If the interface already exists, the compat code would fail. Fix that
and clean up the function. Also double-check that an existing VLAN
interface that has the name we expect also has the master and VLAN
ID we expect.
libnl2 and earlier do not implement rtnl_link_get_kernel() and
thus we need compat code to determine whether an interface is
a bond or a VLAN. Previously, the VLAN code would simply assert
and cause NM to exit when running with libnl2 or earlier because
the interface type could not be determined.
The enum generator stuff apparently includes any #include in source or
header files that it's given, which was causing nm-enum-types.c to
include files from the Intel WiMAX SDK even when WiMAX wasn't enabled.
We don't use the default dnsmasq directory because packages often drop files
there that don't take account of NM's specific use-case and end up conflicting
with the specific local caching nameserver functionality that NM uses dnsmasq
for. NM's private dnsmasq is orthogonal to whatever global dnsmasq
daemon may be running, and with that daemons configuration.
(dcbw: change directory to private one)
If bluez is started by systemd but for some reason is not set to
be D-Bus activated (as seems to be the case on Fedora 16 and later),
then don't emit a warning.
NM was requiring that bond slaves have either no IP config or an
explicit "none"/"disabled" config. But the system scripts just ignore
any IP config that is present on a slave, so change NM to do that too
(but warn about it).
https://bugzilla.redhat.com/show_bug.cgi?id=838907
Unbreak 'make -j' by declaring the dependency that exists between the
two generated vapi files (forcing building of the second one to wait
until after the first one has been built).
https://bugzilla.gnome.org/show_bug.cgi?id=680374
If a class implements init_async, it should implement init_finish too,
rather than assuming the default implementation will do the right
thing (which it briefly didn't in glib 2.33).
The whole callout is pure dbus, not dbus-glib, so there's really no need
for glib code there. Even so, DBUS_LIBS contains dbus-glib and gobject,
so we're still linking to glib/gobject.
The DNS change frequency reduction patches mistakenly changed the signature
of the VpnStateChanged signal. Fix that, since we try really really
hard not to break the D-Bus API in stable branches. My bad...
