Given the unintuitive way that flags are implemented for GstObject and
GstElement (ie. enums that specify a shift instead of the actual bitmask for the
flag), we should under no circumstances export an accessor API that gives access
to the flags bitmask. That is just begging for wrong usage like

  if ((GST_FLAGS(obj) & FOO_IS_BLA_FLAG)) {
    ...
  }

IMHO we should make sure that flags can only be tested with our own
GST_FLAG_IS_SET, GST_FLAG_SET and GST_FLAG_UNSET. For the same reason, the
'flags' member of GstObject should be moved into the private block IMHO.

Cheers
 -Tim