Mailing List Archive

message with autolearn=no is ignored by sa-learn
Hi,

On a machine with spamassassin 3.4.6 under Debian 11, a new spam
arrived, and the headers showed:

X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on joooj.vinc17.net
X-Spam-Level: ****
X-Spam-Status: No, score=4.9 required=5.0 tests=BAYES_50,HTML_IMAGE_RATIO_02,
HTML_MESSAGE,KHOP_HELO_FCRDNS,MIME_HTML_ONLY,SPF_HELO_NEUTRAL,
SPF_NEUTRAL,T_SCC_BODY_TEXT_LINE,URIBL_BLACK,YOUR_DELIVERY_ADDRESS
autolearn=no autolearn_force=no version=3.4.6
X-Spam-Language: en

Since it has autolearn=no, I assume that it wasn't learnt as spam.
So I piped it to "sa-learn --spam --no-sync", but I got

Learned tokens from 0 message(s) (1 message(s) examined)

Why "from 0 message(s)"?

--
Vincent Lef?vre <vincent@vinc17.net> - Web: <https://www.vinc17.net/>
100% accessible validated (X)HTML - Blog: <https://www.vinc17.net/blog/>
Work: CR INRIA - computer arithmetic / AriC project (LIP, ENS-Lyon)
Re: message with autolearn=no is ignored by sa-learn [ In reply to ]
On 2022-06-14 12:13:10 +0200, Reindl Harald wrote:
> Am 14.06.22 um 11:52 schrieb Vincent Lefevre:
> > On a machine with spamassassin 3.4.6 under Debian 11, a new spam
> > arrived, and the headers showed:
> >
> > X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on joooj.vinc17.net
> > X-Spam-Level: ****
> > X-Spam-Status: No, score=4.9 required=5.0 tests=BAYES_50,HTML_IMAGE_RATIO_02,
> > HTML_MESSAGE,KHOP_HELO_FCRDNS,MIME_HTML_ONLY,SPF_HELO_NEUTRAL,
> > SPF_NEUTRAL,T_SCC_BODY_TEXT_LINE,URIBL_BLACK,YOUR_DELIVERY_ADDRESS
> > autolearn=no autolearn_force=no version=3.4.6
> > X-Spam-Language: en
> >
> > Since it has autolearn=no, I assume that it wasn't learnt as spam.
> > So I piped it to "sa-learn --spam --no-sync", but I got
> >
> > Learned tokens from 0 message(s) (1 message(s) examined)
> >
> > Why "from 0 message(s)"?
>
> because the exactly same message was likely already trained

When? The message has autolearn=no, so it wasn't trained when
passed via SpamAssassin while it was received. Then it was in
my main mailbox, where there's no training.

--
Vincent Lef?vre <vincent@vinc17.net> - Web: <https://www.vinc17.net/>
100% accessible validated (X)HTML - Blog: <https://www.vinc17.net/blog/>
Work: CR INRIA - computer arithmetic / AriC project (LIP, ENS-Lyon)
Re: message with autolearn=no is ignored by sa-learn [ In reply to ]
On 2022-06-14 13:39:29 +0200, Reindl Harald wrote:
> Am 14.06.22 um 13:36 schrieb Vincent Lefevre:
> > When? The message has autolearn=no, so it wasn't trained when
> > passed via SpamAssassin while it was received. Then it was in
> > my main mailbox, where there's no training
>
> spam messages often appear more than once

I had searched for other copies in my "junk" mailbox, but forgot
that when the spam score was too high, I was directly discarding
the message instead of storing it to this mailbox.

Indeed, I can see in my logs that I received several of them,
including with autolearn=spam, with very high spam scores for
all of them.

Sorry for the noise.

--
Vincent Lef?vre <vincent@vinc17.net> - Web: <https://www.vinc17.net/>
100% accessible validated (X)HTML - Blog: <https://www.vinc17.net/blog/>
Work: CR INRIA - computer arithmetic / AriC project (LIP, ENS-Lyon)