GitLab

When the boss process was unable to connect to the msgq, it would
exit. However, it would leave the msgq process running. This patch
fixes that by killing the msgq if it is running in this case.

resolved Conflicts:
	src/lib/dns/python/Makefile.am
	src/lib/dns/python/rrset_collection_python.cc
	src/lib/dns/python/rrset_collection_python.h
	src/lib/dns/python/rrset_collection_python_inc.cc

Conflicts:
	ChangeLog
	src/bin/dhcp6/tests/dhcp6_srv_unittest.cc

The problem was, msgq started to open more file descriptors than before.
Also, the changes meant the tests didn't shut down their internal copy
of msgq properly, leaking more file descriptors.

* Do the shutdown correctly.
* Wait for msgq to terminate, so there are no race conditions between
  the old (shutting down one) and a new one for another test.
* Close all the data sockets in msgq, to prevent further leaks. The
  other ends probably still leak, but at least the threads on the other
  ends terminate on EOF, so we don't leak the threads too.
* Wait for the other threads too, so we don't spawn too many threads.
* Unrelated cleanup: remove unneeded output-suppression workaround from
  the tests, it is no longer needed.
* Unrelated cleanup: Initialize testing logger for the tests. But it
  still logs for no apparent reason, needs to be examined.

fixed Conflicts:
	src/lib/datasrc/zone_loader.cc

necessary for systest to pass on some environment

Conflicts:
	ChangeLog
	src/bin/dhcp6/config_parser.cc

- This will be moved to 2642.

Let the message queue connect to itself and handle config updates and commands.
No commands or configuration specific to Msgq does not exist yet, but it
handles the generic logging config.

- added include scoped_ptr.hpp
- removed empty destructor

it's just a matter of preference; I don't have a particular one.

so the caller can get highest possible precission.

as commented, #2145 should eliminate the need for it, but until it's fixed
this seems to be the easiest and smallest workaround.

that constructor is provided, but results in an exception.

there are some changes and files that don't exist the latest trac2435 branch.
they are probably because of the intermediate merge of a separate working
branch, which was based on an earlier version of trac2435.

Don't write to the same file at two different rules. That led to running
the two rules in parallel, writing there at the same time, confusing the
script that loaded it just after that.

As proposed by reviewer

Just comments and style. No functionality or code change.

resolved Conflicts:
	src/lib/datasrc/Makefile.am
	src/lib/datasrc/database.cc
	src/lib/datasrc/tests/database_unittest.cc
	src/lib/datasrc/tests/master_loader_callbacks_test.cc
	src/lib/datasrc/tests/zone_loader_unittest.cc
	src/lib/datasrc/zone.h
	src/lib/dns/rrset_collection_base.h