I have verified, to the best of my ability, that the spamassassin -W command is
not functional, at least to the extent that it does not impact scores used by
the AWL mechanism during mailchecks.

It appears that spamassassin -aW <sample_nonspam.txt adds email addresses,
whereas  spamassassin -a uses the email address in conjunction with originating IP.

It would appear from this testing that -W needs to generate the same format of
AWL db entries as runs of email through with the AWL on generate. They currently
are not the same.


Both tools maintain their own "average score" entries. Wntries created by
running the AWL are only affected by checking emails, and are not affected by
runs of -W, and vice versa.

Short summary with relevant log bits (full logs will be attached)

I ran spamassassin -aRD <sample_nonspam.txt 

 debug: auto-whitelist (db-based): dawson@world.std.com scores 0/0

I ran spamassassin -aWD <sample_nonspam.txt:

 debug: auto-whitelist (db-based): dawson@world.std.com scores 0/0
 debug: add_score: New count: 1, new totscore: -100

I ran spamassassin -aD <sample_nonspam.txt:

 debug: auto-whitelist (db-based): dawson@world.std.com|ip=208.192 scores 0/0
 debug: AWL active, pre-score: 1.9, mean: undef, originating-ip: 208.192.102.199
debug: Post AWL score: 1.9
 debug: add_score: New count: 1, new totscore: 1.9


I also repeated the -aWD and -aD steps to demonstrate that the counts and scores
used by each tool are self maintained and not affected by the other.See
attachments for details.

Note: I'm no perl programer, so I don't get the syntax, but it would appear this
is the cause:

EvalWhitelist.pm the sub check_address will modify self->{entry} to reflect
$addr."|ip=.$origip" and that is used when fetching previous scoring and
updating averages.


SpamAssassin.pm uses "find_all_addrs_in_mail" and passes these to
add_known_good_address, find_all_addrs_in_mail does not have any kind of origip
handling, nor does add_known_good_address.