fetchnews -vvv

It then appears that no new articles have arrived and as some groups I
follow don't have many postings I have to wait quite some time - until yet
another posting - before I can read an incoming article when I see:

fetchnews[3380]: store: gmane.network.leafnode.announce: 
		 stale water marks, trying to fix

Is there anyway of manually fixing the watermarks - I can't see any
obvious options in fetchnews or texpire?

Robert

Doug Laidlaw | 1 Feb 12:13 2003
Picon

Re: Newsgroup seems corrupted

I have just got KDE 3.1 running.  The problem seems to have gone.  The thread 
is still there, but all threads including that one, behave correctly.

Doug.

On Thu, 30 Jan 2003 11:30 am, Doug Laidlaw wrote:
> Yes, Pan has no problem, and I have used sylpheed while I was using
> delaymode, but I prefer the GUI of Knode.  According to another posting,
> the problem was reported to KDE some time ago.  They had a chance to fix it
> with KDE3 (I am using KDE 3.0.3) but they haven't done anything about it,
>
> Doug.
>
> On Thu, 30 Jan 2003 08:50 am, you wrote:
> > Doug Laidlaw <laidlaws@...> writes:
> > > "Knode has a bug.
> > >
> > > "If header of the originating message refers back to itself, Knode gets
> > > thrown into a loop and winds up taking CPU time and grinding away at
> > > the hard drive.
> > >
> > > "There is such a message in alt.os.linux.mandrake.  It's in the 'rar
> > > file' thread."
> > >
> > > The "rar" thread is dated 28 January according to a Knode search (which
> > > doesn't open
> > > the article.)
> > >
> > > I have texpire set at 10 days and I haven't read the article he
> > > mentions, so I only need to wait.  I can even set it shotrer for this
(Continue reading)

Doug Laidlaw | 2 Feb 02:14 2003
Picon

Re: Newsgroup seems corrupted

I have just tested KDE 3.1.  My problem has gone.  Yours seems to be a 
different one.

Doug.

On Thu, 30 Jan 2003 03:14 am, ketil vestby wrote:
> <-----Original Message----->
>
> > From: Matthias Andree
> > Sent: 30/1/2003 1:07:24 AM
> > To: leafnode-list@...
> > Subject: Re: [leafnode-list] Newsgroup seems corrupted
> >
> > "Robert Marshall" writes:
> >
> > On Wed, 29 Jan 2003, Doug Laidlaw wrote:
> > >> I am using Leafnode 1.9.24 with Mandrake 9.0.
> > >>
> > >> Suddenly, in Knode, the group alt.os.linux.mandrake makes Knode
> > >> "freeze" whenever I try to expand any thread. I can choos single
> > >> articles O.K. It only happens in that one group, and it doesn't
> > >> happen with Pan.
> >
> > This is a problem with knode and someone who has posted a self
> > referencing article in that group - there's a thread on the subject
> > there but you might find it a little difficult to read!!
> >
> > Do you happen to have any references (URL) I could add to the leafnode
> > FAQ before I release 1.9.33?
> >
(Continue reading)

Matthias Andree | 2 Feb 02:42 2003
Picon
Picon

Re: Newsgroup seems corrupted

Doug Laidlaw <laidlaws@...> writes:

> I have just got KDE 3.1 running.  The problem seems to have gone.  The
> thread is still there, but all threads including that one, behave
> correctly.

Thanks for telling us. Have fun.

--

-- 
Matthias Andree
leafnode-1 download: http://sourceforge.net/projects/leafnode/
leafnode-1 docs/new: http://mandree.home.pages.de/leafnode/
leafnode-2 homepage: http://mandree.home.pages.de/leafnode/beta/

Matthias Andree | 2 Feb 02:45 2003
Picon
Picon

Re: broken newsserver seeming to give more

Robert <robert@...> writes:

> To expand slightly - I get
>
> news.cache.ntlworld.com: connected (200), banner: "200 ntl NNTP news cache (feedback to nntptrial-feedback@...)"
> news.cache.ntlworld.com: checking for new newsgroups
> Reading new newsgroups failed
> alt.comp.periphs: considering 53 articles 81368 - 81420, using XOVER
> Unknown reply to XOVER command: 411 No such newsgroup
> <sudoscript>: line 11:  3226 Broken pipe sudo -u news /usr/local/sbin/fetchnews -vvv

This pretty much looks like a problem in the communication between your
fetchnews and the upstream -- and because leafnode doesn't catch
SIGPIPE, it aborts, and as a consequence, the water marks (actually a
misnomer, they are rather levels) are stale, a stale lock file is lying
around and all that.

fetchnews and the NTL server are getting out of sync, "no such
newsgroup" is not a valid reply to XOVER -- can you set debugmode=129 in
your config file reproduce the problem and send me the resulting debug
log (don't send to the list). I may be able to fix some of the problem.

