Icecast dev - Dec 2004 - listen to port on ipv6 by default[patch]

right now icecast2 listens to 0.0.0.0 by default if bind-address is not
defined. it should rather listen to :: if ipv6 is available.
the change is rather small. i do not know how that works on systems that
do not support IPv6 yet. 
so i added #ifdef AF_INET6, is that the right way to check for ipv6?

attached is a patch against http://svn.xiph.org/icecast/trunk/net/

if it is ok like that it would be good to apply it for the next release.

j
-------------- next part --------------
A non-text attachment was scrubbed...
Name: icecast2-listentoipv6_by_default.patch
Type: text/x-patch
Size: 702 bytes
Desc: not available
Url :
http://lists.xiph.org/pipermail/icecast-dev/attachments/20041210/9e81ccbf/icecast2-listentoipv6_by_default.bin

On Saturday 11 December 2004 07:56, j@v2v.cc wrote:
> right now icecast2 listens to 0.0.0.0 by default if bind-address is not
> defined. it should rather listen to :: if ipv6 is available.
> the change is rather small. i do not know how that works on systems that
> do not support IPv6 yet.
> so i added #ifdef AF_INET6, is that the right way to check for ipv6?
>
> attached is a patch against http://svn.xiph.org/icecast/trunk/net/
>
> if it is ok like that it would be good to apply it for the next release.
>
> j
This doesn't look right to me. This makes the runtime behaviour depend on 
compile-time availability of AF_INET6. 

Also, I'm not sure if the runtime behaviour is correct - I'm not that
familiar
with IPv6. If you have a v6 socket open, will it accept v4 connections?

In short, a more complex patch would be needed - at the least, it'd need to 
query availability of ipv6 sockets at runtime (as well as having the 
compile-time checks for the appropriate functions/constants/etc.).

Mike

On Mon, 2004-12-13 at 10:10 +1100, Michael Smith wrote:
> This doesn't look right to me. This makes the runtime behaviour depend
on
> compile-time availability of AF_INET6. 
right, HAVE_INET_PTON might be ok than, since AF_INET6 is used in the 
#ifdef HAVE_INET_PTON section before.
> Also, I'm not sure if the runtime behaviour is correct - I'm not
that familiar
> with IPv6. If you have a v6 socket open, will it accept v4 connections?
it will, applications listening on ::, listen on all interfaces ipv4 and ipv6.
> In short, a more complex patch would be needed - at the least, it'd
need to
> query availability of ipv6 sockets at runtime (as well as having the 
> compile-time checks for the appropriate functions/constants/etc.).
as AF_INET6 is needed with PTON i am not sure if that is true.
the open question is does 
((struct sockaddr_in6*)&sa)->sin6_addr = in6addr_any;
((struct sockaddr_in6*)&sa)->sin6_family = AF_INET6;
((struct sockaddr_in6*)&sa)->sin6_port = htons(port);
work on ipv4 only if the kernel has no ipv6 support, i will try to find
out, but have not seen a kernel without ipv6 in a long time...

j