and use the holder in the unit tests as well

Half-related change, if available, call mysql_library_end() at the end of the unit test run

the mysql lease manager wouldn't close the mysql database immediately if initialization failed at certain points.

This proposed fix is not entirely complete; it would probably be better to either use such a holder class as the member, or to slightly clean it up so that the return value from release() can be given to mysql_ (and therefore use the holder in the construction code instead of mysql_), but I did not want to touch too much existing code at this point.

should be trivial, so directly committing.

will fix distcheck failure on some platforms.  okayed on jabber

also point to wiki for known issues for DHCP

Replacing pair of group_sendmsg & group_recvmsg by compound method rpc_call in
many cases. It's not possible to convert all cases, due to various hacks in the
code.

Conflicts:
	ChangeLog

statement is from stephen.

This is from http://bind10.isc.org/attachment/ticket/2742/0001-master-Remove-redundant-PYTHONPATH.patch
of ticket #2742

just happen to notice them in review, and since these files will be updated
this is probably a good opportunity to fix them with the branch.

Don't include example.zone test data file in dist tarball.
It is generated from an .in file via ./configure.

This was noticed by emhn during Debian packaging.
(distclean created a source different than extract tarball.)

I didn't get okay, but I mentioned my patch on jabber.
I tested on one system. This seems to be trivial.

the revised code eliminated its use as a side effect.

(noticed when another fix was suggested for same line)

If the remote is not there, it's not "Data error"

In the code before refactoring, when the CfgMgr wasn't there, it did a
SessionTimeout. Now the RPCRecipientMissing is raised instead, but as it
is subclass of RPCError, we need to exclude it from the error ignoring.

Actually, the error ignoring itself is very questionable thing, but it
is as it was before.

As the rpc_call is the only place the exception can be raised from, this
is equivalent, just cleaner.

According to the notes in the review, we need to wait for the answer.

Remove the part of the DDNS_UPDATE_NOTIFY_FAIL message description,
complaining about how bad the timeouts are. The timeouts should be
solved by now and should happen only in really rare circumstances. If
the other module is not there, it should return error right away.

In case the auth wasn't there before the refactoring, it did a session
timeout, which resulted in DDNS_(START|STOP)_FORWARDER_ERROR log
message. So we handle the RPCRecipientMissing in the same branch (and
list above the RPCError, since it is a subclass of that).