Changes needed to roll libjingle r141

jingle/notifier/base/chrome_async_socket.cc

Index: jingle/notifier/base/chrome_async_socket.cc
===================================================================
--- jingle/notifier/base/chrome_async_socket.cc	(revision 136430)
+++ jingle/notifier/base/chrome_async_socket.cc	(working copy)
@@ -105,7 +105,8 @@
 
   DCHECK_EQ(false, weak_factory_.HasWeakPtrs());
 
-  net::HostPortPair dest_host_port_pair(address.IPAsString(), address.port());
+  net::HostPortPair dest_host_port_pair((address.hostname().empty() ?

[akalin, 2012/05/11 18:04:05]

this should actually be:

... dest_host_port_pair(address.hostname(), address.port());

as the IP isn't expected to be resolved at this point -- the proxy resolution
process does the DNS resolution also. (Unfortunately, this looks like a
long-standing bug :/ )

I think that should fix the crashes.  I'm not sure if the other uses of address
suffer from the same proxy-related issue.

[Mallinath (Gone from Chromium), 2012/05/11 18:10:40]

What should happen if hostname is empty? As talk_base::SocketAddress can contain
empty hostname and valid IP.
On 2012/05/11 18:04:05, akalin wrote:

[akalin, 2012/05/11 18:13:02]

Actually, this doesn't look like it was a bug before.  It looks like
IPAsString() printed the hostname first.

But yeah, definitely address.hostname() shouldn't be empty and should be used.

[Mallinath (Gone from Chromium), 2012/05/11 20:52:45]

Done.

+      address.ipaddr().ToString() : address.hostname()), address.port());
 
   transport_socket_.reset(
       client_socket_factory_->CreateTransportClientSocket(