This answer does NOT really relate to bayes!
But nevertheless to a typical 'out of memory'.

On Tue, Jan 20, 2004 at 01:43:30PM +0100, Daniele Nicolodi wrote:
> I have also tryed to upgrade to sa 2.62 and convert the old db format to
> the new one but sa fill all my memory, and conseguently is killed by oom
> killer on the pass two of the proces.

My 'own way' to create this Problem always was to create
selfmade rules which contained 'badly behaving regexps'.
Those are especially all things where 'multiply allowed
somethings' are again 'multiply allowed', like

  /...(...[abcde]+...)*/ or /.....(\S+)*.../ or something like
  /(word1.*word2|word3.*word4)/

In those cases the regexp-analyser walks through horrendous numbers
of alternatives which all have to be 'remembered' until it is known
whether to fail that pattern.

As 'spamassassin' reads the whole body as ONE string, those
'remembered' partial decisions bloated my spamassassin up to
750M for a 2M Mail (crashing the process shortly after).

So I had to fix my private ruleset to never use 'unlimited'('*')
alternatives but ALWAYS and ONLY use 'reasonable limits' like
/...\S{1,50}.../.

May be this helps? 

Stucki


-------------------------------------------------------
The SF.Net email is sponsored by EclipseCon 2004
Premiere Conference on Open Tools Development and Integration
See the breadth of Eclipse activity. February 3-5 in Anaheim, CA.
http://www.eclipsecon.org/osdn
_______________________________________________
Spamassassin-talk mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/spamassassin-talk

Reply via email to