BIND 10 #2887: "make check" fails on FreeBSD 9.1

[BIND 10 Development]

#2887: "make check" fails on FreeBSD 9.1
-------------------------------------+-------------------------------------
            Reporter:  jwright       |                        Owner:
                Type:  defect        |                       Status:  new
            Priority:  medium        |                    Milestone:
           Component:  tests         |  Sprint-20130423
            Keywords:                |                   Resolution:
           Sensitive:  0             |                 CVSS Scoring:
         Sub-Project:  DNS           |              Defect Severity:
Estimated Difficulty:  3             |  Medium
         Total Hours:  0             |  Feature Depending on Ticket:
                                     |          Add Hours to Ticket:  0
                                     |                    Internal?:  0
-------------------------------------+-------------------------------------

Comment (by jinmei):

 anyway, trac2887 is ready for review.

 I've committed to the branch some non controversial fixes and one
 workaround (for a python test).  The latter may be a bit
 controversial, but I thought it's acceptable since (at least for now)
 that's the only case we need to do this kind of hack.

 Tests that still fail are as follows:

 - RRsetTest.addRdata
 - Rdata_TSIG_Test.compare
 - Rdata_MINFO_Test.compare
 - Rdata_SRV_Test.compare
 - Rdata_RP_Test.compare
 - Rdata_NSEC3PARAM_Test.compare
 - Rdata_NSEC3_Test.compare
 - Rdata_NSEC_Test.compare
 - Rdata_DS_LIKE_Test/*.compare
 - Rdata_DHCID_Test.compare
 - Rdata_OPT_Test.compare
 - Rdata_AFSDB_Test.compare
 - Rdata_MX_Test.compare
 - Rdata_TXT_LIKE_Test/*.compare
 - Rdata_IN_AAAA_Test.compare
 - Rdata_IN_A_Test.compare
 - (in datasrc) RdataSerializationTest.badAddRdata
 - (in auth) QueryTestForMockOnly.nxdomainBadNSEC1

 These fail because they explicitly try to check `std::bad_cast`.  We
 could introduce even more hack to work around these, but I'd
 personally rather avoid tweaking our code in these cases.  It's only a
 matter of how we write tests and only for a very limited environment.

 In any case, again, I strongly suggest avoiding FreeBSD 9.1 for our QA
 machine, but until we update/change the OS/version, I suggest simply
 filtering out these tests using an environment variable:

 {{{
 setenv GTEST_FILTER
 '-RRsetTest.addRdata:Rdata_TSIG_Test.compare:Rdata_MINFO_Test.compare:Rdata_SRV_Test.compare:Rdata_RP_Test.compare:Rdata_NSEC3PARAM_Test.compare:Rdata_NSEC3_Test.compare:Rdata_NSEC_Test.compare:Rdata_DS_LIKE_Test/*.compare:Rdata_DHCID_Test.compare:Rdata_OPT_Test.compare:Rdata_AFSDB_Test.compare:Rdata_MX_Test.compare:Rdata_TXT_LIKE_Test/*.compare:Rdata_IN_AAAA_Test.compare:Rdata_IN_A_Test.compare:RdataSerializationTest.badAddRdata:QueryTestForMockOnly.nxdomainBadNSEC1'
 }}}

 And, finally, I guess we need a changelog entry.  Proposal:

 {{{
 599.?   [bug]           jinmei
         Worked around some unit test regressions on FreeBSD 9.1 due to
         a binary compatibility issue between standard and system
         libraries (http://www.freebsd.org/cgi/query-pr.cgi?pr=175453).
         While not all tests still pass, main BIND 10 programs should
         generally work correctly.  Still, there can be odd run time
         behavior such as abrupt crash instead of graceful shutdown
         when some fatal event happens, so it's generally discouraged to
         use BIND 10 on FreeBSD 9.1 RELEASE.  According to the above
         bug report for FreeBSD, it seems upgrading or downgrading the
         FreeBSD version will solve this problem.
         (Trac #2887, git TBD)
 }}}

-- 
Ticket URL: <http://bind10.isc.org/ticket/2887#comment:17>
BIND 10 Development <http://bind10.isc.org>
BIND 10 Development