> It then appears that no new articles have arrived and as some groups I
> follow don't have many postings I have to wait quite some time - until yet
> another posting - before I can read an incoming article when I see:
>
> fetchnews[3380]: store: gmane.network.leafnode.announce: 
> 		 stale water marks, trying to fix
>
> Is there anyway of manually fixing the watermarks - I can't see any
(Continue reading)

Matthias Andree | 2 Feb 02:48 2003
Picon
Picon

Re: Fetching - problem

Lukasz Rzanek <lukasz@...> writes:

> But the problem with those servers still exists. I got something like
> this in log files:
> Feb  1 14:35:01 szopka fetchnews[1788]: connected to 81.3.74.23:119,
> reply: 200
> Feb  1 14:35:01 szopka fetchnews[1788]: news.atozedsoftware.com: 0
> articles posted
> Feb  1 14:35:01 szopka fetchnews[1788]: Reading server info from
> /var/spool/news/leaf.node/news.atozedsoftware.com
> Feb  1 14:35:01 szopka fetchnews[1788]: Merging in
> /var/spool/news/leaf.node/news.atozedsoftware.com~ from previous run
> Feb  1 14:35:02 szopka fetchnews[1788]: atozedsoftware.intraweb.v5beta:
> considering articles 1 - 3120
> Feb  1 14:35:03 szopka fetchnews[1788]: Unable to parse
>
> And it freezes.

Can you obtain a stack backtrace?

Run: typescript
then type: gdb fetchnews 1788 (replace 1788 by the process ID of fetchnews)
then type: backtrace full
then type: quit
then type: exit or logout (depending on your shell).

Then please send me the resulting typescript file.

It may also help if you set debugmode=1 in your config file before you
do this, and also send me the debug log (see the README file, section
(Continue reading)

clemens fischer | 2 Feb 13:27 2003

Re: broken newsserver seeming to give more

Matthias Andree <ma@...>:

> This pretty much looks like a problem in the communication between your
> fetchnews and the upstream -- and because leafnode doesn't catch
> SIGPIPE, it aborts, and as a consequence, the water marks (actually a
> misnomer, they are rather levels) are stale, a stale lock file is lying
> around and all that.

btw, i never really understood the waterlevels.  could somebody
please explain what they mean?

  clemens

Matthias Andree | 2 Feb 15:14 2003
Picon
Picon

Re: broken newsserver seeming to give more

clemens fischer schrieb am Sunday, den 02. February 2003:

> Matthias Andree <ma@...>:
> 
> > This pretty much looks like a problem in the communication between your
> > fetchnews and the upstream -- and because leafnode doesn't catch
> > SIGPIPE, it aborts, and as a consequence, the water marks (actually a
> > misnomer, they are rather levels) are stale, a stale lock file is lying
> > around and all that.
> 
> btw, i never really understood the waterlevels.  could somebody
> please explain what they mean?

OK, news readers have two means to access an article:

1. by message-id (uncommon for newsreaders except for the "lookup by
   message-id" function and maybe for references that cannot be
   satisfied in the currently displayed newsgroup)

2. by GROUP and ARTICLE commands, where the ARTICLE commands take
   numbers that are server- and group-specific -- and also deposited in
   the Xref: header that isn't propagated between servers.

It is assumed by newsreaders that the article numbers are strictly
monotonically increasing, and therefore, leafnode stores them in the
groupinfo file.  The levels (low/high) note the lowest (less critical)
and highest (critical) article number available on the server, where the
highest number must not be lowered even if articles with high numbers
expire or are cancelled (only leafnode-2 actually executes cancels,
leafnode-1 just transports them as regular postings).
(Continue reading)


Gmane