BIND 10 #1462: check SOA serial in xfrout

[BIND 10 Development]

#1462: check SOA serial in xfrout
-------------------------------------+-------------------------------------
                   Reporter:         |                 Owner:  jinmei
  jinmei                             |                Status:  reviewing
                       Type:  task   |             Milestone:
                   Priority:  major  |  Sprint-20120110
                  Component:         |            Resolution:
  xfrout                             |             Sensitive:  0
                   Keywords:         |           Sub-Project:  DNS
            Defect Severity:  N/A    |  Estimated Difficulty:  3
Feature Depending on Ticket:  AXFR-  |           Total Hours:  0
  out                                |
        Add Hours to Ticket:  0      |
                  Internal?:  0      |
-------------------------------------+-------------------------------------

Comment (by jinmei):

 Replying to [comment:10 kevin_tes]:

 Shane already reassigned the ticket back to me (as expected), but in
 case you didn't know it rather than forgetting it: in general, if you
 have addressed review comments the next step is to give the ticket
 back to the original reviewer.  If you change it to "unassigned", the
 ticket could be dangling.

 > > - before doing that, add a test case that fails due to the integer
 > >   comparison (e.g. 0xffffffff "<" 1).  see, again, xfrin.  Then make
 > >   the above change, and then confirm it fixes the problem.
 > I think this case for a while,and i think it does not to add this test.
 > The first reason:The CURRENT_SOA_SERIAL=2011112001,SOA=2011119000 can
 test this case,and isc::dns::Serial has such test too.
 > If add a test case that fails due to the integer comparison (e.g.
 0xffffffff "<" 1), i think it needs to change the  CURRENT_SOA_SERIAL,I
 think it does not need to change this exited codes too.

 I personally don't think it (skipping the explicit test) a good idea.
 First, if I understand it correctly, the test case with
 CURRENT_SOA_SERIAL=2011112001, SOA=2011119000 cannot detect an error
 that integer comparison is incorrectly used instead of comparison by
 Serial objects.  Second, the main purpose of using a test case like
 0xffffffff "<" 1 is not for testing the Serial object works correctly
 but to check we really use Serial objects instead of integers and
 really compare serials as Serial objects not as integers.  It will
 also detect if and when introduce a regression we accidentally
 reintroduce the same type of bug in future.  So the fact that it's
 tested in the Serial class implementation cannot be a reason for
 skipping it here.

 Also, I didn't suggest changing the definition of CURRENT_SOA_SERIAL,
 etc.  For the test like 0xffffffff "<" 1 I expected to introduce
 another constant (or hardcoded values if they are only used in that
 specific tests).

 > > - ixfr-to-axfr test cases shouldn't be removed (we need to _SERIAL the
 > >   test parameters and check the fallback behavior, not just remove
 it).
 > > - I believe we need a changelog for it.
 > I have recovered this test and changed the test parameters.

 I don't see the recover of the ixfr-to-axfr (and axfr-style of ixfr)
 test in the latest revision (e99849d).  Perhaps you forgot pushing
 your local change?

 Please also show proposed changelog entry.

-- 
Ticket URL: <http://bind10.isc.org/ticket/1462#comment:13>
BIND 10 Development <http://bind10.isc.org>
BIND 10 Development