I applied the two patches related to this matter. I only renamed the structure you defined, stripping '_t' from its name as this suffix is commonly used across the code as a typedef'ed struct name.
Cheers, Daniel
On 10/18/13 10:01 AM, Øyvind Kolbu wrote:
On 17.10.2013 19:09, Daniel-Constantin Mierla wrote:
looking at the first patch, I see that PROTO_NONE was removed from those switch statements on proto variables. I guess it was the reason of the issue, as you pointed it out, that a non-existing udp srv records results in failure. Isn't it?
No, the reason was that it can never be true, as just before *proto is set to PROTO_UDP if not set.
if (proto){ /* makes sure we have a protocol set*/ if (*proto==0) *proto=srv_proto=PROTO_UDP; /* default */ else srv_proto=*proto; }else{ srv_proto=PROTO_UDP; }The problem with the old code was that, unless you specified a protocol, it would only attempt to search up _sip._udp and not continue with _sip._tcp and possible other protocols.
To be sure I got properly the new logic, if proto var is PROTO_NONE, then create_srv_pref_list(..) will build a list of protocols to try based on config preferences. The it will be a dns lookup of the values in the list. Right?
Yes, if the original *proto is PROTO_NONE, then call create_srv_pref_list and then stop on the first match in the generated list.