http://issues.apache.org/SpamAssassin/show_bug.cgi?id=4477
------- Additional Comments From [EMAIL PROTECTED] 2006-06-29 18:06 -------
hmm. this may need some work still. here's COlm's comment I missed:
> The attached patch looks sorta o.k., but there are some easy fixes that
> should be included;
>
> Instead of
>
> + char host[255]; /* hostname, for logging */
> + char port[255];
>
> there should be;
>
> char host[NI_MAXHOST];
> char port[NI_MAXSERV];
>
> those constants should be used in the call to getnameinfo() itself too
> (dropping the sizeof(char) multiplications is safe too).
>
> Based on the comments in the patch, it chooses returns service-names
> where available (ie registered in /etc/services or whatever). Usually
> this confuses users, but I'll leave that up to you guys as to whether
> you want this or not. If not, ORing the last argument to getnameinfo
> with NI_NUMERICSERV.
>
> The lines;
>
> + getaddrinfo("localhost", port, &hints, &res);
> + tp->hosts[0] = res;
>
> Don't really gel with what RFC3493 has in mind. Also by using the name
> it depends on name-services working, which is a new dependency.
> IN6ADDR_LOOPBACK_INIT may be better choices for this INADDR_LOOPBACK,
> but that would considerable change to how the hosts array is managed.
>
> There are two other things I thought I'd mention. One is how the
> max-hosts stuff is done, eg;
>
> if(tp->nhosts == TRANSPORT_MAX_HOSTS) {
>
> the patch makes hosts an array of linked-lists (which is what the
> addrinfo structures are). When for example a DNS record with multiple A
> records is used, the linked list will be expanded to that list of
> records. So, even though the array may contain less than
> TRANSPORT_MAX_HOSTS entries, there may be more actual hosts than that
> stored overall, and hence more attempted connections.
>
> This has a particularly strong effect on _randomize_hosts, which
> basically becomes pointless for multi-entry DNS RRs. To keep traditional
> behaviour, there really needs to be more changes to how hosts are
> stored. Either store the hosts as a single large linked list, or a
> single large array, but not a mixture of both :-)
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.
