BIND 10 #2904: [kean] check violation of DO_NOT_USE_127.0.0.1:47807

[BIND 10 Development]

#2904: [kean] check violation of DO_NOT_USE_127.0.0.1:47807
-------------------------------------+-------------------------------------
            Reporter:  jinmei        |                        Owner:  kean
                Type:  defect        |                       Status:
            Priority:  medium        |  reviewing
           Component:  Unclassified  |                    Milestone:
            Keywords:                |  Sprint-20131015
           Sensitive:  0             |                   Resolution:
         Sub-Project:  DNS           |                 CVSS Scoring:
Estimated Difficulty:  2             |              Defect Severity:  N/A
         Total Hours:  0             |  Feature Depending on Ticket:
                                     |          Add Hours to Ticket:  0
                                     |                    Internal?:  0
-------------------------------------+-------------------------------------

Comment (by kean):

 Replying to [comment:11 muks]:
 > The change of all IPv6 configuration to IPv4 is also fine for the
 builders, but this appears as a regression. IIRC, ::1 was used in lettuce
 tests first to avoid a port problem on Jinmei's machine, but we have since
 then assumed that IPv6 is exercised by lettuce due to the presence of
 these tests.
 There are plenty of other tests that use IPv6. Just do a grep of ::1 and
 you will see.

 The IPv6 changes in 2904 are mostly to undo hacks like this: {{{git diff
 6db3556488ae8a75f5250a175f3507cc609ada19
 34da803fd804c6c0fa7acb90ff46780df69d5898}}}

 > My concern is that if we switch everything to IPv4 loopback address
 without adding corresponding IPv6 tests, we will have a time when we don't
 do any system testing with IPv6 addresses.
 I do not believe that is the case at all. This ticket does not remove all
 IPv6 tests by any means. But what it does also fix is stuff like this:

 {{{
 diff --git a/src/lib/nsas/nameserver_address.h
 b/src/lib/nsas/nameserver_address.h
 index 33a1169..1a8c4df 100644
 --- a/src/lib/nsas/nameserver_address.h
 +++ b/src/lib/nsas/nameserver_address.h
 @@ -74,7 +74,7 @@ public:
      }

      /// \brief Default Constructor
 -    NameserverAddress() : address_(asiolink::IOAddress("::1")) { }
 +    NameserverAddress() : address_(asiolink::IOAddress("127.0.0.1")) { }
 }}}

 Clearly, in that constructor, we don't really mean to be forcing the use
 of IPv6. While its not a supported platform right now, Windows frequently
 has IPv6 disabled completely. Other systems may as well. However they all
 have IPv4 enabled.

 >
 > What I suggest is that you make the port changes alone first on a
 branch, get it reviewed and merge it to master. This will be quick as the
 port change is OK.
 >
 > Then, separate the IPv6 testing by adding new lettuce testcases (such as
 those listed above) for the various components. As part of that ticket,
 cleanup the lettuce IP address configurations.
 The vast majority of 2904 is exactly that. Yes there are some changes to
 remove the IPv6 hack done by #1772 but the largest part of the changes are
 the port number changes.

-- 
Ticket URL: <http://bind10.isc.org/ticket/2904#comment:12>
BIND 10 Development <http://bind10.isc.org>
BIND 10 Development