Bug with HTML messages?

[Alex Broens]

On 2010-07-29 21:04, Denis Beauchemin wrote:
> 
> 
> Le 2010-07-29 14:57, Alex Broens a écrit :
>> On 2010-07-29 20:07, Denis Beauchemin wrote:
>>>  Oops, I switched those 2:
>>>
>>>> SA scoring with HTML-only email: (not cached, score=12.639, requis 
>>>> 4.5, autolearn=spam, DEAR_SOMETHING 1.60, DRUGS_ERECTILE 0.28, 
>>>> DRUGS_ERECTILE_OBFU 1.23, DRUG_ED_CAPS 0.32, HTML_MESSAGE 0.00, 
>>>> KAM_VIAGRA1 3.00, KAM_VIAGRA5 3.10, KAM_VIAGRA6 3.10)
>>>>
>>>> SA scoring with HTML+TXT email: (not cached, score=7.181, requis 
>>>> 4.5, DCC_CHECK 2.17, HTML_MESSAGE 0.00, HTML_TAG_BALANCE_HEAD 1.33, 
>>>> MIME_HTML_ONLY 1.46, TVD_SPACE_RATIO 2.22)
>>>>
>>>
>>> SA scoring with HTML+TXT email: (not cached, score=12.639, requis 
>>> 4.5, autolearn=spam, DEAR_SOMETHING 1.60, DRUGS_ERECTILE 0.28, 
>>> DRUGS_ERECTILE_OBFU 1.23, DRUG_ED_CAPS 0.32, HTML_MESSAGE 0.00, 
>>> KAM_VIAGRA1 3.00, KAM_VIAGRA5 3.10, KAM_VIAGRA6 3.10)
>>>
>>> SA scoring with HTML-only email: (not cached, score=7.181, requis 
>>> 4.5, DCC_CHECK 2.17, HTML_MESSAGE 0.00, HTML_TAG_BALANCE_HEAD 1.33, 
>>> MIME_HTML_ONLY 1.46, TVD_SPACE_RATIO 2.22)
>>>
>>
>> you are not feeding it consistent content.
>> where is the problem?
>>
>> get a real spam wich was tagged as spam and dumped in your quaratine.
>> feed that to spamassassin (without MS), is the result consistent?
>>
>> if not, its usually due to MS's msg chunk settings
>>
>>
> 
> Alex,
> 
> When I feed the emails to SA they get scored much higher than through MS.
> 
> My point is that an HTML-only email with quite common spam words are not 
> being scored, while an HTML+TXT email with the same spam words get 
> scored.  This looks quite suspicious to me.
> 
> Denis
> 

what are your MS "chunk" settings?
are you sure MS is sending the full message to SA?

pls post the sample message you're using in pastebin so ppl can try to 
reproduce.

Alex