GitLab

there seems to be a portability issue without it.

This disables the test to receive data over a closed socket on BSD systems.
The 8.1-RELEASE version of FreeBSD has a bug which causes select() not to
return EBADF error code when invalid socket descriptor is provided. The
bug number is kern/155606.

The socket function returns socket descriptor on success and value of -1
on error. This implies that 0 is a valid socket descriptor and the unit
test which verifies it should treat 0 as success. Previous version of
IfaceMgr unit test expected positive result.

Conflicts:
	src/lib/dhcp/Makefile.am

it's described in the context of findZone() method.

added a note on the numbering.
and made one piggyback cleanup: define public static const explicitly.

The ETHERTYPE_IP is defined in different header files on different OSes.
Since it is just a single constant that is needed, it is not practical
to locate the proper header file in configure.ac to pull single constant.

the mock findZone() implementation was buggy and wouldn't work correctly
if there were actually more than one zone.  fixed it by porting
MockDataSourceClient::findZone() from datasrc/tests/mock_client.cc.

per style guideline.

this allows the caller to know whether a load error happens and log the
reason for it if it constructed the writer to catch load error exceptions
internally.  ConfigurableClientList::configure() needs it to log these
events.  the catch block in the method is now dead code and was removed.

Packet will be sent to broadcast address if broadcast flag is set by the
client. Also, DHCPNAK is always sent to broadcast address.

Fix CSV-related test for usermanager that failed on some versions of
python (seen on 3.2.4 on Gentoo).

...and will treat it as a broken (rather than nonexistent) zone for later
lookups.

okayed on jabber.

updated addZone() and findZone() behavior so they can handle the concept of
empty zones well.