Okay did some more testing:

both mails, same ruleset but this time using -L to disable network
checks. Scantime on 4.0.2 is down

Mail A (looot of URI) from 2.5min to 19s
Mail B (funny html) from 1min to 6s

on 4.0.1 both mails do not show much different timings between with -L
and without

Cheers

tobi

On Wed, 2025-09-03 at 09:56 +0200, [email protected] wrote:
> On 9/3/25 9:04 AM, Tobi wrote:
> > Hello list
> > 
> > since beginning of this week we use the new spamassassin 4.0.2
> > Since
> > then we can see that processing time (at least for particular
> > mails)
> > massively increased. We use a timeout of 30s in our setup to
> > communicate with SA.
> > 
> > Two mails we now have that cannot be processed anymore in that time
> > and
> > therefore get stuck in our processing.
> > 
> > Mail A has a huge load of links/mailtos in header and body. About
> > 7k.
> > The timing in processing with the very same rules is from 6s (in
> > 4.0.1)
> > to 2.5min (in 4.0.2). I counted the number of URI spamassassin
> > detected
> > and the number of DNS queries in both versions and they are more or
> > less equal. So I guess the increase in processing time is not DNS
> > related.
> > 
> > Mail B does not have that many URI but a quite complex HTML
> > attachment
> > with a load of styles with almost every div and span tag. The
> > processing time is 4s (in 4.0.1) and about 1min (in 4.0.2).
> > 
> > I will check in our timings logs to see if we see a general
> > increase in
> > timings or if it's more only related to specific mails.
> > 
> > Have a good one
> > 
> Hi,
> one commit that can degrade performance is this commit
> https://github.com/apache/spamassassin/commit/4a6643c72fe5cd6296b2e67eb0ccc368a4987caa
> This tries to find CNAMEs for all hosts found in the email.
>  From my tests this can slow down email processing a bit but not that
> much.
> Could you send me those emails in private ?
>   Thanks
>    Giovanni

Reply via email to