GitLab

There is a new function used in keactrl tests which waits for the server
to shutdown, rather than check if it is down already. This is to avoid
race condition when the process may not completely terminate when we
check that it is.

distcheck was failing to tests/dhcp(4/6)_process_tests.sh.
Since these files are now constructed from a .in file,
they must be pathed in abs_builddir.
dhcp(4/6)/tests/Makefile.am updated.

 - dhcp6_test.py removed
 - src/lib/testuitls/testdata/*.wire files are now in repo,
   no longer need python3 script to generate
 - src/lib/cc/proto_defs.h added to repo, no longer autogenerated
 - Added comment to get-rdatacode.py
 - removed LIBRARY_PATH_PLACEHOLDER hacks from several Makefiles.
 - ChangeLog entry added.

Also, a few corrections in Makefiles.

Starting lease manager in the server constructor implied the use of the
default memfile for a short period until the server is configured. The
default memfile configuration used the persistence (writing leases to disk)
which in many cases required root privileges. This caused some unit tests
to fail.

 - ChangLog updated (b10-dhcp4 => DHCPv4 server)
 - stand-alone mode removed in Kea4,Kea6
 - getInstance() used instead of server_
 - ctrl_dhcp4_srv.cc: one stringstream instance now shared by all error messages
 - ctrl_dhcp4_srv.h: comments cleaned up
 - Added log message for received configuration
 - kea_controller.cc: Unnecessary header/using namespace removed
 - *.json files are now cleaned up in src/bin/dhcp4/tests
 - added comment that explains lack of tests for Bundy backend
 - kea_controller_unittest.cc now uses the same filename everywhere

Changes also applied to Kea6 (where applicable).

The goal of these tests is to exercise error handling of a failed NCR send
in Kea4/Kea6.  Because the test is for a UDP sender is there is no simple way
to make the send fail.  The only way found so far is to attmpt to send it to
0.0.0.0/port 0.  This fails everywhere except Solaris 11. For now the tests
have been disabled for Solaris 11