GitLab

As agreed, we don't provide the authority section for the original
DNAME/CNAME, only for the last target. As we don't chain, we don't
provide it at all.

It seems the RFC 2672 isn't too clear about it, but all checked
authoritative servers don't provide the authority section in case of
YXDOMAIN caused by too long synthesized name.

Return YXDOMAIN instead.

Conflicts:
	src/bin/auth/query.cc

in a function call that triggered typeerror exception.
okayed on jabber.

points to be considered with chaining.

Getting the NS returned delegation, it checked only if the callback is
enabled there and it was enabled because of DNAME.

The tests worked, but the timeouts were so small that on somewhat slower
systems the tests could fail. Increased them a bit.

The tests fail on assert in zonecutCallback currently.

In the previous commit I resolved a few merge conflicts, but too
hastily, this fixes 'em good

Just empty test bodies now.

Conflicts:
	src/bin/resolver/resolver.cc
	src/lib/asiolink/asiolink.cc
	src/lib/asiolink/asiolink.h
	src/lib/asiolink/tests/asiolink_unittest.cc

They are almost the same right now, but they will be different once the
CNAME chaining is implemented.

directly commiting to master, skipping review.

trivial doc fix, skipping review.

Also made it more portable not to use select.KQ_* until the add/delete
methods are actually called.

No chaining is implemented. While RFC 1024 indicates we should do it at
last inside the same zone, we don't now and leave it out for later.

The CNAME is handled in the same way as as SUCCESS with slight
modification about additional data (which probably never happen in real
life anyway).

No chaining yet.

* Fixed a comment
* Modified condition in MockZone to match the one in MemoryZone.

* kqueue maybe doesn't survive fork on MacOS
* add_kqueue_socket wanted socket, not fileno
* Some systems don't have MSG_DONTWAIT, so we set the socket nonblocking