[Live-devel] ip 0.0.0.0 in rtps url

[Ben Lear]

Just a little comment inline;

>>When we run test*Server the rstp URL we are given tells us to connect 
>>to ip 0.0.0.0? If we check ifconfig everything looks allright, 
>>connecting over ssh to the ip from ifconfig works fine too.
>>
>>What can be the problem that makes the test servers think the 
>ip of our 
>>robot is 0.0.0.0 and how can we fix this?
>
>The code uses the function "ourSourceAddressForMulticast()" (in
>"groupsock/Groupsock.cpp") to find its IP address.  It does 
>this by sending a multicast packet, then receiving it, and 
>looking at the IP source address.  Somehow this is failing.
>
>It appears that you do not have multicast configured correctly 
>in your kernel, as illustrated by:
>
>>Unable to determine our source address: setsockopt(IP_ADD_MEMBERSHIP)
>>error: No such device
>
>Make sure that you have a route for "224.0.0/4" (to "eth0").

It appears that the intent of the function "ourSourceAddressForMulticast" is
to first attempt to use a multicast loopback method to determine the local
IP, else if this fails use "gethostname". A problem in the test logic *may*
exist in that if the multicast WRITE fails then the second method(calling
"gethostname") is skipped, though if the multicast READ fails then
"gethostname" is used as expected.

Would it not be better to try "gethostname" regardless of what
failed(Write/Read) ? Maybe in the case of multicast failure something could
be flagged and/or a warning message stating multicast failed/not available..
whatever. The reasoning is that multicast is not required for *all*
implimentations and it seems unreasonable to fail when it is not available.

Cheers,

Ben.