http://bugzilla.spamassassin.org/show_bug.cgi?id=3997
------- Additional Comments From [EMAIL PROTECTED] 2005-04-12 02:55 ------- Sidney, That patch does disable persistence the way I did it. Irrespective of any ad hoc testing, I assure you that patch does fix problems, at least in my environment. That's not to say there aren't other problems around, or problems in other people's environments. I debugged the problem I had, and the need to disable persistent TCP/UDP is not a guess, but absolutely correct even if possibly not sufficient. See my description in bug 4249. Oh, you've made another reply... yes bg_send creates a new socket each time. However the first DNS query is made using "search" which ends up calling "send_udp" not bg_send. The issue is that the port used by that lookup (the one using "search") is reused (due to persistent_udp) for the first lookup using bg_send. Subsequent lookups with bg_send will indeed create a new port. It's the first use of bg_send that fails. The effects of this were all directly observed by using "tcpdump port domain" and "spamassassin -D" and also setting the debug flag in Net::DNS. Producing stats of lots of runs that are just saying "it's broke" isn't terribly useful - at least not unless you know what lookups actually did happen. I strongly recommend that anyone debugging this problem also run "tcpdump port domain" in another window and when there's a false positive, look (and post here) the excerpt from around that time to confirm that the correct DNS requests and responses are being made, and in light of my problem, with the correct sockets. Even better if all the debug info from above is also turned on. ------- You are receiving this mail because: ------- You are the assignee for the bug, or are watching the assignee.
