Milter-greylist

Bruncsak, Attila <attila.bruncsak@...> wrote:

> I do not see any bug.

I tracked it down to this minimal example:

racl blacklist rcpt archibald.haddock@... msg "blacklisted"
racl whitelist auth /.*/ report "Authenticated sender"
racl whitelist default

Sending as an authenticated user, I get the following result:
archibald.haddock@... -> blacklisted, as expected
emmanuel.dreyfus@... -> accepted

Now if I start sending with an accepted user first:
emmanuel.dreyfus@... -> accepted
archibald.haddock@... -> wrongly accepted

My understanding is that the offending code is at the begining of
real_envrcpt():
        
        if ((priv->priv_sr.sr_whitelist & EXF_WHITELIST) &&
            (priv->priv_sr.sr_whitelist &
             (EXF_NONIP | EXF_AUTH | EXF_STARTTLS | EXF_SPF)))
                goto exit_accept;

That reuses the status from previous recipient. I made the change below,
which avoids it based on a global configuration parameter, but I wonder
if it makes sense to preserve the original behavior. Is it just a plain
bug, or can it have some merit?

--- milter-greylist-4.5.1/milter-greylist.c 
+++ milter-greylist-4.5.1p1/milter-greylist.c 
@@ -638,8 +638,16 @@
         */
        prop_clear(priv, UP_CLEARPROP);
 #endif
 
+       /*
+        * If we re-evaluate racl for each recipient, forget
+        * about previous decision.
+        */
+       if (conf.c_multiracl)
+               priv->priv_sr.sr_whitelist &=
+                    ~(EXF_WHITELIST|EXF_GREYLIST|EXF_BLACKLIST);
+
        if ((priv->priv_sr.sr_whitelist & EXF_WHITELIST) &&
            (priv->priv_sr.sr_whitelist &
             (EXF_NONIP | EXF_AUTH | EXF_STARTTLS | EXF_SPF)))
                goto exit_accept;


-- 
Emmanuel Dreyfus
http://hcpnet.free.fr/pubz
manu@...