Jelte Jansen [Tue, 27 Mar 2012 12:29:24 +0000 (14:29 +0200)]
[master] forgot CLEANFILES entry in merge of #963
this should fix the distcheck failure reported in http://git.bind10.isc.org/~tester/builder/BIND10-distcheck/20120327113000-FreeBSD8-i386/logs/distcheck.out
It's shocking how much is not tested here. No surprise the missing
command was not discovered. Still, not adding tests for the others,
they'd probably fail and it would make the ticket a lot bigger.
Jelte Jansen [Tue, 27 Mar 2012 08:41:49 +0000 (10:41 +0200)]
[1491] copy default named-set data to local when editing
if a child member is set, and the named-set data is default, it would 'lose' the other default values if it wasn't copied to local. For succesful operations this wouldn't be a problem (as this only happens 'locally' in bindctl), but it can cause problems when updates fail due to mistyped data
JINMEI Tatuya [Mon, 26 Mar 2012 23:36:21 +0000 (16:36 -0700)]
[1772] used ::1 instead of 127.0.0.1 2nd bind10 instance and the example test.
127.0.0.1:47807 is used by a MacOS application and the conflict makes the
test fail in environments that uses the app. Using an IPv6 address is not
a cleanest solution, but unless we make it fully configurable we need to
handle this type of thing by some workaround anyway. Also, I believe
it's good to use IPv6 transport for some tests in any event.
JINMEI Tatuya [Mon, 26 Mar 2012 17:39:14 +0000 (10:39 -0700)]
[1836] also added unittest that would fail the fixed in this branch.
To make this test workable, I needed to change the returned flag
for the NXRRSET case in MockZoneFinder::find() from setting NSEC_SIGNED
unconditionally to setting any SIGNED flags. this doesn't affect other
tests.
Jelte Jansen [Fri, 23 Mar 2012 11:19:59 +0000 (12:19 +0100)]
[963] improve exit status
0 on success (or upgrade aborted by user)
1 on check and need update
2 on check and version of db too high
3 on command-line error
4 on db read error
5 on upgrade error
JINMEI Tatuya [Fri, 23 Mar 2012 00:18:52 +0000 (17:18 -0700)]
[1579] suggested change for an awkward comment line:
- s/NSEC3PARAMETER/NSEC3PARAM/
- "definitely" is too strong in this context
- add a space after a comma
- "zonefile" is awkward (there's no "zonefile" in this context)
- "NSEC3 method" is an awkward term
- "this zone signature" is also awkward
JINMEI Tatuya [Thu, 22 Mar 2012 23:35:40 +0000 (16:35 -0700)]
[1775] ensure wildcard expanded RRsets are considered of same kind if they are.
in the previous version, if one is returned from find() and the other is
from getAdditional() they are not recognized as of the same kind, because
the underlying pointers are different.
JINMEI Tatuya [Thu, 22 Mar 2012 21:52:49 +0000 (14:52 -0700)]
[1775] handled corner cases with wildcard expansion: empty or multiple nodes.
The previous version crashes if the wildcard node is empty. Also, the
previous version didn't work well if there were multiple wildcard matches
because the auxiliary tree changes as we add more nodes and stored
pointer can become invalid. The fix to this issue is a bit complicated;
I needed to make the process two-stage.
JINMEI Tatuya [Thu, 22 Mar 2012 17:13:04 +0000 (10:13 -0700)]
[1688] (unrelated cleanup) make RESERVE_RRSETS a private member of Query class.
This will prevent it from causing definition bloat even if the linkder
doesn't combine the constant def. It would also make more sense to make
it inaccessible from others anyway, because it's basically an
implementation-internal constatnt.
Jelte Jansen [Thu, 22 Mar 2012 14:36:43 +0000 (15:36 +0100)]
[1443] Addressed review comments
- if self.clear_config is None was wrong, should just be if self.clear_config
- 'back up' should have been 'create backup of'
- default value for unused parameter rename_config_file in b10-cfgmgr_test.py.in MyConfigManager should have been False, not None
- reordered message file
JINMEI Tatuya [Thu, 22 Mar 2012 06:54:04 +0000 (23:54 -0700)]
[1688] re-enabled lettuce tests that failed due to missing dup suppression.
note: there's an error in the test description itself. In "7.2.2 other",
the result should be NOERROR, not NXDOMAIN, because it should match
*.w.example (BIND 9 indeed returns NOERROR, for that matter). I fixed
this, then all tests passed.
JINMEI Tatuya [Thu, 22 Mar 2012 06:33:23 +0000 (23:33 -0700)]
[1688] use BOOST_FOREACH instead of dedicated for loops.
this could be slightly more efficient because it avoids redundant calls
to vector::end(). In my experiments I actually didn't see much difference,
but it doesn't harm anyway.
JINMEI Tatuya [Thu, 22 Mar 2012 04:45:00 +0000 (21:45 -0700)]
[1688] suggested reorg: merge response building to the inserter class.
that way we don't have to expose mutable member variables of `Query`.
now the responsibility of the "inserter" is enlarged, also renamed it
to represent the new job more appropriately.
JINMEI Tatuya [Thu, 22 Mar 2012 04:06:52 +0000 (21:06 -0700)]
[1688] suggested cleanup: use vector::insert instead of copy().
also avoid referring to an invalid range of iterator with operator+; use
end() instead (although the original code was probably safe at least in
practice).