[Spambayes] Spambayes 1.1a1 too many problems?

[Tony Meyer]

> 1) Spam/Ham training through the online server did not cause 
> any change in the statistics but all messages ended up in the 
> pop3proxy-ham-cache or pop3proxy-spam-cache directory 
> depending whether the messages were trained as ham or spam.

Someone else already reported this one, and it's fixed in CVS.  The
statistics would actually get updated, but only next time you started
sb_server (i.e. the persistent bit worked, but not the current session bit).

> 2) Once all the training messages ended up in the cache 
> directories, it took ages to start the Spambayes server.

Longer than with the same number of messages in the cache with 1.0.x?  I
can't think of any reason that would be.  If anyone else is reading this and
still using 1.1a1, do you find this too?  Is it actually related to the
number of files in the caches, or something else?

> 3) Intermittently, I saw e-mail message headers with the 
> following errors:
> X-Spambayes-Exception: Traceback (most recent call last):
[...]
>  .DBRunRecoveryError: (-30978, 'DB_RUNRECOVERY: Fatal error,
>  run database recovery -- PANIC: Invalid argument') 

As Tim said, this means that the database is dead, and you have to retrain
from scratch.  Did you start with the same database you had with 1.0.x, or
delete that and start fresh?  This problem occurs with 1.0.x, too (and is
one of the reasons for 1.1a1 including other database options), although not
often, and most cases are caused when something unsupported (concurrent
access from two spambayes processes, for example) is done.  Was there
anything unusual about the way you ran 1.1a1?  Was 1.0.x definitely not
running at the same time?

> 4) Error messages in the logfile like:
[...]
> error: uncaptured python exception, closing channel 
> <spambayes.Dibbler._HTTPHandler connected at 0x98437b0> 
> (socket.error:(10053, 'Software caused connection abort') 
> [asynchat.pyc|handle_read|89] [asyncore.pyc|recv|343])
> Almost immediately after this error, the Page File increased 
> to over 1.3GB!!! and my system slowly died. 

Again, if anyone else is reading this, have you experienced anything like
this?  It's hard to track down from a single report like this; it would be
much easier if someone could keep running 1.1a1 while we figure out what the
problem is.  I don't suppose you know *when* the messages appeared in the
log?  (i.e. when you were just using the web interface, when mail was being
checked, when training was occurring).  Is your network connection
unreliable? (My best guess is that there were regular problems with
connecting to the POP3 server, but I need more information to know for
sure).

As Tim said, thanks for the report.

=Tony.Meyer

-- 
Please always include the list (spambayes at python.org) in your replies
(reply-all), and please don't send me personal mail about SpamBayes.
http://www.massey.ac.nz/~tameyer/writing/reply_all.html explains this.