[Fwd: [SAtalk] Bad bug in Perl or spamd.]

[Gerry Doris]

I included the following message for the sake of those using RH 8.0 and
spamassassin.  It sounds like Redhat's use of UTF-8 is causing some
interesting problems for spamassassin.

Gerry


-------- Original Message --------
Subject: [SAtalk] Bad bug in Perl or spamd.
From: Gustave Eiffel <gustave at 3web.net>
Date: Fri, October 25, 2002 1:11 pm
To: Spamassassin-talk at lists.sourceforge.net

This will come up alot on RedHat 8.0 users and high volume users.....

After much trouble shooting and trial.....

All users running perl 5.8 and spamassassin will find this:
I am running RedHat 8.0 with spamassassin ( was 2.32 that comes with
RedHat 8.0 ) 2.43 and the default Perl 5.8.

I receive about 100,000 emails a day on each server.  I tried using
spamassassin but always found that after a few minutes that the CPU
would load way up ( 30 + ) on spamd and essentialy kill the server.

I saw UTF-8 encoding errors:

Oct 24 13:37:09 blacktip spamd[24511]: Malformed UTF-8 character
(unexpected
> > > continuation byte 0xa0, with no preceding start byte) in
> transliteration (tr/ //)
> > > at /usr/lib/perl5/site_perl/5.6.1/Mail/SpamAssassin/EvalTests.pm
> line 1787, <STDIN> line 199.

but saw that this should not be an issue beyond the error message.  This
is not correct.

I went to the bugzilla site and looked at bug # 1147

TEMPORARY WORKAROUND:

You can change your locale to regular en_US (en_US.iso885915) instead of
en_US.UTF-8 and this seems to fix the problem.  RedHat 8 defaults to
using UTF-8 locales.  You can use locale_config if you have XFree86
installed, or you can just edit your /etc/sysconfig/i18n file.



and put the change in place.

I not only no longer get the error but it has fixed the CPU load
problem.  I rarely ever see a load much above .4 and never see spiked in
CPU at all.

I am not sure if this is a Perl 5.8 problem or a SA problem but seeing
as how RedHat 8.0 now includes SA this will come up alot.  It may go
un-noticed in low volume sites but SA will become known as a CPU pig for
sure.  This is a great product and this should be fixed ASAP and a patch
submited to RedHat for thier errata to avaoid this I think.

Thoughts?

Thanks all for your help over the last few days with this BTW!

Mark