On Tue, 26 Jul 2011 18:05:57 +0300, Tsantilas Christos wrote:
On 07/26/2011 02:01 AM, Amos Jeffries wrote:
On Mon, 25 Jul 2011 17:59:48 +0300, Tsantilas Christos wrote:
I am sending a version 4 of the patch which is the same with Amos
patch but is a little smaller. IT is easier to see the changes.
But applying this patch I found some problems in icp* related code.
This problems has as result an assertion failure inside icp code.
I am attaching a patch which is required to allow icp works with
isAnyAddr patch, but maybe fixes bugs.
From what I can understand reading about the icp protocol, it
supports only the IPv4 protocol: It contains in icp messages header
a
32bit number which represents the ip address of sender. This is
mean
that only ipv4 addresses can used.
The icp-isanyaddr.patch:
1) Use the ip::Address::SetIPv4 method to always convert to IPv4
the
ip addresses used to set the hostid field of the ICP header
message.
2) Inside neighbors_init() function there is the line:
theIcpPublicHostID.GetInAddr( *((struct in_addr*)&echo_hdr.shostid)
);
The above line sets the hostid field of the header of the icp echo
message, to theIcpPublciHostID. The problem is that this ipaddress
is
not initialized yet. Will be initialized inside
icpIncomingConnectionOpened when this function called.
The icp-isanyadd.patch replaces the above line with a simple:
echo_hdr.shostid = 0;
which is absolutely equivalent (but probably not the correct).
The icp-isanyaddr.patch is not a finished patch but I am posting it
here to show two (possible) problems:
- The icp does not work with IPv4 so we may add checks to prevent
using ipv6 addresses
- The echo_hdr message initialized with an IP address which is not
initialized yet.
What you have so far on the ICP looks correct.
OK I post a separate patch for icp.
Is it OK to commit the isAnyAddr patch to trunk?
Hmm, of itself its ready. But I'm not inclined to +1 it until the ICP
patch is in.
I think that is not many minutes work away though.
Amos
