[Mimedefang] Correction about Autolearn Headers

[Kevin A. McGrail]

Many persons, including myself are interested in having the headers in a
MD+SA installation be as similar to SA's original headers as possible.

One of the differences I never solved was the autolearn status for Bayesian
tokens.  I was forwarded some code that originated on the SATalk email which
I unfortunately took on faith because it looked correct.

Unfortunately, after researching this further, it is my opinion that the
auto_learn_status is not available in the MimeDefang SubFilter and certainly
not through the SASpamTester Variable.  Whoever posted the original email on
SATalk had this completely wrong for a MIMEDefang installation.  It is
however, correct logic for the SA Learn status variable.

To go further, the SA Learn Status is a function of the SpamAssassin
PerMsgStatus that gets returned from a call to SpamAssassin's check()
function.

To clarify this in terms of MIMEDefang, the call to SA's check routine is
made in the subroutine spam_assassin_status and the status variable is
returned to spam_assassin_check where it is scoped locally AND finished
prior to executing sub spam_assassin_check.

Therefore, the only way to make this work that I can see is to modify
mimedefang.pl.  It would be a simple change to return $status, comment out
the $status->finish, and return $status to the subfilter where you would
need to run a $status->finish on it in the subfilter.  I will leave that to
Roaring Penguin to decide if they want to make that change.

Regards,
KAM