GitLab

As per review, don't use self._session for the check to see if there is
an existing message queue running. Rather, use a localally scoped
variable that will get cleaned up when the function exits.

Check to ensure there is no existing daemon running before removing
the socket endpoint. Fail to start if one is. Added a new test to
check this behaviour.

Put the traceback handler in the top level of all the python daemons and
some utility programs. The programs that don't do logging are not
included.

And adjust the main code so it is possible to test it.

Remove the code that used poll and kqueue, inserted sockets there and
such. We just use select not (and create the list of sockets on the
spot).

Remove the outdated tests. We don't do read & write in the same
iteration any more.

The spec file is not generated, so use B10_FROM_SOURCE, not
B10_FROM_BUILD.

Make sure msgq does not send notification when a client tries to
unsubscribe from a group it did not subscribe to.

Also, make sure we don't block long time on a function holding a mutex.
We use select for that.

Return a list of groups and instances the socket was unsubscribed from
from the subscription manager's unsubscribe_all.

Needed in follow-up work.

Test that notifications for unsubscription are sent even in case a
client disconnects without explicit unsubscription.

It'll be needed in the following work.

mostly in documentation or in comments

reviewed by muks via jabber

includes output changes for two isc_throw.

Convert the msgq to use the protocol constants instead of hardcoded
magic strings.

Don't put the resetUnitTestRootLogger in production program. This was
just a leftover from some debugging.

The success cases of sending, receiving, subscribing, etc.

reviewed by jelte via jabber.

Changes some log message descriptions and some
function arguments and some output.

Do a hack to make the tests run with separate build and source
directories. The generated file ended up in a different directory than
the rest of the modules and the __init__.py. It therefore didn't work.
We use the same trick as log_messages, a forwarder module and addition
to PYTHON_PATH to find it.

Only the real file is installed.

If an EPIPE is received in attempt to send a message, consider it
there's no recipient. This is still not perfect, since there may be data
in queue and before the socket is handled, it would be just appended.
Or, if the other socket is closed before the whole queue is sent there,
the messages are silently blackholed. But fully tracking each message
would be too much work for this ticket.